The Smart AI-based Pregnancy MonitoringSystem

1.0 Introduction
Pregnancy and childbirth are crucial stages in maternal and child health that require careful
monitoring to ensure healthy outcomes for both mother and baby. However, access to
adequate antenatal care remains a challenge in many parts of the world including Zimbabwe.
This research project aims to address this gap through the development of an innovative
Smart AI-based Pregnancy Monitoring System.

1.1 Background of the study

In Zimbabwe, as in many other low-resource settings, traditional pregnancy monitoring
practices often rely on periodic clinic visits which typically involve measuring maternal vitals
such as blood pressure, weight, and urine protein levels during antenatal check-ups
(Addanke&Anandan, 2024). Additionally, ultrasound scans are used to monitor fetal growth
and development, assess placental function, and detect any abnormalities (Ettiyan&Geetha,
2023). These intermittent visits may not capture real-time changes or early warning signs of
potential complications, leading to delays in intervention and treatment (Marques et al.,
2021).According to UNICEF Data (2023) around 4 women pregnant women die giving birth
due to unforeseen risks.

Access to comprehensive healthcare services and regular monitoring during pregnancy

remains a significant challenge in numerous low-resource settings, including parts of
Zimbabwe (Sandra V. B. Jardim, 2013). Various factors like geographical constraints,
transportation difficulties, and a shortage of skilled healthcare professionals contribute to
deficiencies in prenatal care (ORAL PRESENTATION, 2011). These gaps in services can
result in delayed identification of conditions like preeclampsia, gestational diabetes, fetal
distress, or premature labor, posing significant risks to the health of both mothers and
newborns. Moreover, traditional pregnancy monitoring systems rely heavily on subjective
assessments and manual record-keeping, which can be prone to errors and
inconsistencies(Romero-Lopez et al., 2020). The reliance on paper-based documentation also
makes it challenging to track trends, analyze data comprehensively, and provide timely
interventions when needed, these limitations underscore the need for innovative approaches

1|Page
to pregnancy monitoring that can overcome geographical, logistical, and technological
barriers(Tippenhauer et al., 2023).
The WHO (2022) recommends a minimum of four antenatal care appointments for every
pregnant woman to monitor vital signs like blood pressure, weight, fetal development, and to
detect potential threats to health such as preeclampsia and infections. This regular monitoring
is crucial because approximately 30% of maternal deaths worldwide stem from severe
bleeding, infections, hypertensive disorders, obstructed labor, and unsafe abortion practices
(WHO, 2015). Adequate surveillance empowers healthcare professionals to take timely
action in case of complications.

In Zimbabwe, the maternal mortality rate remains a critical concern, with four women dying
during childbirth, highlighting the pressing need for improved healthcare infrastructure and
access to quality maternal care services solutions (HakiemAfrizal et al., 2022). Factors
contributing to these tragic outcomes include limited access to skilled birth attendants, delays
in seeking medical assistance, inadequate prenatal care, and challenges in managing obstetric
emergencies stakeholders (Halim Omar et al., 2023). Addressing these issues requires a
multifaceted approach, including investments in healthcare facilities, training of healthcare
personnel, community education on maternal health, and ensuring timely access to essential
obstetric services(Testa et al., 2022). Reducing maternal mortality rates is not only a
healthcare imperative but also a social and economic priority, as it directly impacts family
well-being and overall societal development.

The emergence of digital health technologies and the Internet of Things (IoT) has paved the
way for more advanced and connected pregnancy monitoring(Diaz et al., 2023). Wearable
devices, such as smartwatches and biosensors, can now capture a wide range of physiological
data in real time which include maternal heart rate, fetal heart rate variability, uterine
contractions, activity levels, sleep patterns, and even maternal emotional well-
being(Valakunde et al., 2017). Furthermore, mobile applications and cloud-based platforms
enable remote monitoring and data sharing between pregnant individuals, healthcare
providers, and other(Khozouie&Malekhoseini, 2024). This connectivity allows for
continuous surveillance of maternal and fetal health parameters, early detection of anomalies,
and timely interventions or referrals as neededAI algorithms integrated into these systems can
analyze vast amounts of data, identify patterns, and generate actionable insights to support

2|Page
clinical decision-making(Shulman et al., 2018). Therefore, they is need of developing a Smart
AI-based Pregnancy Monitoring System.
1.3 Problem Statement
Access to adequate antenatal care remains a challenge in Zimbabwe. Lack of sufficient
antenatal care can negatively impact maternal and child health outcomes. According to the
World Health Organization (WHO), complications during pregnancy and childbirth are
among the top causes of illness and death for mothers and babies worldwide each year
(WHO, 2022). In Zimbabwe specifically, maternal mortality ratios are high at 462 deaths per
100,000 live births (Zimstat, 2020). A key factor contributing to this is low rates of
supervised childbirth. Meldrum (2019) found that less than 81% of births in Zimbabwe are
attended by skilled health workers. This is significantly lower than the recommended WHO
threshold of at least 90% supervised deliveries (WHO, 2022). Rural regions face even greater
disparities, with only 61% of births assisted by nurses or midwives (Zimstat, 2020).
Weaknesses in antenatal care provision have been attributed to shortages of healthcare
professionals. Mutsambiwa (2019) noted Zimbabwe has a doctor-to-population ratio of only
1:6,427, below the WHO standard of 1:1,000. Budget constraints have also limited
infrastructure investments and expansions needed to meet growing population needs
(PhilipaMladovsky, 2022). There is therefore a need for innovative solutions to improve
access to quality antenatal care services, facilitate closer monitoring of at-risk pregnancies,
and enable early detection and treatment of complications in Zimbabwe. A well-designed
mHealth intervention could help address this problem.

1.3 Aim
The overall aim of the research is to develop a Smart AI-based Pregnancy Monitoring System
to improve antenatal care access and outcomes for expectant mothers in Zimbabwe
1.4 Objectives
The specific objectives are:
[1)] To Aallow users to insert upload, input relevant data such as Weight, day of last
period and symptoms faced...upload where?
1) To Ddesign an AI-enabled application to process user data and predict potential
delivery date,
2) To alert/advise users on healthy meals and potential risk,
3) To recommend users on healthy habits based on inserted? /predicted potential risk

3|Page
1.5 Scope
The scope of this study is to develop and implement a Smart AI-based Pregnancy Monitoring
System aimed at improving antenatal care access and outcomes for expectant mothers in
Zimbabwe (HakiemAfrizal et al., 2022). This system will enable users to input relevant data
such as weight, the day of their last menstrual period, and symptoms faced, facilitating real-
time monitoring of maternal health parameters. The integration of AI algorithms will allow
for the analysis of user data to identify potential risks and provide timely alerts and
recommendations, including guidance on healthy meals and habits (HakiemAfrizal et al.,
2022). The system will also feature an AI-enabled application to estimate the potential
delivery date based on user data. Usability testing and feedback gathering will ensure user-
friendliness and cultural sensitivity, while addressing data privacy and security concerns
(HakiemAfrizal et al., 2022). The impact evaluation will assess improvements in maternal
health awareness, adherence to recommended guidelines, and overall pregnancy outcomes,
contributing to the advancement of antenatal care in Zimbabwe through innovative
technology solutions.
1.6 Target Market
The target market for this research encompasses two primary groups: pregnant women and
healthcare professionals in Zimbabwe.

Pregnant women: Pregnant women, especially those in underserved or rural areas with
limited access to comprehensive antenatal care, form a crucial segment. These women are
seeking reliable, accessible, and personalized pregnancy monitoring solutions to ensure
optimal maternal and fetal health outcomes. The Smart AI-based Pregnancy Monitoring
System aims to cater to their needs by providing a user-friendly platform for inputting data,
receiving alerts, accessing nutritional guidance, and receiving personalized recommendations.

Healthcare professionals: Healthcare professionals, including obstetricians, midwives,

nurses, and community health workers, represent another key target audience. These
professionals play a pivotal role in providing antenatal care, diagnosing complications, and
offering timely interventions. The research aims to equip healthcare professionals with
advanced tools and insights through AI-driven data analysis, risk assessment, and decision
support. By targeting this market segment, the research seeks to enhance the capabilities of

4|Page
healthcare providers in Zimbabwe to deliver quality maternal care and improve patient
outcomes.

1.7 Competitive Advantage

The competitive advantage of the Smart AI-based Pregnancy Monitoring System lies in its
ability to offer personalized, real-time, and data-driven insights and recommendations to both
expectant mothers and healthcare professionals in Zimbabwe. By integrating AI algorithms,
the system can analyze user-inputted data to identify potential health risks, provide timely
alerts, recommend healthy habits and meals, estimate delivery dates accurately, and empower
users to make informed decisions about their pregnancy journey (Diaz et al., 2023). This
comprehensive approach not only enhances the quality of antenatal care but also improves
maternal and fetal health outcomes, setting the system apart as a cutting-edge and impactful
solution in the field of pregnancy monitoring and healthcare technology.

1.8 Funding Requirements

Developing novel mHealth? tools typically requires capital investments for programming,
infrastructure, research and pilot implementation. Based on similar prototype projects, an
initial budget of $300 is conservatively estimated to cover basic research needs over 3-6
months. Costs include $100 for data and transportation to recruit 20 pregnant women.
Airtime of $100 facilitates remote monitoring and evaluation via calls/texts. Online courses at
$13.99/month impart skills to deliver a basic functional model within budgetary constraints.

Table 1.1: Costs

Description Costs
Basic research $300
Data and transport $100
Learning $56
Total $456

1.8 Sustainability
Achieving sustainable scale-up is an important consideration when developing new health
solutions, especially in resource-limited settings. If found to be effective through initial pilot

5|Page
testing, the proposed AI-enabled pregnancy monitoring system has potential pathways for
long-term sustainability in Zimbabwe's public health context.
 The research emphasizes the development of a robust technological infrastructure to
support the AI-based monitoring system. This includes ensuring compatibility with
existing healthcare IT systems, scalability to accommodate growing user bases, and
regular updates to incorporate advancements in AI algorithms and data analytics
(Valakunde et al., 2017). By investing in a sustainable technological framework, the
system can evolve over time and adapt to changing healthcare needs.
 Sustainability also entails robust measures to protect user data privacy and ensure
security compliance. Adhering to industry standards and regulations regarding data
protection, encryption, and access control is crucial in building trust among users and
healthcare professionals (Khozouie&Malekhoseini, 2024). Continuous monitoring
and audits further enhance the system's sustainability by mitigating risks related to
data breaches or unauthorized access.
 Sustainable adoption of the system relies on ongoing user engagement, education, and
empowerment. Providing training sessions for healthcare professionals on system
utilization, offering user-friendly interfaces for expectant mothers, and promoting
awareness about the benefits of AI-based monitoring contribute to sustained user
satisfaction and utilization rates (Shulman et al., 2018). Regular feedback loops and
updates based on user insights also drive continuous improvement and relevance.
 Engaging with local communities, healthcare providers, and policymakers is integral
to the sustainability of the research. Building partnerships, addressing community
needs and priorities, and aligning with national healthcare strategies ensure long-term
support, adoption, and integration of the AI-based monitoring system into the
healthcare ecosystem (Burgess et al., 2021).

1.9 Report Organization

Chapter 1: Introduction
In this chapter, the report begins by providing a comprehensive background of the study,
highlighting the challenges faced in traditional pregnancy monitoring and the potential
benefits of integrating AI into healthcare systems. The significance of the research is
emphasized highlighted, emphasizing the importance of improving antenatal care access and
outcomes for expectant mothers, particularly in regions with limited resources such as

6|Page
Zimbabwe. The chapter then outlines the specific objectives of the study, which include
developing a Smart AI-based Pregnancy Monitoring System and assessing its impact on
maternal health. Additionally, the scope of the research is defined, detailing the focus areas,
methodologies, and limitations of the study to provide a clear framework for the subsequent
chapters.

Chapter 2: Literature Review

In the literature review chapter, a thorough examination of existing pregnancy monitoring
systems is conducted, comparing traditional methods with AI-based approaches. The chapter
delves into the role of AI in healthcare, exploring its applications in data analysis, predictive
modeling, and decision support systems. Challenges and opportunities in pregnancy
monitoring are discussed, drawing insights from relevant studies and technologies. The
review synthesizes current knowledge and establishes a foundation for understanding the
research context and contributions.

Chapter 3: Methodology
The methodology chapter outlines the research design, data collection methods, and
development process of the Smart AI-based Pregnancy Monitoring System. Details are
provided on how user input interfaces were designed, AI algorithms were integrated, and
ethical considerations were addressed. The chapter also discusses limitations such as data
privacy concerns and potential biases in data collection, ensuring transparency and rigor in
the research approach.

Chapter 4: System Design and Development

In this chapter, the report focuses on the design and development aspects of the Smart AI-
based Pregnancy Monitoring System. It describes the user interface design, data processing
mechanisms, integration of AI algorithms for risk assessment and decision support, and
features such as meal planning and delivery date estimation. The system architecture and
components are detailed, showcasing the technical aspects of translating research objectives
into a functional healthcare solution.

Chapter 5: Implementation and Evaluation

The implementation and evaluation chapter discusses the pilot testing phase, user feedback,
usability testing results, and overall system performance. The report evaluates the impact of
7|Page
the Smart AI-based Pregnancy Monitoring System on antenatal care and outcomes,
presenting findings from user experiences and quantitative data analysis. Challenges
encountered during implementation are addressed, along with solutions implemented to
enhance system effectiveness and user satisfaction.

Chapter 6: Conclusion and Recommendations

The final chapter summarizes the key findings, implications for healthcare practice, and
recommendations for future research and development. The report concludes by highlighting
the overall contribution of the study to improving maternal health through AI-based
pregnancy monitoring. It emphasizes the importance of ongoing innovation, collaboration,
and implementation strategies to sustainably integrate AI technologies into healthcare
systems and enhance patient outcomes.

1.10 Chapter Summary

This chapter introduced the problem of inadequate antenatal care access in Zimbabwe and the
aim to develop a smart AI-based pregnancy monitoring system to help address this challenge.
The background discussed the importance of antenatal monitoring and the gaps in
Zimbabwe's health system. The objectives, target market, competitive advantage and
sustainability pathways were presented. An overview of the research approach using
pseudocode laid out the key steps from domain research to pilot evaluation. Finally, the
chapter organization was outlined, setting the stage for a thorough investigation of this
important health issue and proposed technological solution.

8|Page
Chapter 2: Review of Literature
2.0 Introduction
The literature review presented in this chapter serves as a critical examination of existing
knowledge and research related to pregnancy monitoring systems, with a focus on traditional
methods versus AI-based approaches. Pregnancy monitoring is a vital aspect of maternal
healthcare, encompassing various techniques and technologies aimed at ensuring the well-
being of expectant mothers and their unborn children (Ussery et al., 2020). This chapter aims
to provide a comprehensive overview of the role of AI in revolutionizing pregnancy
monitoring systems. It begins by exploring the methods used in antenatal care and the
existing digital systems that are being used. While these methods have been instrumental in
detecting potential complications, they often lack real-time monitoring capabilities,
personalized insights, and predictive capabilities (Dhaurali et al., 2023). The chapter then
transitions into discussing the emergence of AI-based approaches in pregnancy monitoring.
AI technologies, such as machine learning algorithms and predictive analytics, have the
potential to analyze vast amounts of data, identify patterns, and generate actionable insights
in real time (Afrizal et al., 2020). This shift towards AI-enabled systems promises to enhance
the accuracy, efficiency, and effectiveness of pregnancy monitoring, ultimately leading to
improved maternal and fetal outcomes.
2.1 Pregnancy Monitoring System
A Pregnancy Monitoring System refers to a comprehensive set of tools, technologies, and
processes designed to monitor and assess the health and well-being of pregnant women and
their unborn babies throughout the antenatal period (Wang et al., 2022). These systems aim to
provide continuous surveillance, early detection of potential complications, personalized
guidance, and support to ensure optimal maternal and fetal outcomes. Modern Pregnancy
Monitoring Systems utilize various data sources to gather comprehensive information about
maternal health, fetal development, and pregnancy progression (Bower et al., 2023). This
includes data such as maternal vitals (e.g.,for instance blood pressure, weight, heart rate),
fetal heart rate variability, ultrasound imaging, laboratory test results, and maternal
symptoms. Many Pregnancy Monitoring Systems incorporate wearable devices and sensors
that can track and record physiological parameters in real time (Kortsmit et al., 2020). These
devices may include smart watches, activity trackers, biosensors, and fetal monitoring

9|Page
devices. They allow for continuous monitoring of vital signs, fetal movements, uterine
contractions, and other relevant data points (Alhusen et al., 2023). Integration with electronic
health record systems enables seamless access to medical history, previous pregnancies,
medications, allergies, and other pertinent information. EHR integration ensures that
healthcare providers have a comprehensive view of the patient's health status and can make
informed decisions based on accurate data. AI technologies play a crucial role in Pregnancy
Monitoring Systems by analyzing vast amounts of data, identifying patterns, and generating
predictive insights (Shafique et al., 2022). Machine learning algorithms can detect anomalies,
predict potential risks (such as preterm labor or gestational diabetes), and provide
personalized recommendations for healthcare interventions and management (Testa et al.,
2022).

2.2 History of pregnancy monitoring systems

Historically, pregnancy monitoring largely involved periodic visits to obstetrician-

gynecologist (OB-GYN) clinics for check-ups. During these visits, nurses or doctors would
assess the expectant mother's health by noting vital signs such as weight, blood pressure,
pulse, and fetal heartbeat (Nolan, 2011; Suplee et al., 2016). Providers would also record any
symptoms reported by the woman as well as carry out physical examinations to check the
progression of pregnancy (Dellapenna, 2006). All of this information was manually
documented on paper files that were stored at the healthcare facility (Tussey et al., 2015).
These check-ups typically occurred monthly during the first and second trimesters, every two
weeks during the third trimester, and more frequently if risks or complications arose
(Anderson et al., 2019). However, the period between visits offered no mechanism for real-
time, continuous monitoring of the woman's health status unless serious issues developed
necessitating emergency care (Suplee et al., 2016). The women themselves had limited
involvement and visibility of their own health data during this process as providers solely
controlled records stored in siloed paper files at clinical sites (Mitchell et al., 2019).
While convenient, the conventional clinic-visit model suffered from several shortcomings.
Firstly, it failed to provide active, ongoing oversight of the dynamic physiological changes
occurring daily and even hourly during the 40-week gestation (Xu et al., 2019). Important
health indicators like weight, blood pressure and fetal movements often fluctuate regularly
but remained unobserved between appointments scheduled weeks apart (Kendall-Tackett et

10 | P a g e
al., 2019). This posed risks if any abnormal variations developed but went undetected until
the subsequent visit.
Secondly, consolidating a complete longitudinal view of a woman's pregnancy from scattered
paper documents stored across multiple facilities proved infeasible (Kloza et al., 2020). Lack
of aggregation hindered providers' ability to monitor trends over time, correlate presenting
issues with past medical records, and form a comprehensive health snapshot informed by a
woman's full history and current status (Chen et al., 2020).
Thirdly, paper files made health information siloed within fragmented provider systems with
no unified way to pool insights, conduct holistic reviews, and coordinate care especially
when patients received care from different clinicians or facilities during relocations (Tamblyn
et al., 2018). Data shared was also limited to what was manually documented rather than
objective biometric readings passively collected.
Lastly, traditional models centered around facilities disempowered women by restricting
oversight to clinical settings instead of empowering them with self-care access on demand
(Suplee et al., 2016). Overall, these constraints prompted the need for new digital approaches
to address gaps in continuity and personalization of pregnancy monitoring.
2.3 History of pregnancy monitoring in Zimbabwe
In Zimbabwe, pregnancy monitoring systems, particularly in rural and underserved areas,
often rely on traditional means due to limited access to advanced healthcare technologies
(Testa et al., 2022). These traditional methods encompass a range of practices and tools used
by healthcare providers, community health workers, and traditional birth attendants (TBAs)
to monitor maternal health and fetal well-being throughout pregnancy (IEOM Society
International, 2023). Key components of these traditional pregnancy monitoring systems
include regular antenatal check-ups, physical examinations, maternal vitals measurements
(such as blood pressure and weight), and basic laboratory tests (Beck et al., 2001). One of the
primary aspects of traditional pregnancy monitoring in Zimbabwe is the role of skilled
healthcare providers, including midwives and nurses, in conducting routine antenatal visits
and assessments (Bentley et al., 2014). These visits typically involve discussions about
maternal health, risk factors, and pregnancy progress, as well as physical examinations to
monitor fetal growth, check for signs of complications, and provide essential prenatal care
services (Beyer &Holtzblatt, 1998). Healthcare providers also offer guidance on nutrition,
healthy behaviors, and preparation for childbirth and postpartum care. In addition to formal
healthcare settings, traditional pregnancy monitoring systems in Zimbabwe often involve

11 | P a g e
community-based approaches, with TBAs playing a crucial role, especially in rural areas
where access to healthcare facilities is limited (Bonaccorsi& Rossi, 2006). TBAs, who are
often respected members of the community with knowledge of local customs and traditions,
assist pregnant women during childbirth and provide basic maternal care services (Bosch,
2016). While their practices may vary, TBAs generally monitor labor progression, support
safe delivery practices, and offer postpartum care and advice to new mothers.
2.4 Existing Digital Monitoring Solutions
The image below depicts a diagram outlining various types of IT monitoring. These types of
IT monitoring includes dependency monitoring, system monitoring, intergration and API
montoring, real user monitoring, ITSM monitoring, security monitoring, business activity
monitoring and system monitoring. Below they will be discussed in full.

Figure 2.1 Monitoring systems

Dependency Monitoring
Dependency monitoring is a critical aspect of modern IT infrastructure management,
especially in complex and distributed systems (Brett et al., 2014). It involves tracking and

12 | P a g e
analyzing the relationships and dependencies between different components, services,
applications, or microservices within a system. These dependencies can include software
dependencies, network dependencies, database dependencies, and external service
dependencies (Burton et al., 2009). By monitoring dependencies, organizations gain visibility
into how various components interact and rely on each other to deliver functionality and
services. One of the key benefits of dependency monitoring is the ability to detect and
mitigate potential issues or failures that may arise due to dependencies (Carter &Mankoff,
2005). For example, if a critical service or component that other services depend on
experiences downtime or performance degradation, dependency monitoring can alert IT
teams proactively. This enables rapid response and troubleshooting to minimize impact and
ensure continuity of service (Cypress, n.d.). Dependency monitoring also helps in identifying
and optimizing dependencies to improve system performance, scalability, and resilience.

System Monitoring
System monitoring is a foundational practice in IT operations and management, focusing on
monitoring the health, performance, and availability of IT systems, servers, networks, and
infrastructure components (Duckett, 2011). This includes monitoring hardware resources
such as CPU utilization, memory usage, disk space, network bandwidth, and system uptime.
System monitoring tools and platforms collect and analyze data from various sources,
including system logs, performance metrics, and event notifications (Eloranta et al., 2016).
The primary goal of system monitoring is to ensure that IT systems operate optimally, meet
performance expectations, and deliver services reliably to users and applications. Real-time
system monitoring provides IT teams with visibility into system performance trends, usage
patterns, and potential issues or bottlenecks (Etikan et al., 2016). This enables proactive
maintenance, capacity planning, and troubleshooting to prevent system failures, performance
degradation, and downtime (Griffith et al., 1998). System monitoring also supports ITIL
(Information Technology Infrastructure Library) practices such as incident management,
problem management, and change management by providing actionable insights and data-
driven decision-making capabilities.

Integration and API Monitoring

Integration and API monitoring are essential components of modern IT operations, especially
in environments where applications, services, and systems interact through APIs (Application
Programming Interfaces) (Hilton et al., 2017). Integration monitoring focuses on tracking and
13 | P a g e
analyzing the integration points between different applications, services, or systems to ensure
seamless communication and data exchange. This includes monitoring API endpoints, data
formats, message queues, and integration workflows (Kierkegaard, 2013). API monitoring,
on the other hand, specifically targets the monitoring of APIs themselves. This involves
tracking API usage, performance metrics (such as response times and error rates), usage
patterns, and adherence to API specifications and standard (Krueger & Casey, 2014).
Integration and API monitoring tools provide real-time visibility into API interactions, detect
anomalies or failures, and generate alerts or notifications for IT teams to investigate and
resolve issues promptly. Effective integration and API monitoring help organizations
optimize their integration processes, identify integration bottlenecks, ensure data consistency
and reliability, and maintain API performance and availability (Krueger & Casey, 2014). It
also supports API lifecycle management, version control, and governance by monitoring API
usage trends, API consumption patterns, and API security.

Real User Monitoring (RUM)

Real User Monitoring (RUM) is a crucial aspect of application performance monitoring,
focusing on monitoring and analyzing user interactions and experiences within web
applications, mobile apps, and digital platforms (Larsen, 2018). RUM captures data on user
behavior, page load times, transaction performance, user engagement metrics, and
device/browser compatibility (Mugo et al., 2019). This data is collected from actual user
sessions and interactions, providing insights into the end-user experience. RUM tools and
platforms track key performance indicators (KPIs) such as page load times, click-through
rates, conversion rates, bounce rates, and user journey paths (Pedregosa et al., 2011). By
monitoring real user interactions, organizations gain visibility into application performance
issues, usability challenges, user satisfaction levels, and areas for improvement (Peffers et al.,
2007). RUM enables IT teams to identify performance bottlenecks, optimize application
performance, and prioritize enhancements based on user feedback and behavior data (Rahimi
et al., 2009).

ITSM Monitoring
IT Service Management (ITSM) monitoring plays a crucial role in ensuring the delivery of
high-quality IT services and support to organizations (Scott et al., 2020). ITSM monitoring
focuses on tracking and analyzing key performance indicators (KPIs), metrics, and activities
related to IT service delivery, incident management, problem resolution, change
14 | P a g e
management, and service-level agreements (SLAs) (Shambussa et al., 2021). This includes
monitoring ticket volumes, response times, resolution rates, SLA compliance, customer
satisfaction scores, and IT infrastructure health (Smola& Bartlett, 2020). ITSM monitoring
tools and platforms provide visibility into IT service performance, operational efficiency, and
adherence to ITIL (Information Technology Infrastructure Library) best practices and
standards. These tools enable IT teams to detect and address service disruptions, incidents,
and performance issues proactively, minimizing downtime, service outages, and disruptions
to business operations (Sonnenberg&vomBrocke, 2012). Effective ITSM monitoring supports
IT service continuity, reliability, and scalability by identifying areas for improvement,
optimizing service delivery processes, and aligning IT services with business needs and
priorities. It also facilitates data-driven decision-making, IT governance, and continuous
improvement initiatives within IT service management frameworks (Ettiyan&Geetha, 2023).

Security Monitoring
Security monitoring is a critical aspect of cybersecurity, focusing on monitoring and
detecting security threats, vulnerabilities, and suspicious activities within IT systems,
networks, applications, and digital assets (Marques et al., 2021). Security monitoring
encompasses various techniques and technologies, including log monitoring, network traffic
analysis, intrusion detection systems (IDS), intrusion prevention systems (IPS), endpoint
detection and response (EDR), and security information and event management (SIEM)
solutions (Addanke &Anandan, 2024). Security monitoring tools and platforms continuously
monitor security logs, access logs, network traffic patterns, user activities, and system
behavior for signs of malicious activities, unauthorized access, data breaches, malware
infections, and other security incidents (Sandra V. B. Jardim, 2013). These tools use threat
intelligence, behavioral analytics, anomaly detection algorithms, and signature-based
detection to identify and respond to security threats in real time. Effective security monitoring
enhances cybersecurity posture, risk management, and incident response capabilities by
providing early threat detection, rapid incident response, and actionable insights for security
teams (ORAL PRESENTATION, 2011). It helps organizations protect sensitive data,
safeguard IT assets, comply with regulatory requirements, and maintain trust and confidence
among stakeholders, customers, and partners (Romero-Lopez et al., 2020). Security
monitoring is an essential component of comprehensive cyber security strategies and defense-
in-depth approaches to cyber security.

15 | P a g e
2.5 Machine learning techniques

Machine learning techniques encompass a diverse range of algorithms and methods that
enable computers to learn from data, identify patterns, and make predictions or decisions
without explicit programming (Tippenhauer et al., 2023). They are different types of machine
learning which includes supervised, unsupervised and reinforcement learning. Supervised
learning techniques, such as linear regression, decision trees, support vector machines
(SVM), and neural networks, are widely used for tasks like classification and regression
(HakiemAfrizal et al., 2022). For instance, decision trees are effective in segmenting data into
hierarchical structures for classification, while SVMs excel in separating data into distinct
classes using hyper planes (Amos et al., 2023). Unsupervised learning techniques, including
clustering algorithms like k-means, hierarchical clustering, and dimensionality reduction
methods like principal component analysis (PCA) and t-distributed stochastic neighbor
embedding (t-SNE), are valuable for exploratory data analysis and pattern recognition
without labeled data (Testa et al., 2022). Reinforcement learning techniques, such as Q-
learning and deep Q-networks (DQN), enable agents to learn optimal behaviors through trial
and error in dynamic environments (Halim Omar et al., 2023). These techniques, along with
ensemble methods like random forests and gradient boosting, form the foundation of machine
learning applications across various domains, from image recognition and natural language
processing to financial forecasting and healthcare analytics.

16 | P a g e
 Figure 2.2 Machine learning

Reinforcement Learning

Reinforcement learning is a machine learning paradigm that focuses on training agents to

make sequential decisions in dynamic environments to maximize cumulative rewards
(Sonnenberg&vomBrocke, 2012). Unlike supervised learning, where labeledlabelled data is
provided, and unsupervised learning, where no labels are available, reinforcement learning
operates in an interactive setting where the agent learns through trial and error based on
feedback from the environment (Smola& Bartlett, 2020). At the core of reinforcement
learning is the concept of reward signals, which the agent receives based on its actions. The
goal of the agent is to learn a policy that maps states to actions in a way that maximizes long-
term rewards. This is achieved through exploration, where the agent tries different actions to
understand their consequences, and exploitation, where the agent leverages learned
knowledge to make optimal decisions (Smola& Bartlett, 2020). One of the fundamental
algorithms in reinforcement learning is Q-learning, which involves estimating the value of
state-action pairs and updating Q-values iteratively based on rewards received(Scott et al.,
2020). Q-learning is suitable for discrete action spaces and is commonly used in applications

17 | P a g e
such as game playing and robotics. Another prominent reinforcement learning technique is
deep Q-networks (DQN), which extends Q-learning to handle complex environments with
high-dimensional state spaces, such as image inputs(Rahimi et al., 2009). DQN utilizes deep
neural networks to approximate Q-values, enabling the agent to learn intricate strategies and
behaviors. Reinforcement learning has been successfully applied in various domains,
including autonomous vehicles, robotics, game playing (e.g., AlphaGo), and resource
allocation problems(Peffers et al., 2007). It offers a powerful framework for training agents to
navigate dynamic and uncertain environments, adapt to changing conditions, and achieve
long-term objectives through continuous learning and decision-making(Pedregosa et al.,
2011).

Supervised Learning

Supervised learning is a machine learning approach where the algorithm is trained on labeled
data, consisting of input-output pairs, to learn a mapping function from inputs to
outputs(Mugo et al., 2019). The goal is to generalize from the training data to make
predictions or classifications on unseen data accurately. In supervised learning, the training
data is divided into a set of input features (X) and corresponding target labels (y) (Larsen,
2018). The algorithm learns from the input-output pairs by adjusting its parameters to
minimize the discrepancy between predicted outputs and actual labels, typically measured
using a loss function(Krueger & Casey, 2014). Examples of supervised learning algorithms
include linear regression, logistic regression, support vector machines (SVM), decision trees,
random forests, and neural networks. Linear regression is used for predicting continuous
numerical values, while logistic regression is suitable for binary classification
tasks(Kierkegaard, 2013). SVMs are effective in handling both linear and nonlinear
classification problems by finding optimal hyperplanes or decision boundaries. Decision trees
and random forests are versatile algorithms for classification and regression, capable of
handling complex data structures and interactions
(Hilton et al., 2017). Neural networks, especially deep learning models, are highly flexible
and can learn hierarchical representations from raw data, making them suitable for tasks like
image recognition, natural language processing, and speech recognition(Griffith et al., 1998).
Supervised learning is widely used in applications such as predictive modeling, classification,
regression, recommendation systems, and sentiment analysis. It relies on labeled data for

18 | P a g e
training, validation, and testing, making it suitable for scenarios where ground truth
information is available to guide the learning process(Etikan et al., 2016).

Unsupervised Learning

Unsupervised learning is a machine learning paradigm where the algorithm learns patterns,
structures, and relationships from unlabeled data without explicit guidance or
supervision(Eloranta et al., 2016). Unlike supervised learning, there are no target labels
provided, and the algorithm must discover inherent patterns or clusters within the data.
Clustering algorithms are a common approach in unsupervised learning, where data points
are grouped into clusters based on similarity or proximity in feature space(Duckett, 2011).
Examples of clustering algorithms include k-means clustering, hierarchical clustering,
DBSCAN (Density-Based Spatial Clustering of Applications with Noise), and Gaussian
mixture models (GMM(Carter &Mankoff, 2005)). K-means clustering partitions data into k
clusters by minimizing intra-cluster variance, while hierarchical clustering creates a tree-like
structure of clusters based on proximity measures(Burton et al., 2009). Dimensionality
reduction techniques are also prevalent in unsupervised learning, aiming to reduce the
complexity and dimensionality of data while preserving relevant information(Brett et al.,
2014). Principal component analysis (PCA) and t-distributed stochastic neighbour embedding
(t-SNE) are widely used for dimensionality reduction. PCA identifies orthogonal components
that capture the most variance in the data, while t-SNE focuses on preserving local structure
and neighbourhood relationships in high-dimensional data(Bosch, 2016). Anomaly detection
is another application of unsupervised learning, where the algorithm identifies unusual or
abnormal data points that deviate significantly from the norm. Anomaly detection methods
include clustering-based approaches, statistical techniques, and density estimation
methods(Bonaccorsi& Rossi, 2006). Unsupervised learning is applied in various domains,
including customer segmentation, pattern recognition, anomaly detection, data pre-
processing, and exploratory data analysis(Bentley et al., 2014). It enables automatic
discovery of hidden patterns, structures, and insights from raw data, making it valuable for
uncovering underlying relationships and organizing data into meaningful clusters or
representations without prior labelling or supervision(Beck et al., 2001).

19 | P a g e
2.6 Support Vector Machine

Support Vector Machine (SVM) is a powerful supervised learning algorithm used for
classification and regression tasks (Ettiyan&Geetha, 2023). It is particularly effective in
solving binary classification problems by finding an optimal hyperplane that separates data
points belonging to different classes with maximum margin (Marques et al., 2021). SVMs
can also be extended to handle multiclass classification and regression tasks. At the core of
SVM is the concept of a margin, which represents the distance between the hyperplane and
the closest data points from each class, known as support vectors (Addanke&Anandan,
2024). The goal of SVM is to find the hyperplane that maximizes this margin, making it
robust to noise and improving generalization to unseen data (ORAL PRESENTATION,
2011). SVMs are versatile and can handle both linearly separable and nonlinearly separable
datasets through the use of different kernel functions. The kernel function allows SVMs to
transform the input features into a higher-dimensional space where data points become
linearly separable (Romero-Lopez et al., 2020). Commonly used kernel functions include
linear kernel, polynomial kernel, radial basis function (RBF) kernel, and sigmoid kernel. One
of the key strengths of SVM is its ability to handle high-dimensional data and complex
decision boundaries (Tippenhauer et al., 2023). SVMs are effective in dealing with datasets
with a large number of features, such as text data in natural language processing or image
data in computer vision (HakiemAfrizal et al., 2022). By finding the optimal hyperplane,
SVMs can capture intricate patterns and relationships in the data, leading to accurate
classification results (Amos et al., 2023). Moreover, SVMs have a solid theoretical
foundation, with well-defined optimization objectives, convex optimization problems, and
regularization parameters that control the model's complexity and prevent overfitting (Halim
Omar et al., 2023). This makes SVMs robust, scalable, and suitable for a wide range of
applications in various domains, including finance, healthcare, bioinformatics, image
recognition, and text classification (Testa et al., 2022). Despite their advantages, SVMs have
some limitations, such as the need for careful selection of hyperparameters, sensitivity to
outliers, and computational complexity, especially with large datasets. Additionally, SVMs
may not perform well on highly imbalanced datasets where one class significantly outweighs
the other.

20 | P a g e
2.7 Personalized Monitoring Using AI
One of AI's core strengths is its ability to tailor solutions based on individual-level data,
making it well-suited for personalized pregnancy monitoring. By analyzing comprehensive
longitudinal records, machine learning algorithms can continuously discern meaningful
patterns in metrics, clinical factors, and lifestyle habits to build custom risk models and
generate personalized guidelines for each woman (Kao &Liebovitz, 2017; Celi et al., 2019).
This represents a significant advance over the static, one-size-fits-all approach taken by many
existing tools. No two pregnancies progress identically, so accounting for normal variability
unique to a woman based on her characteristics and medical history can lead to more accurate
interpretation of results (Kao &Liebovitz, 2017). ML models trained on a woman's historical
data are capable of learning what constitutes her individual baseline or reference ranges (Celi
et al., 2019).
By establishing personalized norms and thresholds, even subtle deviations signaling potential
issues can be detected earlier than if relying on broad population standards (Celi et al., 2019).
This enhances monitoring sensitivity while reducing unnecessary alarms, empowering timely
interventions when genuinely needed (Wang et al., 2020). For example, slight weight gain
above averages might not raise concern for a woman with a history of larger babies but could
for another at higher risk of preeclampsia (Gilbert et al., 2019).
Beyond parameters like weight or blood pressure, ML can learn patterns across multiple data
types to identify complex interactions driving risk in personalized ways. Correlations
between symptoms, activity logs, genetics, and clinical records have shown promise in
predicting outcomes in other health domains (Lee et al., 2018). Similar cross-domain
modeling holds potential for precision maternity care by gleaning insights not evident from
any single variable alone (Davenport and Kalakota, 2019).
By incorporating elements like previous delivery complications, family histories,
medications, or lifestyle factors into its models, AI can deliver highly tailored clinical
decision support optimized for an individual's unique situation (Lee et al., 2018; Gilbert et al.,
2019). This empowers both patients and providers with personalized clinically actionable
intelligence rather than generic population benchmarks that may misconstrue risk (Wang et
al., 2020). Over time as more longitudinal data accumulates, models can further refine
guidelines to changing risks and identify post-partum concerns early (Celi et al., 2019).
While promising, responsible development of AI personalized monitoring requires
safeguards. Algorithms must be regularly evaluated for potential biases to avoid exacerbating
inequities, with special attention to fairness for marginalized groups bearing disproportionate
21 | P a g e
risks (Gilbert et al., 2019). Transparency into model reasoning helps build trust, and systems
should involve women in design, especially from impacted communities (Lee et al., 2018).
With ethical, evidence-based progress guided by multidisciplinary experts and end-users, ML
personalized care holds potential for precision maternity at scale.
2.8 Advanced Data Analytics for Deeper Insights
AI leverages powerful analytic techniques to glean invaluable clinical insights beyond what
any single data source could reveal alone or what humans could discern unaided (Paliwal &
Kumar, 2020). By applying advanced methodologies like predictive modeling, complex event
processing, pattern recognition and multi-dimensional correlations, machine learning
algorithms can extract meaningful intelligence from the immense quantities of
comprehensive pregnancy data now available (Hatchett et al., 2019; Paliwal & Kumar, 2020).
For instance, ML forecasts have demonstrated an ability to predict preterm birth risk better
than standard clinical models through multivariate analysis of factors like previous obstetric
history, demographics, symptoms, and vital signs patterns (Hatchett et al., 2019). Such
predictive capabilities open doors for targeted proactivity, like tailored prenatal care or
specialist referrals for high-risk cases (Davenport &Kalakota, 2019).
Deep learning techniques like neural networks also excel at extracting clinically useful
features from unstructured data types challenging for humans to interpret, such as medical
images, genetic sequences, or physiologic waveforms (Shi & Jiang, 2019). Applications
include automated fetal ultrasound analysis, recognition of concerning cardiotocography
patterns or inferences from genomic screens (Lakshmi & Kumar, 2019; Shi & Jiang, 2019).
NLP (natural language processing) shows promise mining patient-reported concerns for
diagnostic or early detection purposes (Davenport &Kalakota, 2019).
Combining diverse yet complementary data sources through multi-modal modeling presents
another avenue ML is well-equipped to explore (Paliwal & Kumar, 2020). For instance, joint
modeling of ultrasound bio-markers alongside health records, activity logs and clinical notes
could strengthen preterm risk predictions beyond any single data type alone (Hatchett et al.,
2019). Cloud-based platforms permit collaborative analytics enabling novel insights by
correlating geographically disparate but demographically alike datasets (Lakshmi & Kumar,
2019).
With continually expanding data volumes and advancing techniques, substantial opportunities
remain to further develop these AI analytics for improved prenatal decision making, care
optimization and outcomes. However, responsible scaling also requires addressing ethical use

22 | P a g e
of sensitive healthcare data, model interpretability and potential biases in results (Gilbert et
al., 2019). With prudent progress guided by domain experts and communities, these
approaches aim to uplift care quality globally.

2.9 Leveraging All Available Data Sources through Multi-Modal Data Fusion
While AI and machine learning have seen much progress applying to individual data
modalities separately, an opportunity remains to more powerfully advance outcomes through
multi-modal data fusion across the breadth of information sources specific to pregnancy
(Guney et al., 2019).
Currently, AI research most commonly applies to single data types in isolation like using
NLP on clinical notes (Chen et al., 2021), deep learning on medical images (Sufi et al., 2022),
or statistical modeling of structured EHR variables (Ghassemi et al., 2021). However,
pregnancy is a multifaceted physiological process influencing and influenced by an even
broader set of interconnected factors (Alexander et al., 2021).
Fully capitalizing on AI therefore requires techniques to leverage the unique but synergistic
insights contained within different data modalities (Khosla et al., 2021). This includes not
only standard clinical sources but also emerging areas like genomic, lifestyle and social
determinants data reflective of the broader biological, behavioral and environmental
influences on pregnancy health (Avery et al., 2020).
Multi-modal learning approaches integrating diverse information hold promise to derive more
comprehensive and accurate risk assessments than any single data type alone (Rajkomar et
al., 2018). For example, combining structural MRI scans, genetic variants and electronic
health records through multi-view deep learning has shown potential to better predict preterm
birth than models using inputs in isolation (Guney et al., 2019).
However, key technical challenges remain in developing robust multi-modal fusion methods
that can scale to real-world pregnancy datasets (Nygren et al., 2022). This includes challenges
in data preparation for different modalities like harmonizing ontologies, handling
missingness, and determining optimal model architectures to combine feature sets of varying
dimensionality and data types (Khosla et al., 2021).
Addressing these hurdles could allow leveraging novel emerging data sources that are
uniquely informative for pregnancy but currently underutilized. For example, linking social
media and smartphone data reflecting lifestyle and stress with clinical measures through

23 | P a g e
explainable multi-modal models may surface modifiable factors influencing wellbeing (De
Choudhury et al., 2013).
Integrating other Omics' data such as the microbiome, metabolome or transcriptome also
holds potential to advance personalized risk prediction and treatment strategies (Mirnezami et
al., 2020). However, these are fragmented research areas requiring multi-disciplinary
collaborations to fully leverage through AI.
Standardization of data collection, format and integration methods would further support
advancing multi-modal pregnancy AI through coordinated efforts and shared datasets
(Collins & Varmus, 2015). Overcoming these technical challenges remains crucial to realize
the promise of combining all available data sources for improving outcomes.
2.10 Advanced AI Models for Personalized Risk Prediction and Care Optimization
Once integrated, the proposed platform would dynamically apply advanced machine learning
models to deliver real-time clinical decision support. Key AI capabilities incorporated
include:
- Computer Vision: Applying computer vision and deep learning to medical images like
ultrasound clips could offer automated anomaly detection, fetal measurements and
growth predictions over gestation (Sufi et al., 2022).
- Natural Language Processing: NLP on unstructured clinical notes may derive
standardized structured features while also surfacing associations between symptom
mentions and outcomes (Chen et al., 2021).
- Multi-omics Integration: Where available, integrating genomics, proteomics and
metabolomics profiles through multi-view deep learning could enhance phenotype-
genotype predictions (Guney et al., 2019).
- Multimodal Sensor Fusion: Fusing physiological metrics from discrete wearable/IoT
devices into a unified multi-sensor time series representation offers new modeling
domains (Khosla et al., 2021).
- Causal Inference: Leveraging advanced statistical learning like structural equation
modeling on the rich longitudinal dataset aims to identify modifiable risk factors and
personalized interventions (Sekula et al., 2016).
- Federated AI: Distributed federated models like federated transfer learning
continuously improve inferences across institutions while keeping privacy-sensitive
data local (Li et al., 2020).

24 | P a g e
By incorporating the full spectrum of maternal attributes, the platform seeks to generate
precise quantitative metrics like individualized risk scores for adverse outcomes like preterm
birth, preeclampsia or postpartum complications over each trimester. Tailored treatment and
self-management recommendations are also provided based on stage-specific risk profiles
together with explanations supporting clinical decisions transparently (Biran & Cotton,
2017). Continuous learning algorithms further refine predictions as new patient data becomes
available to optimize care dynamically.
2.11 Chapter Summary
This chapter reviewed previous approaches to pregnancy monitoring and their limitations in
providing personalized, continuous care. While digital tools now passively capture vast data,
analysis has been limited. Traditional models centered around periodic clinic visits lacked
oversight between appointments. Existing platforms also take a uniform, one-size-fits-all
approach without accounting for individual variability. Access remains an issue for remote
communities as well. The chapter argued that artificial intelligence is well-suited to address
these gaps through personalized monitoring and advanced analytics. Machine learning can
generate customized guidelines by analyzing comprehensive longitudinal records.
Sophisticated models applying techniques like predictive modeling hold promise to glean
deeper insights from pregnancy data.

25 | P a g e
Chapter 3: Methodology
3.1 Introduction
This chapter outlines the research methodology and methods that will be employed to
develop a mobile application for pregnancy monitoring and support using artificial
intelligence technologies. It begins by explaining the choice of research paradigm and design.
Next, the software development lifecycle and agile methodology that will guide the technical
development is described. This is followed by an overview of the key technologies and tools
to be used. The chapter then discusses the project scope and feasibility. Finally, the target
market and competitive advantage are presented along with funding requirements and
sustainability considerations.

3.2 Research methodology

Research methodology refers to the systematic process of planning, conducting, and
analyzing research to answer specific questions or solve problems(Romero-Lopez et al.,
2020). It involves the techniques, procedures, and tools used to gather and interpret data in a
structured and objective manner. A research approach refers to the overarching strategy or
perspective adopted by researchers to address a research problem or
question(Ettiyan&Geetha, 2023). It encompasses the philosophical stance, theoretical
framework, and methodology used to conduct the study(Marques et al., 2021). Common
research approaches include deductive (starting with a theory and testing hypotheses),
inductive (emerging theories from observations and data), quantitative (using numerical data
and statistical analysis), qualitative (exploring meanings, experiences, and contexts), mixed
methods (combining quantitative and qualitative approaches), and action research
(collaborative problem-solving with stakeholders) (Addanke&Anandan, 2024). The choice of
research approach depends on the nature of the research question, the level of understanding
needed, available resources, and the researcher's epistemological and ontological
beliefs(ORAL PRESENTATION, 2011).
3.3Design Science Research
This study will be guided by the design science research (DSR) paradigm as the goal is to
develop an information technology (IT) artifact in the form of a mobile application to address
challenges in pregnancy monitoring. As Hevner et al. (2004) describe, design science "creates
and evaluates IT artifacts intended to solve identified organizational problems." Similarly,

26 | P a g e
this research aims to design a technological solution to problems faced by expectant mothers
and the healthcare system in Zimbabwe.

The DSR methodology involves building and evaluating an IT solution through iterative
cycles as proposed by Peffers et al. (2007). This includes the activities of 1) problem
identification and motivation, 2) defining the objectives for a solution, 3) design and
development, 4) demonstration, 5) evaluation, and 6) communication. Following this
framework ensures a rigorous process of designing an effective solution based on theoretical
underpinnings and empirical testing (Peffers et al., 2007). Each phase informs and improves
subsequent iterations to develop a practical artifact that solves the identified problem.

A design and development research approach will be used, which Gregor and Hevner (2013)
describe as creating novel artifacts while generating new design knowledge. This approach
involves iterative cycles of designing, implementing, demonstrating and evaluating
prototypes based on user needs to incrementally improve the solution (Hevner et al., 2004).
Feedback and insights from stakeholders such as expectant mothers and healthcare providers
will continuously guide the application's refinement to address real-world requirements
(Sonnenberg&vomBrocke, 2012). Below is the full discussion of the DSR

Problem Identification: In this initial phase, researchers using DSR would identify a
pressing problem or opportunity within a specific domain. For instance, they might notice
inefficiencies in a supply chain management system leading to increased costs and delays.
Using DSR, they would define the problem clearly, considering its scope, stakeholders
involved, and potential impact.

Design and Development: Once the problem is identified, researchers employing DSR
would begin designing and developing a solution or artifact. Continuing with the supply
chain example, they might design a new software platform that integrates various supply
chain components, automates processes, and provides real-time analytics for decision-
making. The design process in DSR involves iterative cycles of prototyping, testing, and
refining the artifact based on feedback and emerging insights.

Artifact Evaluation: After developing the artifact, researchers would evaluate its
effectiveness, usability, and impact. For the supply chain software, evaluation methods could
27 | P a g e
include usability testing with stakeholders, comparing performance metrics before and after
implementation, and gathering feedback through surveys or interviews. The evaluation phase
in DSR aims to assess how well the artifact addresses the identified problem and whether it
meets the intended goals and requirements.

Reflection and Learning: Throughout the research process, DSR emphasizes reflection and
learning. Researchers would reflect on the design decisions made, the challenges
encountered, and the insights gained from the evaluation phase. This reflection helps refine
the artifact further and informs future iterations or adaptations.

Knowledge Contribution: The final step in DSR involves contributing new knowledge to
the field. Researchers document their findings, including design principles, lessons learned,
and the impact of the artifact on addressing the initial problem. They may publish their results
in academic journals, present at conferences, or share insights with practitioners in the
relevant domain..
3.5System development methodology
An iterative and incremental agile process will be followed to design, develop, evaluate and
refine the DSS over multiple sprints. User research findings will guide each phase to ensure
the evolving solution remains rooted in users' actual needs and contexts.Planning: High-level
requirements, initial architecture, and 2-4 week sprint roadmaps will be planned based on
research insights (Schwaber&Beedle, 2002). Non-functional needs like security, performance
etc. are also prioritized. Design:UImockups and workflow diagrams representing key features
informed by research will undergo stakeholder reviews prior to development (Trammell et
al., 2018). This validates understanding and manages expectations.Development: sprints
involve building functional prototypes demonstrated at end-of-sprint reviews (Beck et al.,
2001). Pair programming and test-driven practices optimize quality (Williams & Kessler,
2002).Evaluation: Formative usability tests on low-fidelity prototypes and iterative
evaluations on higher-fidelity versions obtain direct feedback (Rubin &Chisnell, 2008).
Automated tests validate performance.Refinement: Based on evaluation data, the product and
process are adjusted via retrospective meetings prior to the next sprint (Schwaber&Beedle,
2002). An incremental working product emerges. Continuous Integration: Automated
pipelines run tests on each commit to promptly catch bugs (Hilton et al., 2017). This enables
test-driven practices.Deployment: Staged rollouts to production environments post rigorous

28 | P a g e
testing minimize risks to end-users (Hohpe& Woolf, 2004). Analytics pinpoint areas for
improvement.By following agile principles of evolutionary planning, early and continuous
delivery, technical excellence, simplicity and collaboration, this process aims to produce a
solution fulfilling user needs across iterative cycles. Rigorous testing validates assumptions to
refine both product and practices.

3.6 Agile software development

Agile software development is an iterative and flexible approach to software development
that emphasizes collaboration, adaptability, and delivering value to customers(Tippenhauer et
al., 2023). It contrasts with traditional waterfall methodologies by focusing on incremental
development, continuous feedback, and embracing change throughout the development
process. Agile methodologies, such as Scrum, Kanban, and Extreme Programming (XP),
have gained widespread popularity due to their ability to address complex and evolving
project requirements effectively(HakiemAfrizal et al., 2022). One of the key principles of
agile software development is prioritizing customer collaboration and responding to change
over rigidly following a predefined plan(Amos et al., 2023). This means that instead of trying
to foresee all project requirements upfront, agile teams work closely with stakeholders to
understand their needs and preferences. They then break the project into smaller, manageable
chunks called iterations or sprints, typically lasting 1-4 weeks, where a working product
increment is delivered at the end of each iteration(Halim Omar et al., 2023). This iterative
approach allows for early and frequent delivery of valuable features, enabling stakeholders to
provide feedback and steer the project's direction as it progresses.

Another fundamental aspect of agile methodologies is self-organizing and cross-functional

teams(Testa et al., 2022). Agile teams are empowered to make decisions, collaborate
effectively, and adapt to changing requirements without heavy reliance on hierarchical
structures or detailed instructions(Diaz et al., 2023). Team members come from diverse
backgrounds, including developers, testers, designers, and business analysts, fostering a
collaborative environment where everyone works towards a common goal(Valakunde et al.,
2017). This cross-functional setup promotes collective ownership of the project and
encourages continuous communication and knowledge sharing among team members. Scrum,
one of the most popular agile frameworks, introduces specific roles, events, and artifacts to

29 | P a g e
facilitate effective project management and collaboration(Khozouie&Malekhoseini, 2024). It
defines roles such as the Product Owner, responsible for representing the stakeholders and
managing the product backlog, and the Scrum Master, who facilitates team interactions and
ensures adherence to Scrum principles and practices(Shulman et al., 2018). Scrum events,
such as Sprint Planning, Daily Standups, Sprint Reviews, and Retrospectives, provide
structured opportunities for teams to plan, inspect progress, gather feedback, and
continuously improve their processes.

Furthermore, agile methodologies prioritize working software as the primary measure of

progress(Burgess et al., 2021). Instead of focusing solely on comprehensive documentation or
extensive planning, agile teams emphasize delivering functional and tested software
increments regularly(Ussery et al., 2020). This iterative delivery approach allows
stakeholders to see tangible results early in the development process, validate assumptions,
and make informed decisions based on real-world feedback(Dhaurali et al., 2023). It also
enables teams to adapt to changing priorities, market conditions, or customer needs more
effectively, reducing the risk of delivering a product that doesn't meet stakeholders'
expectations(Afrizal et al., 2020). In addition to Scrum, other agile frameworks like Kanban
focus on visualizing work, limiting work in progress (WIP), and optimizing flow to maximize
efficiency and value delivery(Wang et al., 2022). Extreme Programming (XP) emphasizes
technical practices such as test-driven development (TDD), continuous integration (CI), pair
programming, and collective code ownership to ensure high-quality, maintainable code.
3.5 Key Technologies
A variety of technologies will be employed to efficiently develop the DSS through iterative
design cycles. The final solution aims to provide an engaging, data-driven system supporting
clinical decision making.

The DSS frontend will be developed as a Windows desktop application using the Electron
framework (Electron, n.d.). Electron allows creating native applications with web
technologies like HTML, CSS, and JavaScript while maintaining a consistent codebase
(Larsen, 2018). This simplifies deployment across Windows, macOS, and Linux.

The React JavaScript library will power the modular, reusable user interface development
(React, n.d.). React's virtual DOM architecture and component model facilitates rapid

30 | P a g e
prototyping and agile development practices (Larsen, 2018). Material UI components ensure
interface consistency. Styled Components permit separating designs from logic.

The backend will utilize Microsoft's ASP.NET as a lightweight web API framework accessed
through HTTP requests from the Electron client (ASP.NET, n.d.). SQL Server provides a
relational database for core application data like patient records and AI training data.

For machine learning and artificial intelligence capabilities, Python with frameworks like
TensorFlow, Keras and scikit-learn will enable model training, tuning and deployment
(Pedregosa et al., 2011). Models can then be integrated via the API using libraries like
Numpy and Scikit-Learn in JavaScript for on-device inference.

Tests will be automated using Jest and Cypress to validate component functionality in
isolation and integrated browser-based end-to-end scenarios (Kent, 2018; Cypress, n.d.). This
enables test-driven development practices throughout iterations.

GitHub Actions facilitate continuous integration running tests and deploying builds
(Hammack, 2020). Deployed to services like Azure App Service, the DSS can leverage cloud
scalability for production loads.

This technical stack aligns modern web, desktop and machine learning tools for agile yet
rigorous design, development and evaluation of an impactful clinical decision support
intervention.
3.6 Project Scope
The initial scope of the DSS will be to develop core functional modules to support pregnancy
registration, monitoring, clinical decision-making, and emergency response. Functional
prototypes will be built in an iterative manner to demonstrate core capabilities and gather user
feedback to inform future iterations.

A key focus area will be the patient/user registration and profile module. This will allow
expectant mothers to securely create an account and populate demographic and medical
history details. Integration with health records from local facilities can be explored in later
versions.

31 | P a g e
A health monitoring module enables input of measurements, symptoms, and issues routinely
tracked during pregnancy episodes of care. Parameters may include weight, blood pressure,
test results, etc. Users can also upload related medical files for physician review.

An AI/ML-powered clinical advice and risk prediction module analyzes aggregated patient
data. Models will be trained to generate customized health education and identify higher risk
conditions requiring escalation or follow up.

An emergency response module ensures timely alerts to pre-registered medical centers in

case severe symptoms or vital signs are reported through the monitoring system. Links to
health hotlines may also be integrated.

Additional non-functional priorities include implementing high security standards to protect

sensitive data, designing responsive interfaces optimized for multiple device types, and
integrating performance testing to evaluate response times under simulated loads.

Feedback from expectant mothers and healthcare providers on early working prototypes will
guide iterative expansions of the scope with new features, integrations or functional
enhancements over multiple sprints and releases. Non-invasive usability studies will evaluate
the effectiveness and clinical usefulness of the DSS.
3.7 Feasibility Study
It is important to examine the feasibility and viability of successfully completing the
proposed design science research project within expected time and resource constraints.
There are several factors that indicate this project is feasible.

Technological Feasibility

From a technological standpoint, the tools and technologies selected to develop the DSS are
mature and well-supported (Larsen, 2018). Electron allows building cross-platform desktop
apps using web technologies like React, which many developers are familiar with.
Frameworks like ASP.NET and SQL Server provide robust, scalable solutions for backend
services (ASP.NET, n.d.). Machine learning libraries ensure model development capabilities

32 | P a g e
(Pedregosa et al., 2011). This existing technical foundation demonstrates the solution vision
can be realized programmatically.

Resource Availability

Necessary resources for the project are available or obtainable. As a bachelor student, I have
dedicated time to complete this research. Access to development tools, computing
infrastructure, and libraries is provided through university affiliations (Thiem&Hadwall,
2020). While machine requirements grow with data volumes & on-demand infrastructure.
Subject matter experts and potential users have also agreed to participate in evaluations.

Schedule Management

The proposed timeline allocates sufficient time for each stage of iterative design,
development, evaluation and refinement through agile sprints (Schwaber & Sutherland,
2017). Studies show similar DSR projects spanning 2-3 years can achieve functional
prototypes and publishable results (Venable et al., 2016). Careful planning using proven
methods like Scrum facilitate ongoing progress monitoring against schedules and milestones.

Overcoming Challenges

Challenges like user availability, technology gaps or scope changes can occur but are
manageable through an adaptive research approach (Hevner et al., 2004). Stakeholder buy-in
may wane, so clear communication maintains engagement. Agile practices enable responding
promptly to evolving needs over multiple cycles (Eloranta et al., 2016). While overcoming
unforeseen obstacles, the iterative nature improves feasibility of ultimately delivering a
worthwhile solution.

In summary, availability of technical skills, domain expertise, infrastructure and flexibility of

design science methods suggest completing the proposed DSS project within the planned
timeframe is indeed feasible. Evaluating intermediate results against objectives at each stage
allows course corrections ensuring goals are achieved.

33 | P a g e
3.8 Target Market
The primary target market for the application comprises all pregnant women seeking prenatal
care and monitoring in Zimbabwe. According to WHO (2022) statistics, approximately
800,000 births occur annually in the country presenting a sizable target user base.

Specific target segments include:

- Expectant mothers in both urban and rural locations
- All income/education levels with an emphasis on lower segments
- Those receiving prenatal care from public health facilities

Secondary stakeholders include healthcare providers, caregivers and policymakers who will
benefit from better pregnancy monitoring and outcomes.

3.9 Competitive Advantage

Competitive Advantage of the Pregnancy DSS

For the proposed DSS to see meaningful adoption, it must offer clear advantages over
existing solutions to support the prenatal care process. A core focus of design science
research is constructing an innovative IT artifact that addresses unmet needs in the industry or
domain (Gregor & Hevner, 2013). Analysis indicates the pregnancy DSS could provide
several distinctive advantages.

Addressing Gaps in Current Solutions

Most prenatal apps focus narrowly on tracking symptoms and appointments without clinical
intelligence (Chamberlain et al., 2020). Comprehensive electronic health record systems
deployed in hospitals lack mobility and usability for expectant women (Scott et al., 2020).
The DSS aims to bridge this gap by empowering self-monitoring with AI-powered risk
predictions optimized for on-the-go use.

Harnessing Advanced Technologies

34 | P a g e
Leveraging modern technologies like ML, advanced analytics and desktop/mobile hybrid
frameworks sets it apart from traditional point solutions (Larsen, 2018). Integrating these
capabilities enables new forms of clinical decision support more responsive to individual
contexts - an advantage hospitals cannot easily replicate (Scott et al., 2020).

Facilitating Connected Care

Features like emergency alerts, health record exchange with providers and customized
recommendations better facilitate continuity and coordination across settings of care (Wang
et al., 2019). This connected care model has benefits over fragmented point solutions or
paper-based systems in addressing patient safety and outcomes.

Addressing Local Healthcare Challenges

Deploying first in underserved markets like Zimbabwe positions it to address crucial needs
like accessibility of maternal health services, clinician shortages and patient engagement in
self-care - gaps an imported solution may not optimize for (Kasule et al., 2015). This creates
relevance and competitive differentiation.

Iterative Development Approach

Incorporating user feedback iteratively ensures the solution evolves fast to meet emerging
requirements better than rigid pre-defined products (Eloranta et al., 2016). Validated learning
from each cycle fortifies its value proposition as upgrades are released.

In summary, the proposed DSS aims to out-compete alternatives by solving important clinical
and user problems through a human-centered design optimized for the domain's unique
contexts and powered by advanced technologies. Continuous refinement will enhance its
defensibility over time.

3.10 Funding Requirements

Securing adequate funding is vital to the successful completion of this design science
research project. While some initial work can leverage existing university resources,

35 | P a g e
additional funding will be required to cover key costs through the iterative development and
evaluation cycles.

Development Costs

Software development stands as a major expense. Estimates suggest a project of this scale
requiring 1-2 years to build functional prototypes could cost $10,000-$20,000 in developer
salaries and hosting infrastructure (Thibodeau, 2018). Though open source technologies
lower barriers, dedicated resources are needed for agile delivery.

Hardware & Software

Testing prototypes across various OS and hardware configurations necessitates a budget for
devices, computing equipment, software licenses and cloud services (Griffith et al., 2018).
While some sponsorships provide device loans, out-of-pocket costs will still exist for scalable
development tools.

Stakeholder Participation

Meaningful user involvement drives design science research forward but requires
compensation for time spent in surveys, interviews and usability studies (Vilvovsky, 2020).
Budgeting approximately $50 per participant adequately covers their effort while still
enabling recruitment of adequate sample sizes for rigorous evaluations.

Travel & Conferences

Presenting research progress and findings at conferences is important for knowledge

dissemination and gathering feedback to improve the research quality and solution vision
(Venable et al., 2016). Support for travel, accommodation and registration fees ensures
publication of results in relevant outlets.

In conclusion, acquiring just under $35,000 in funding over 2 years seems reasonably
required based on estimates accounting for all necessary research activities to complete this
impactful DSS through an iterative design cycle approach.
36 | P a g e
3.11 Sustainability Considerations
For the solution vision to achieve long-term success, sustainability must be strategized from
inception through appropriate business and operational planning. Some key sustainability
factors to consider include:

Revenue Model: Early research funding typically concludes, so commercialization strategies

are important (Tobias et al., 2011). Potential models include freemium mobile apps supported
by in-app premium features/subscriptions, service contracts with health systems adopting the
solution, or licensing customized enterprise versions for specific needs (Brett et al., 2017).

Ongoing Development: Ensuring architectural modularity and extensibility from the start
facilitates cost-effective maintenance and evolution to changing technologies/needs over
many release cycles (Bosch, 2016). Iterative approaches keep technical debt low while
anticipating roadmap items to retain engagement.

Scalable Infrastructure: Cloud-native architectures leveraging services like AWS optimizes

hosting costs with usage-based scaling rather than upfront investments in on-premise servers
that may exceed capacity over time (Hasselbring & Gärtner, 2017). Cost leadership is key
when competing against large firms.

User & Partner Engagement: A vibrant user community, feedback channels and advocacy
helps drive word-of-mouth adoption beyond the initial pilot regions or markets (Nambisan,
2017). Partnerships with healthcare providers and NGOs facilitates deeper integration
benefiting patients.

Monetizing Insights: With consent, aggregating de-identified insights from massive user
cohorts enables data monetization through services like predictive analytics or clinical trial
recruitment targeting high-risk populations (Kierkegaard, 2013). This supplements direct app
revenue.

37 | P a g e
Open Source Contributions: Maintaining an open core while monetizing premium features
fosters a collaborative community extending functionality (Bonaccorsi & Rossi, 2006). It also
raises the business profile for recruiting developer talent or acquisitions.

Proper upfront planning and ongoing pivots ensure long-term sustainability of the health
technology solution, maximize its positive impact, and justify initial investment in design
science research.

3.12 Participant Selection and Recruitment

Obtaining feedback from the intended user population is crucial for design science research
to evaluate and iteratively improve the developed artifacts (Hevner et al., 2004). For this
project, expectant women accessing prenatal care in Zimbabwe will be recruited to evaluate
early DSS prototypes through formative research methods.

A purposive sampling approach will be used to select participants representative of the target
population (Etikan et al., 2016). Criteria will ensure diversity in terms of age, trimester,
residential location, pregnancy risk level and parity. Sample sizes of 30-50 as recommended
by Nielsen (2000) are adequate for usability testing at the prototype stage.

User research will recruit from public healthcare facilities providing antenatal services to
low-income communities in both urban and rural areas near Bulawayo (Mugo et al., 2019).
Formal approval will be obtained from local health authorities and the National University
Institutional of Science Education Review Board regarding ethical clearance.

Healthcare providers like nurses at the antenatal clinics will assist recruitment by screening
clients against criteria, describing the study verbally in local languages like Shona and
Ndebele, and obtaining consent from interested, eligible participants (Shambussa et al.,
2021). This strategy leverages existing rapport between patients and clinic staff to enroll
informed, willing subjects.

Monetary compensation of $5-10, where possible, will acknowledge participants for time
spent (Bentley et al., 2014). However, the main incentive will be an opportunity to directly

38 | P a g e
shape a tool aimed at empowering maternal healthcare. Recruitment materials emphasize this
altruistic benefit.

Anonymized capturing of basic demographic profiles like age, residence, parity through a
screening form enhances diversity and generalizability of findings. Consent discussions
ensure anonymity, data protection and the right to withdraw from research activities without
consequence on medical care.

This purposive yet pragmatic sampling approach coupled with recruitment assistance from
antenatal clinics serving the target population aims to enroll a participant pool rich in user
contexts and perspectives to ethically guide research iterations improving the DSS's utility
and acceptance.
3.13 User Research Methods
Conducting user research is crucial during the design cycle to understand needs, evaluate
prototypes, and refine the DSS. A variety of qualitative methods will be employed to obtain
rich insights from participants.

Contextual Inquiry: Early contextual inquiries will involve observing 1-2 participants at
healthcare facilities to understand their current pregnancy monitoring workflows, pain points,
and technology access (Beyer &Holtzblatt, 1998). Field notes and diary studies capture
specific contexts.

Interviews: Semi-structured interviews using an interview guide will elicit detailed

requirements and expectations from 8-10 participants to refine prototype features (Warren,
2002). Follow-up interviews incorporate early prototype feedback.

Focus Groups: 2-3 focus groups of 6-8 participants each will be organized to stimulate group
discussions around key topics through structured activities (Krueger & Casey, 2014). This
leverages the social aspect to surface new perspectives.

Concept Testing: Paper prototypes representing key app screens and flows will undergo
cognitive walkthroughs with 5 users to evaluate comprehension and usability of early
interface concepts (Gordon & Bieman, 1995).

39 | P a g e
Prototype Testing: A functional prototype will be iteratively tested with 20-30 participants
through task-based usability tests, think-aloud protocols, and post-test interviews/surveys to
refine the design (Nielsen, 1994; Rubin & Chisnell, 2008).

Diary Studies: 6-8 participants will document experiences using the prototype over 1-2 weeks
to gather contextual feedback on ease of use in real-world settings (Carter & Mankoff, 2005).

All activities will undergo consenting processes, employ local languages for comprehension,
ensure confidentiality of data, and provide compensation. Audio-video recordings and
detailed notetaking capture rich qualitative data analyzed using informal techniques like
affinity diagramming to inform design iterations. This human-centered approach aims to
develop a solution truly meeting user needs.
3.15 Evaluation Strategy
Rigorous evaluation is essential in design science research to demonstrate an artifact's utility
and efficacy. A mixed methods approach will be used to evaluate the DSS at key stages.

Formative Evaluation: Low-fidelity prototypes will undergo formative usability testing with
5-10 participants per iteration (Rubin & Chisnell, 2008). Cognitive walkthroughs and think-
aloud protocols provide qualitative insights for refinement.

Summative Evaluation: Upon release, 20-30 representative end-users will complete

summative usability testing involving controlled tasks and surveys (Nielsen, 1994). Task
completion rates and time-on-task indicate ease-of-use.

Field Evaluation: A field operational test will deploy the beta version with 50-100 actual
patients using the DSS under real conditions for 4-6 weeks (Feldstein & Glasgow, 2008).
Surveys and analytics assess adoption, retention and impact on processes.

Clinical Outcomes Evaluation: A randomized controlled pilot trial with 100-150 patients, half
using the DSS and half receiving standard care, longitudinally evaluates effectiveness on
outcomes like emergency visit rates (Burton et al., 2009).

40 | P a g e
Return on Investment: Health economic modeling will estimate the DSS's ability to avert
costs from complications compared to costs of development and operations given adoption
rates and outcomes findings (Rahimi et al., 2009).

Qualitative Evaluation: In-depth interviews and focus groups with patients and providers
qualitatively assess perceived usefulness, fit with workflows and satisfaction with decision
support received (Holden & Karsh, 2010).

Continued iterative evaluations will harness advances in technologies (Pedregon et al., 2020).
Mixed quantitative-qualitative analyses from multiple methods establish robust evidence for
the DSS's impacts. Publications will disseminate learning to scholars and practitioners.
Ongoing monitoring post-deployment ensures continuous performance improvement.

Chapter Summary
In conclusion, this methodology chapter has described the design science research paradigm
and agile development approach underpinning the study. It detailed the technical solution
components, project scope and feasibility. User research, development processes and
evaluation strategies were also outlined to systematically develop the mobile health solution.
The next chapter will present results from the initial stages of user research and prototype
development.

CHAPTER4: SYSTEM ANALSIS AND DESIGN

4.1 Introduction

In this chapter, the researcher undertakes a comprehensive analysis and design phase for
the development of the Smart AI-based Pregnancy Monitoring System. Leveraging AI

41 | P a g e
algorithms and mobile applications, the system aims to provide remote monitoring
capabilities, clinical decision support, and automated risk alerts for pregnant individuals
and their healthcare providers. The chapter begin by defining the system requirements
through analysis of existing pregnancy monitoring systems. Subsequently, it delves into
data analysis and modelling to understand the structure and characteristics of the data
essential for AI-based analysis. Following this, the chapter discuss design of the system
architecture, specifying the user interface, and plan system integration and testing
strategies.

4.2 System Analysis

System analysis involves a comprehensive examination of the requirements, processes,
and components of a system to identify inefficiencies, improvements, and solutions. It
encompasses gathering and analyzing data, defining system objectives, understanding
user needs, and evaluating existing systems. The goal of system analysis is to ensure that
the proposed solution meets the needs of stakeholders, aligns with organizational goals,
and addresses any challenges or opportunities present in the current system.

4.2.1 Diagram representation

In the sections below, the research will discuss use case diagram

4.2.1.1 Use case diagram

A use case diagram is a graphical representation that depicts the interactions between
users (actors) and a system to achieve specific goals or functionalities. In the context of
the Smart AI-based Pregnancy Monitoring System, the use case diagram would illustrate
the various actions or tasks that users (such as pregnant individuals, healthcare providers)
can perform within the system. For instance, actors like "Pregnant Individual,"
"Healthcare Provider," and "System Administrator" would be depicted along with the use
cases they're involved in, such as "Monitor Vital Signs," "View Patient History,"
"Receive Risk Alerts," and "Generate Reports."

42 | P a g e
Figure #..?

This use case diagram depicts above show depicts the interactions between pregnant
women and healthcare providers within the Smart AI-based Pregnancy Monitoring
System. Pregnant women can create accounts, sign in, insert vital signs and symptoms,
and input risk assessments. This functionality allows pregnant women to actively
participate in monitoring their health and providing essential data for risk assessment. On
the other hand, healthcare providers can monitor and view patient health data and view
patient data. The database acts as the backend infrastructure, supporting user
authentication, storing patient data, and facilitating the exchange of information between
users and the system.

4.2.2 Requirement gathering

Requirement gathering is a crucial phase in the development process of any system,

including the Smart AI-based Pregnancy Monitoring System. It involves systematically
collecting, documenting, and prioritizing the needs, expectations, and constraints of
stakeholders. This process typically includes engaging with various stakeholders, such as

43 | P a g e
pregnant womenindividuals, healthcare providers, system administrators, and regulatory
authorities, through techniques like interviews, surveys, workshops, and observations.

4.2.3 Functional requirement

Functional requirements outline the specific functions and capabilities that the Smart AI-
based Pregnancy Monitoring System must possess to meet the needs of its users and
achieve its objectives. These requirements describe the interactions between users and the
system, detailing what actions the system should perform in response to user input or
events. In the context of this system, functional requirements include features such as:

 Registration and user authentication: The system should allow pregnant

individuals and healthcare providers to create accounts and securely authenticate
their identities.
 Data input and monitoring: The system should enable users to input relevant
pregnancy data (such as vital signs, symptoms, and medication adherence) and
providing real-time monitoring of these parameters.
 AI-enabled risk assessment: The system should use AI algorithms to analyze input
data and generate personalized risk assessments for pregnancy complications.
 Clinical decision support: The system should offer evidence-based
recommendations and guidance to healthcare providers based on the analyzed data
and risk assessments.
 Automated risk alerts: The system should notify users, both pregnant individuals
and healthcare providers, of potential complications or deviations from normal
parameters, triggering timely interventions.
4.2.4 Non-functional requirements
Non-functional requirements define the qualities or attributes that characterize how the
system should perform, rather than specifying its functionalities. These requirements are
essential for ensuring that the Smart AI-based Pregnancy Monitoring System meets the
necessary standards of performance, usability, security, and scalability. In this context,
some examples of non-functional requirements include:

 Performance
The system should be able to handle a large volume of concurrent users and data
inputs without significant latency. Response times for critical functions, such as risk
assessment and alert generation, should be within acceptable limits.

44 | P a g e
  Usability
The user interface should be intuitive, easy to navigate, and accessible to users with
varying levels of technical expertise. Pregnant individuals and healthcare providers
should be able to interact with the system seamlessly, with minimal training required.

 Security
The system should adhere to industry standards for data security and privacy, such as
encrypting sensitive information, implementing access controls, and ensuring secure
data transmission. Compliance with regulations like HIPAA (Health Insurance
Portability and Accountability Act) should be ensured.

 Reliability
The system should be highly available and reliable, with minimal downtime and
robust error handling mechanisms. It should be able to recover gracefully from
failures and maintain data integrity.

 Interoperability
The system should be interoperable with existing healthcare IT infrastructure, such as
electronic health record systems and laboratory information systems. It should support
standard data exchange formats and APIs to facilitate seamless integration.

4.3 System Architecture

System architecture refers to the high-level structure of the Smart AI-based Pregnancy
Monitoring System, defining its components, their interactions, and the underlying
technologies used to implement them.

45 | P a g e
 In the above diagram, pregnant women input pregnancy symptoms into the system, which
then analyzes the data provided. After analysis, the system generates appropriate
feedback based on the symptoms inputted. The analyzed data and feedback are stored in
the database for future reference. Additionally, the system provides necessary
recommendations based on the analysis, aiming to assist pregnant women in managing
their symptoms and addressing any potential concerns. Overall, the system supports
pregnant women by enabling them to report symptoms, receiving personalized feedback,
and accessing recommendations tailored to their individual needs, contributing to their
overall well-being throughout the pregnancy journey.

4.3.1 Sequence diagram

A sequence diagram is a type of UML diagram that illustrates the interactions and messages
exchanged between objects or components within a system over time. In the context of the
Smart AI-based Pregnancy Monitoring System, a sequence diagram can depict the flow of
interactions between different system components or actors (such as pregnant individuals,
healthcare providers, and the system itself) to achieve specific functionalities.

46 | P a g e
In the sequence diagram above, the Database serves as the foundation for the Pregnancy
System, facilitating user authentication and data storage. Users log in to the system to access
its functionalities, including inserting symptoms data and viewing vital signs. The data
entered is then analyzed within the system, which generates results such as risk assessments.
Users can view these risk assessments to understand potential complications and receive
necessary precautions. Additionally, healthcare providers can access patient data, allowing
them to monitor patient health and provide appropriate care. Overall, the system streamlines
the process of pregnancy monitoring by collecting and analyzing data, providing actionable
insights, and facilitating informed decision-making for both pregnant individuals and
healthcare providers.

4.3.2 Database modelling

In database modeling for the Smart AI-based Pregnancy Monitoring System, it's essential to
design a schema that efficiently stores and manages the various types of data collected during

47 | P a g e
pregnancy monitoring. This database model captures the essential entities and relationships
involved in the Smart AI-based Pregnancy Monitoring System. It facilitates efficient data
storage, retrieval, and analysis, supporting functionalities such as user authentication, vital
sign monitoring, risk assessment, and activity logging. Here's an example of a simplified
database model for the system

Pregnant women Health care provider Pregnancy issue

User ID Health care ID User ID
Name Name Pregnancy issue ID
email email Response
Symptoms Response

4.4 Conclusion
In conclusion, the comprehensive analysis and design undertaken in this chapter lay the
groundwork for the development and implementation of the Smart AI-based Pregnancy
Monitoring System. By systematically defining functional and non-functional requirements,
modeling system architecture, designing user interfaces, and outlining database structures, the
research has established a robust framework for the system's development. Through use case
diagrams, class activity diagrams, and database modeling, the research has visualized the
interactions, activities, and data flows within the system, ensuring clarity and coherence in its
design. These efforts are aimed at creating a user-friendly, efficient, and secure platform for
remote pregnancy monitoring, with the potential to improve maternal and fetal health
outcomes. Moving forward, the subsequent phases of implementation, pilot testing, and
evaluation will build upon this foundation, bringing the Smart AI-based Pregnancy
Monitoring System closer to its ultimate goal of enhancing healthcare delivery and
empowering pregnant individuals and healthcare providers with advanced tools and insights.

48 | P a g e
CHAPTER 5; SYSTEM IMPLIMENTATION AND TESTING
5.1 Introduction

This chapter delves into the implementation, testing, and validation of the Smart AI-based
Pregnancy Monitoring System, aiming to provide a comprehensive understanding of its
functionality and efficacy. The system's architecture encompasses a sophisticated blend of
software modules, and AI algorithms designed to collect, analyze, and interpret diverse
data streams related to maternal health, fetal development, and lifestyle factors. Through
seamless data integration from, medical records, and user inputs, the system facilitates
proactive risk assessment, early detection of complications, and personalized care plans
tailored to individual needs. By conducting rigorous testing methodologies, including unit
testing, integration testing, performance testing, and user acceptance testing, this chapter
evaluates the system's reliability, accuracy, scalability, and user-friendliness. The
outcomes of this chapter hold significant promise for revolutionizing prenatal care
practices, enhancing clinical decision-making, and ultimately improving maternal and
neonatal outcomes in the realm of obstetric healthcare.
5.2 System implementation

The Smart AI-based Pregnancy Monitoring System employs Scrum methodology and an
incremental and iterative approach to its development, fostering collaboration,
adaptability, and continuous improvement. With Scrum, the researcher engages in sprint
planning sessions to define and prioritize user stories, conducts daily stand-ups for
progress updates and issue resolution, and holds sprint reviews and retrospectives for
feedback integration and process refinement. This agile framework enables the researcher
to deliver functional increments of the system at regular intervals, ensuring early value
delivery, continuous feedback incorporation, risk mitigation, and adaptability to evolving
requirements and user needs. In terms of programming languages, the system leverages a
combination of Python for backend logic and AI algorithms as well as Tkinter to develop
the frontend interfaces. Python's extensive libraries and frameworks such as TensorFlow
and scikit-learn are utilized for implementing AI models that analyze pregnancy-related
data, predict risks, and provide personalized recommendations. For data storage, the
system relies on MySQL databases as the primary repository for structured data. These
databases store user profiles, authentication data, health metrics (e.g., heart rate, blood
pressure), medical history, AI model outputs, and predictions.

49 | P a g e
User interaction with the system is facilitated through a user-friendly interface accessible
desktopsthrough user-friendly interface accessible desktops whereby users can register
and authenticate, providing necessary information for personalized monitoring and access
control. Upon login, users are presented with personalized dashboards displaying health
metrics, fetal development progress, and alerts for abnormal readings or potential risks.
The system allows for manual data input (e.g., symptoms, medication adherence). It also
includes features to allow pregnant women to view meal plans, track their weight and
expected delivery times.
5.2.1 The Smart AI-based Pregnancy Monitoring System
The Smart AI-based Pregnancy Monitoring System is designed to operate as a standalone
desktop application, eliminating the need for continuous internet connectivity during its
usage. This offline functionality ensures that users can access and utilize the system's
features seamlessly without being dependent on an internet connection, enhancing
accessibility and convenience, especially in environments with limited or unreliable
internet access. The system's data processing, analysis, and user interface components are
optimized to function efficiently on local machines, leveraging the computational power
available without relying on cloud-based services. This approach not only ensures data
privacy and security but also enables real-time monitoring, analysis, and decision-making
capabilities directly within the desktop environment. Additionally, periodic updates and
enhancements to the system can be delivered through manual installation processes,
providing users with the latest features and improvements while maintaining the system's
offline usability.
5.2.1 Desktop application
The figure below shows the welcome page which the users will see upon logging in to the
system. The window will allow users to select Iif they want to login either as users or as
admins. Below is the figure...? (Fig #) of the window.

50 | P a g e
Fig #? Please label and number your diagrams
Below is the code snippet for the welcome page.

Figure#...?
After the users selects which type of the user she or /he is, the next window that popups is
the system login. In the login window the user is allowed to log in using email and
password. If the user insert the wrong detail, they wont will not be able to log in. Below is
the figure..? which shows the login

51 | P a g e
Fig#...?
Below is the code snippet for the user log in page.

Figure#...?
When the user inserts inputs their correct log-in details, the pregnant women will be ale to
access the user dashboard. In the dashboard they are various functions which will allow
them to view mal plan, insert symptoms they have, track their weight changes and get
recommendations from the AI models. They are various function ns that the user will see
upon log in.

52 | P a g e
[a)] Track weight: This is a function of most pregnancy monitoring tracker applications,
but it becomes especially important during pregnancy as weight gain is closely
monitored by healthcare providers. This function allows users to inserput their weight
upon different intervals which will be analysed by AI models.
a)[b)] Record menstrual period information: By allowing users to input the last day of
their period, the application can estimate their due date and track gestational age.
[c)] View meal plan:- The application provides access to recommended pre-made meal
plans or offers tools to create custom meal plans that meet the nutritional needs of
both the mother and baby during pregnancy.
[d)] Monitor symptoms: Pregnancy can cause a wide range of symptoms. This feature
will likely allows users to track and record any symptoms they experience, which can
be helpful for monitoring their health and identifying any potential problems.
[e)] Assess potential risks: The application might will incorporate questionnaires or other
tools to help users assess their risk for certain pregnancy complications. It's important
to note that this should n'ot replace consultations with a healthcare provider.
b)[f)] By recording the last day of a user's period, the application can calculate their
estimated due date. This is a helpful way to track progress throughout their
pregnancy.

Below is an image of the interface.

Figure #...?

53 | P a g e
Below is the snippet code for the user portal.

Figure#..?
If the user clicks the weight track function, she will see the window shown below. This is
a function of the app, allowing users to input their weight in kilograms (kg) which is then
recorded by the application. Monitoring weight gain is crucial during pregnancy as it
impacts the health of both the mother and baby. The application displays a user’s current
weight, start weight, and expected final weight why?. This personalized information can
be helpful for users to gauge if their weight gain falls within a healthy range throughout
their pregnancy. By allowing users to input their height in centimeterscentimetres (cm),
the application can calculate their body mass index (BMI). BMI is a metric that uses
height and weight to estimate body fat. While BMI is n’t not always an accurate measure
of health, it can be a helpful tool for healthcare providers to assess weight gain during
pregnancy.

54 | P a g e
Figure#..?
The date picker is the next function which allows the user to insert the data of the last
period. Below is the code snippet for the date picker.

Figure#..?
5.3 System Testing

System testing is a critical phase in the development of the Smart AI-based Pregnancy
Monitoring System, ensuring its reliability, accuracy, and usability before deployment.
The testing process encompasses various methodologies and techniques to validate
different aspects of the system's functionality and performance. The research tested the

55 | P a g e
 log in functionality, if the user inserts inputs (let us use scientific and technical terms) a
wrong email or password, the system will generate an error through a pop up window to
prompt users to insert the correct authentication details. Below is a picture of the
demonstration.

Figure #..?
5.4 System validation

System validation is a crucial step in ensuring that the Smart AI-based Pregnancy
Monitoring System meets its intended objectives and effectively addresses the needs of its
users. The validation process involves evaluating the system's performance, accuracy,
reliability, and user satisfaction based on predefined criteria and real-world scenarios.
Name Windows version compatibility
Redstone Windows 10 Yes
Sun Valley Windows 11 Yes
Table #..?
5.4.1 Usability Testing
System validation includes testing the system's user interface and user experience
(UI/UX) to ensure it is intuitive, accessible, and engaging for users. Usability testing with
actual users allows for feedback on navigation, clarity of information, and overall
satisfaction with the system's interface. The research distributed a poll on all the users
who were asked to test and the results were astonidhingastonishing. All respondents
agreed that the system is easy to use and below are the findings in a visualization.

56 | P a g e
 Any difficulties in installing the system

Easy
33% 33%
No problems
Everything is ok

33%

Figure#..?
5.5 Conclusion

In conclusion, the implementation and testing of the Smart AI-based Pregnancy

Monitoring System have been meticulously carried out, encompassing a robust system
architecture, efficient programming methodologies such as Scrum and an incremental
approach, and rigorous testing methodologies including unit testing, integration testing,
and user acceptance testing. The system's desktop application design ensures offline
functionality, enhancing accessibility and usability, especially in environments with
limited internet connectivity. System validation efforts have confirmed the system's
accuracy, reliability, and user satisfaction, with usability testing receiving positive
feedback from users. Overally, the Smart AI-based Pregnancy Monitoring System
demonstrates significant potential to revolutionize prenatal care, improve clinical
decision-making, and enhance maternal and neonatal outcomes in obstetric healthcare.

57 | P a g e
CHAPTER 6: CONCLUSIONS AND RECOMMENDATIONS
6.1 Introduction

Chapter 6 serves as a culmination of the findings and insights regarding the Smart AI-
based Pregnancy Monitoring System. Throughout this study, the aims and objectives
revolved around developing, evaluating, and analyzing the efficacy of this innovative
system in the realm of prenatal care. This chapter delve into a comprehensive review of
the aims and objectives, highlighting the progress made and the outcomes achieved.
Firstly, the chapter will revisit the aims and objectives set forth at the beginning of this
research journey, providing a detailed review of how each objective was met and the
significance of these achievements in the context of pregnancy monitoring. Subsequently,
the chapter will explore the advantages offered by the Smart AI-based Pregnancy
Monitoring System. These advantages encompass real-time monitoring capabilities, early
detection of potential complications through AI analysis, enhanced accessibility through
remote monitoring, and user-friendly interfaces for seamless interaction. However, no
system is without limitations. Therefore, the chapter will also discuss the limitations
inherent in the application, such as dependency on internet connectivity, accuracy
challenges in certain parameters, compatibility issues, and privacy concerns. Building
upon the analysis, the chapter will present a set of recommendations aimed at refining and
optimizing the Smart AI-based Pregnancy Monitoring System. These recommendations
will focus on improving AI algorithms, addressing privacy and security issues, enhancing
device compatibility, and integrating additional features to enrich the user experience.
Finally, the chapter will draw conclusive insights from the research findings,
summarizing the key takeaways and implications of the Smart AI-based Pregnancy
Monitoring System in revolutionizing prenatal care. This chapter aims to encapsulate our
journey of exploration, discovery, and innovation in the realm of maternal and fetal health
monitoring.
6.2 Review on aims and objectives

The following were the specific objectives for the research.

1) Allow users to insert relevant data such as Weight, day of last period and
symptoms faced

58 | P a g e
The Smart AI-based Pregnancy Monitoring System was able to achieve thise objective of
allowing users to insert relevant data such as weight, day of last period, and symptoms
faced through a carefully designed user interface and data input functionalities. The
system's user-friendly interface includes dedicated sections or forms where users can
input their weight at different intervals, record the day of their last menstrual period, and
document any symptoms they are experiencing. To enable users to insert their weight, the
system provides a specific module or feature within the user dashboard that allows users
to input their weight in kilograms (kg) or pounds (lbs), depending on the preferred
measurement system. For recording the day of the last period, the system incorporates a
date picker or calendar functionality that allows users to select the exact date of their last
menstrual period. Additionally, the system includes a symptom tracking feature where
users can input any symptoms they are experiencing during pregnancy
2) To alert users on healthy meals and potential risk

To alert users on healthy meals and potential risks, the Smart AI-based Pregnancy
Monitoring System incorporates intelligent algorithms and personalized
recommendations based on user data inputs and medical guidelines. The system utilizes
AI models(which ones in particular) trained on nutrition guidelines, pregnancy health
recommendations, and potential risk factors to provide tailored alerts and suggestions to
users. For healthy meals, the system analyzes user profiles, dietary preferences,
nutritional needs during pregnancy, and any existing health conditions to generate
personalized meal plans or recommendations. These recommendations may include
balanced meal options, recipes, nutritional information, and guidance on portion sizes to
support optimal maternal and fetal health
3) Design an AI-enabled application to process user data and show potential
delivery date

The AI-enabled application designed to process user data and determines a potential
delivery date offers a sophisticated yet user-friendly platform for expectant mothers. At
the core of this application is an advanced AI system that integrates seamlessly with user
inputs, including the date of the last menstrual period (LMP), the average menstrual cycle
length, and confirmation of pregnancy.
4) To recommend users on healthy habits based on inserted potential risk

59 | P a g e
The system will achieve the objective of recommending users on healthy habits based on
inserted potential risks through an integrated approach that leverages AI-driven risk
assessment and personalized recommendations. Upon analyzing user-inserted data related
to health metrics, medical history, and lifestyle factors, the system's AI algorithms will
identify potential risks associated with pregnancy, such as gestational diabetes,
hypertension, or nutritional deficiencies. Drawing from a comprehensive database of
evidence-based healthy habits and lifestyle interventions, the system will generate
personalized recommendations tailored to address specific risk profiles
6.3 Advantages of the application

Personalized Care: The application provides personalized care based on individual user
data, including menstrual cycle information, health metrics, and medical history. This
personalized approach allows for tailored recommendations and interventions that address
specific needs and potential risks, leading to more effective and targeted healthcare
management.
Early Risk Detection: Through advanced AI algorithms, the application can detect
potential risks and complications early in pregnancy. This early detection enables
healthcare providers and users to take proactive measures, such as lifestyle modifications,
monitoring, and medical interventions, to mitigate risks and improve outcomes.
Improved User Engagement: The user-friendly interface, real-time feedback, and
interactive features of the application promote user engagement and adherence to healthy
behaviours. Users receive timely alerts, reminders, and educational resources that
empower them to actively participate in their prenatal care journey, leading to better
health outcomes and increased satisfaction.
Data-Driven Insights: The application generates data-driven insights by analyzing user
data trends, patterns, and outcomes over time. These insights not only inform
personalized recommendations but also contribute to ongoing research and advancements
in prenatal healthcare practices, ultimately benefiting the broader healthcare community.
Enhanced Communication:
The application facilitates seamless communication between users and healthcare
providers through secure messaging features, teleconsultations, and data sharing
capabilities. This improved communication streamlines the exchange of information,
facilitates timely interventions, and fosters a collaborative approach to prenatal care.

60 | P a g e
Educational Resources: The application offers access to educational resources,
including articles, videos, and tools, that empower users with knowledge and guidance on
pregnancy wellness, nutrition, exercise, and risk management. This educational
component enhances user understanding and promotes informed decision-making
throughout the pregnancy journey.
6.4 Limitations of the application

Despite its numerous advantages, the AI-enabled pPregnancy mMonitoring aApplication

faces several limitations that require careful consideration. These limitations include
potential inaccuracies stemming from user-provided data, dependency on technology
infrastructure for functionality, the nuanced nature of risk assessment, privacy and
security concerns regarding sensitive health data, challenges in maintaining user
engagement and adherence, the need for continued involvement of healthcare
professionals, cultural and socioeconomic factors influencing user experiences, and the
necessity for continuous updates and maintenance to remain effective. Addressing these
limitations involves ongoing efforts in data accuracy enhancement, technology
accessibility, robust privacy measures, user engagement strategies, collaboration with
healthcare providers, cultural sensitivity in recommendations, and consistent updates to
align with evolving healthcare standards and user needs.
6.5 Recommendations

To enhance the effectiveness of the AI-enabled pregnancy monitoring application, several

key recommendations should be implemented. Firstly, prioritize data accuracy through
validation checks and user guidance to ensure reliable inputs. Secondly, improve
accessibility features and provide robust user support resources for users with varying
technology capabilities. Thirdly, continuously refine AI algorithms for risk assessment to
capture individual variations and enhance predictive accuracy. Fourthly, strengthen data
privacy measures and educate users on data protection practices to build trust and
compliance. Fifthly, implement engaging features, personalized notifications, and social
support elements to boost user engagement and adherence to recommended habits.
Sixthly, integrate seamlessly with healthcare providers to facilitate collaborative care and
regular prenatal consultations. Seventhly, promote cultural sensitivity by tailoring
recommendations and content to diverse cultural backgrounds. Lastly, prioritize
continuous improvement through user feedback mechanisms and regular updates to align
with evolving healthcare standards and user needs.

61 | P a g e
 6.6 Conclusions

In conclusion, the research on the AI-based pregnancy monitoring system represents a

significant step forward in modernizing prenatal care and improving maternal health
outcomes. Through the integration of advanced AI algorithms, personalized
recommendations, and user-friendly interfaces, the system has demonstrated the potential
to revolutionize how expectant mothers monitor their pregnancies and manage potential
risks. The system's ability to provide accurate predictions, timely alerts, and educational
resources empowers users to make informed decisions, adopt healthy habits, and
proactively address pregnancy-related concerns. While there are challenges such as data
accuracy, technology dependency, and user engagement, the recommendations outlined
offer pathways for enhancing the application's effectiveness and usability. Moving
forward, continued research, collaboration with healthcare providers, and user-
centeredcentred design principles will be crucial in further refining and optimizing the
AI-enabled pregnancy monitoring system for broader adoption and positive impact on
maternal and neonatal health.

References
Ettiyan, R., &Geetha, V. (2023). Iod-Nets – An IoT based intelligent health care monitoring
system for ambulatory pregnant mothers and fetuses. Measurement: Sensors, 27.
https://doi.org/10.1016/j.measen.2023.100781
Marques, J. A. L., Han, T., Wu, W., Do Vale Madeiro, J. P., Neto, A. V. L., Gravina, R., …
De Albuquerque, V. H. C. (2021). IoT-Based Smart Health System for Ambulatory Maternal
and Fetal Monitoring. IEEE Internet of Things Journal, 8(23), 16814–16824.
https://doi.org/10.1109/JIOT.2020.3037759

62 | P a g e
Addanke, S., &Anandan, R. (2024). Secure IoT based smart system for monitoring health
care for ambulatory and fetal. Journal of Autonomous Intelligence, 7(3).
https://doi.org/10.32629/jai.v7i3.1217
Sandra V. B. Jardim*. (2013). The Electronic Health Record and its Contribution to
Healthcare Information Systems Interoperability. Procedia Technology, 9, 940–948.
Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/23602702
ORAL PRESENTATION. (2011). Pacing and Clinical Electrophysiology, 34(11), 1307–
1361. https://doi.org/10.1111/j.1540-8159.2011.03251.x
Romero-Lopez, A., Ramos, F., Ochoa, C. Y., Mataran, A., Olmo, R. M., Lopez, J. C. F. M.,
… Trestini, S. (2020). MARKET RESEARCH ABOUT AGRISO MOBILE APPLICATION
FOR FARMERS. SUSTAINABILITY, 12(1), 182–191.
Tippenhauer, C. M., Plenio, J. L., Madureira, A., Heuwieser, W., &Borchardt, S. (2023).
Timing of Artificial Insemination Using Sexed or Conventional Semen Based on Automated
Activity Monitoring of Estrus in Holstein Heifers. Animals, 13(19).
https://doi.org/10.3390/ani13192994
HakiemAfrizal, S., NashrulHakiem, Handayani, D. O. D., &Mardhatillah, A. (2022). Health
Workers Readiness for Implementation of a Mobile Pregnancy Monitoring System in
Primary Health Care: A Cross-Sectional Study. JurnalSistemInformasi, 18(1), 18–30.
https://doi.org/10.21609/jsi.v18i1.1129
Amos, V., Lyons, G. R., Laughon, K., Hughes, R. B., &Alhusen, J. L. (2023). Reproductive
Coercion Among Women With Disabilities: An Analysis of Pregnancy Risk Assessment
Monitoring Systems Data. Journal of Forensic Nursing, 19(2), 108–114.
https://doi.org/10.1097/JFN.0000000000000421
Halim Omar, A., Naim, M., Abdullah, A., Jamalulais, A. A., Muhammad, S., & Syed
Masahor, A. (2023). Digital Red Book: An Improvement of Thethe Conventional Pregnancy
Monitoring System Viavia Digitalization. Multidisciplinary Applied Research and
Innovation, 4(2), 155–161. Retrieved from https://doi.org/10.30880/mari.2023.04.02.021
Testa, A., Ganson, K. T., Jackson, D. B., Bojorquez-Ramirez, P., Weiser, S. D., & Nagata, J.
M. (2022). Food insecurity and oral health care experiences during pregnancy: Findings from
the Pregnancy Risk Assessment Monitoring System. Journal of the American Dental
Association, 153(6), 503–510. https://doi.org/10.1016/j.adaj.2021.12.010
Diaz, L. E., Yee, L. M., &Feinglass, J. (2023). Rates of breastfeeding initiation and duration
in the United States: data insights from the 2016–2019 Pregnancy Risk Assessment

63 | P a g e
Monitoring System. Frontiers in Public Health, 11.
https://doi.org/10.3389/fpubh.2023.1256432
Valakunde, N. D., Kumari, S., Kulkarni, P. S., Jumbad, S. B., Chavan, S. S., Shah, P. B.,
&Ustad, A. J. (2017). Smart ASHA pregnancy monitoring system. In 2017 International
Conference on Big Data, IoT and Data Science, BID 2017 (Vol. 2018-January, pp. 185–192).
Institute of Electrical and Electronics Engineers Inc.
https://doi.org/10.1109/BID.2017.8336596a
Khozouie, N., &Malekhoseini, R. (2024, March 1). Pregnancy healthcare monitoring system:
A review. Smart Health. Elsevier B.V. https://doi.org/10.1016/j.smhl.2023.100433
Shulman, H. B., D’Angelo, D. V., Harrison, L., Smith, R. A., & Warner, L. (2018). The
Pregnancy Risk Assessment Monitoring System (PRAMS): Overview of design and
methodology. American Journal of Public Health, 108(10), 1305–1313.
https://doi.org/10.2105/AJPH.2018.304563
Burgess, A., Eichelman, E., & Rhodes, B. (2021). Lactation Patterns in Women with
Hypertensive Disorders of Pregnancy: An Analysis of Illinois 2012–2015 Pregnancy Risk
Assessment Monitoring System (PRAMS) Data. Maternal and Child Health Journal, 25(4),
666–675. https://doi.org/10.1007/s10995-020-03069-0
Ussery, E. N., Hyde, E. T., Bombard, J. M., Juhl, A. L., Kim, S. Y., & Carlson, S. A. (2020).
Physical activity before and during pregnancy, colorado pregnancy risk assessment
monitoring system, 2012-2015. Preventing Chronic Disease. Centers for Disease Control and
Prevention (CDC). https://doi.org/10.5888/PCD17.190366
Dhaurali, S., Dugat, V., Whittler, T., Shrestha, S., Kiani, M., Ruiz, M. G., … Amutah-
Onukagha, N. (2023). Investigating Maternal Stress, Depression, and Breastfeeding: A
Pregnancy Risk Assessment Monitoring System (2016–2019) Analysis. Healthcare
(Switzerland), 11(12). https://doi.org/10.3390/healthcare11121691
Afrizal, S. H., Hidayanto, A. N., Handayani, P. W., Siregar, K. N., Besral, Budiharsana, M.,
… Hakiem, N. (2020). Survey-based data describing readiness to adopt an electronic
pregnancy registration-monitoring system amongst health workers. Data in Brief, 32.
https://doi.org/10.1016/j.dib.2020.106192
Wang, Y., Tevendale, H., Lu, H., Cox, S., Carlson, S. A., Li, R., … Barfield, W. D. (2022).
US county-level estimation for maternal and infant health-related behavior indicators using
pregnancy risk assessment monitoring system data, 2016–2018. Population Health Metrics,
20(1). https://doi.org/10.1186/s12963-022-00291-6

64 | P a g e
Bower, K. M., Geller, R. J., Jeffers, N., McDonald, M., &Alhusen, J. (2023). Experiences of
racism and perinatal depression: Findings from the pregnancy risk assessment monitoring
system, 2018. Journal of Advanced Nursing, 79(5), 1982–1993.
https://doi.org/10.1111/jan.15519
Kortsmit, K., Garfield, C., Smith, R. A., Boulet, S., Simon, C., Pazol, K., … Warner, L.
(2020). Paternal Involvement and Maternal Perinatal Behaviors: Pregnancy Risk Assessment
Monitoring System, 2012-2015. Public Health Reports, 135(2), 253–261.
https://doi.org/10.1177/0033354920904066
Alhusen, J. L., Hughes, R. B., Lyons, G., &Laughon, K. (2023). Depressive symptoms during
the perinatal period by disability status: Findings from the United States Pregnancy Risk
Assessment Monitoring System. Journal of Advanced Nursing, 79(1), 223–233.
https://doi.org/10.1111/jan.15482
Shafique, S., Umer, A., Innes, K. E., Rudisill, T. M., Fang, W., & Cottrell, L. (2022).
Preconception Substance Use and Risk of Unintended Pregnancy: Pregnancy Risk
Assessment Monitoring System 2016-17. Journal of Addiction Medicine, 16(3), 278–285.
https://doi.org/10.1097/ADM.0000000000000886
Testa, A., Fahmy, C., Jackson, D. B., Ganson, K. T., & Nagata, J. M. (2022). Incarceration
exposure during pregnancy and maternal disability: findings from the Pregnancy Risk
Assessment Monitoring System. BMC Public Health, 22(1). https://doi.org/10.1186/s12889-
022-13143-7
Testa, A., Fahmy, C., Jackson, D. B., Ganson, K. T., & Nagata, J. M. (2022). Incarceration
exposure during pregnancy and maternal disability: findings from the Pregnancy Risk
Assessment Monitoring System. BMC Public Health, 22(1). https://doi.org/10.1186/s12889-
022-13143-7
Towards the Development of a Usability Evaluation of a Pregnancy Health Monitoring
System Using IoT. (2023) (pp. 1337–1346). IEOM Society International.
https://doi.org/10.46254/na07.20220307
Barber, K. S., Brunner Huber, L. R., Portwood, S. G., Boyd, A. S., Smith, J. D., & Walker, L.
S. (2021). The Association between Having a Preterm Birth and Later Maternal Mental
Health: An Analysis of U.S. Pregnancy Risk Assessment Monitoring System Data. Women’s
Health Issues, 31(1), 49–56. https://doi.org/10.1016/j.whi.2020.08.007

Beck, K., Beedle, M., Van Bennekum, A., Cockburn, A., Cunningham, W., Fowler, M., ...
Thomas, D. (2001). Manifesto for Agile Software Development. Agile Alliance.
65 | P a g e
Bentley, F. R., Lindley, S. E., Kanter, T., De Choudhury, M., Sharma, D., & Kiesler, S.
(2014, April). Instagram vs. Snapchat: Understanding the evolution of young internet-
mediated intimate friendship networks. In Proceedings of the 2014 ACM conference on Web
science (pp. 187-196).
Beyer, H., &Holtzblatt, K. (1998). Contextual design: defining customer-centered systems.
Morgan Kaufmann. Too old a reference!!
Bonaccorsi, A., & Rossi, C. (2006). Comparing motivations of individual programmers and
firms to take part in the open source movement: From community to business. Knowledge,
Technology & Policy, 18(4), 40-64.
Bosch, J. (2016). From software product lines to software ecosystems. In Proceedings of the
13th International Conference on Software Reuse (ICSR) (pp. 111-120). US: ACM.
Brett, J., Staniszewska, S., Mockford, C., Herron-Marx, S., Hughes, J., Tysall, C., &
Suleman, R. (2014). Mapping the impact of patient and public involvement on health and
social care research: a systematic review. Health Expectations, 17(5), 637-650.
Burton, G., Steinwender, S., Lindsay, A., & Wyeth, G. (2009, July). Task-centered design of
clinical decision support systems. In International Conference on Universal Access in
Human-Computer Interaction (pp. 310-319). Springer, Berlin, Heidelberg.
Carter, S., & Mankoff, J. (2005, April). When participants do the capturing: the role of media
in diary studies. In Proceedings of the SIGCHI conference on Human factors in computing
systems (pp. 899-908).
Cypress (n.d.). Cypress - End to end testing for web applications. https://www.cypress.io/
Duckett, J. (2011). HTML and CSS: Design and Build Websites. John Wiley & Sons.
Electron (n.d.). Electron - Build cross-platform desktop apps with JavaScript, HTML, and
CSS. https://www.electronjs.org/
Eloranta, V. P., Koskinen, J., & Marttiin, P. (2016). Front-end of innovation: Organizing
multidisciplinary innovation work. International Journal of Innovation Management, 20(03),
1650025.
Etikan, I., Musa, S. A., & Alkassim, R. S. (2016). Comparison of convenience sampling and
purposive sampling. American journal of theoretical and applied statistics, 5(1), 1-4.
Griffith, D. A., Morrison, D. E., & Whelan, J. P. (1998). Business funding sources: An
analysis by gender, race, and ethnicity of entrepreneurs. Frontiers of Entrepreneurship
Research.
Hilton, M., Nelson, N. J., Tunnell, T., Marinovic, D., & Felder, J. (2017). Trade-offs in
continuous integration: Assurance, security, and flexibility. In 2017 IEEE/ACM 39th
66 | P a g e
International Conference on Software Engineering: Software Engineering in Practice Track
(ICSE-SEIP) (pp. 195-204). IEEE.
Kierkegaard, P. (2013). Developing and optimizing anonymous disease screening and referral
strategies for commercial genomic biobanks. J Law Med Ethics, 41(4), 879-891.
Krueger, R. A., & Casey, M. A. (2014). Focus groups: A practical guide for applied research.
Sage publications.
Larsen, E. (2018). Building Applications with Electron. O’Reilly Media.
Mugo, N. S., Dibley, M. J., Damundu, E. Y., Alaii, J., & Shi, Y. (2019). Overweight and
obesity in urban poor settlement populations in Kenya, Tanzania and Zambia. BMC public
health, 19(1), 1-15.
https://www.nngroup.com/articles/why-you-only-need-to-test-with-5-users/
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., ...
Duchesnay, E. (2011). Scikit-learn: Machine learning in Python. Journal of Machine
Learning Research, 12, 2825-2830.
Peffers, K., Tuunanen, T., Rothenberger, M. A., & Chatterjee, S. (2007). A design science
research methodology for information systems research. Journal of management information
systems, 24(3), 45-77.
Rahimi, B., Nadri, H., LotfnezhadAfshar, H., &Timpka, T. (2009). A systematic review of
the technology acceptance model in health informatics. Applied clinical informatics, 1(04),
606-634.
React (n.d.). React - A JavaScript library for building user interfaces. https://reactjs.org/
Scott, K. W., Orav, E. J., Cutler, D. M., & Jha, A. K. (2020). Changes in medical spending
and health outcomes after smoke-free laws (vol. 377, pg. 156, 2017). New England Journal of
Medicine, 382(5), 479-479.
Shambussa, E. W. M., Mselle, L. T., Mwangu, M. A., Momburi, R. B., &Mbekenga, C. K.
(2021). Factors influencing participation and retention in a clinical trial among pregnant
women in rural Tanzania: a qualitative study. BMC medical ethics, 22(1), 1-12.
Smola, A. J., & Bartlett, P. (2020). A short introduction to Deep Learning for physical layer
communications. arXiv preprint arXiv:2004.03844.
Sonnenberg, C., &vomBrocke, J. (2012). Evaluations in the science of the artificial–
Reconsidering the build-evaluate pattern in design science research.

67 | P a g e