11 V May 2023

https://doi.org/10.22214/ijraset.2023.53137
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue V May 2023- Available at www.ijraset.com

Obstructive Sleep Apnea Prediction Using Machine

Learning
Prof. Balaji Bodkhe1, Mandar Phirke2, Sankalp Oimbe3, Jayesh Khairnar4, Amol Kengar 5
1, 2, 3, 4, 5
Department of Computer Engineering, Modern Education Society's College of Engineering, Pune, Maharashtra, India

Abstract: A common sleep problem called obstructive sleep apnea (OSA) is characterized by periodic breathing pauses while
you're asleep. For OSA to be effectively treated and related health concerns to be avoided, an early and precise diagnosis is
essential. Machine learning approaches have become effective tools in recent years for managing and diagnosing OSA. The
research that used machine learning techniques for OSA detection and prediction are reviewed in detail in this publication. A
large dataset made up of polysomnography recordings and clinical data from OSA sufferers was gathered in order to achieve
this. Using this dataset, many machine learning models were trained to categorize the severity levels of OSA or predict the
occurrence of OSA, including support vector machines, random forests, and deep neural networks. The findings demonstrated
the powerful ability of machine learning algorithms to recognize OSA and distinguish between various severity levels. These
models outperformed conventional diagnostic techniques in terms of high sensitivity and specificity rates. To further shed insight
on the underlying patterns and processes of the illness, feature selection approaches were used to determine the most pertinent
physiological parameters for OSA detection. The effectiveness and precision of screening and therapy strategies may be
improved by incorporating machine learning algorithms into OSA diagnosis. This study advances our understanding of the use
of machine learning in sleep medicine and paves the door for the creation of automated and individualized OSA diagnostic tools.
Keyword: Machine Learning, algorithms, apnea-hypopnea index (AHI), Obstructive sleep apnea (OSA) disease, Support vector
machine, random forest algorithm.

I. INTRODUCTION
Obstructive sleep apnea (OSA) is a common sleep disorder affecting a significant portion of the population. It is characterized by
recurrent cases of partial or complete apnea during sleep, leading to sleep disturbances and hypoxia. OSA is associated with many
health risks, including cardiovascular disease, daytime sleepiness, cognitive decline, and reduced quality of life. Therefore, accurate
and timely detection of OSA is crucial for effective diagnosis and appropriate treatment.
Machine learning algorithms have demonstrated great potential in a variety of fields, including healthcare, by efficiently analyzing
large data sets and extracting valuable patterns and insights treat. In the context of OSA, machine learning algorithms can use
features extracted from physiological signals, such as electroencephalogram (EEG), electrocardiogram (ECG), airflow and oxygen
saturation, to develop models capable of accurately classifying OSA severity or predicting OSA presence.
One of the main advantages of using machine learning in OSA diagnostics is its automation and scalability. By developing powerful
and accurate machine learning models, it is possible to streamline the diagnostic process and reduce the burden on sleep
laboratories. This could allow broader access to OSA diagnosis and speed the initiation of treatment for those affected. Several
studies have investigated the application of machine learning algorithms in OSA detection and classification. These studies used a
variety of machine learning techniques, including support vector machines, random forests, decision trees, artificial neural networks,
and deep learning models. By training these models on a comprehensive dataset that includes physiological signals and clinical
information from OSA patients, researchers have obtained promising results for accurately identifying OSA and distinguish between
different degrees of severity. Apnea-Hypopnea Index (AHI) index is used to determine the levels of OSA in which (AHI 5), mild
(AHI is in 5–14), moderate (AHI is in 15–30), and severe (AHI 30) respectively.
In addition, feature extraction and selection techniques were used to identify the most suitable physiological features for OSA
detection. This process reduces computational complexity, improves model interpretability, and provides valuable insight into the
underlying OSA models and mechanisms. The integration of machine learning algorithms in OSAS diagnostics has the potential to
revolutionize the field of hypnotics. By harnessing the power of machine learning, clinicians and researchers can develop automated
and personalized diagnostic tools for OSA. These tools can not only improve the efficiency and accuracy of screening and
management, but also contribute to a better understanding of the pathophysiology of OSA. In summary, OSA is a common sleep
disorder with important health implications.

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 6113
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue V May 2023- Available at www.ijraset.com

The use of machine learning techniques in OSA detection and classification holds promise for improving the diagnostic process. By
taking advantage of physiological signals acquired during sleep, machine learning models can accurately identify OSA and
distinguish between severity levels. The development of automated and personalized OSA diagnostic tools could improve access to
diagnosis and enable prompt initiation of treatment. In addition, application of machine learning can provide valuable insights into
the underlying mechanisms of OSA, enhancing our understanding of this complex disorder.

II. PROBLEM STATEMENT

The sleep disorder obstructive sleep apnea (OSA) has long-term effects. Cardiovascular disorders and sleep-related problems are
examples of long-term impacts. A polysomnogram, an overnight sleep test, is frequently used to diagnose OSA. Long diagnostic
wait times can make monitoring expensive. So, in this study, we developed a machine learning-based OSA prediction system to
determine if a patient has obstructive sleep apnea or not.

III. PROPOSED SYSTEM

In the proposed system, we have built machine learning algorithm based OSA prediction system using Random Forest algorithm
and Support vector machine. Based on the patient medical inputs that are entered by the user the disease is predicted. In this
prediction system we are going to predict apnea disease.
In obstructive sleep apnea prediction, there is possible to predict obstructive sleep apnea at a time. So, by which user doesn’t need to
go for heavy systems initially which will save the time and money.

IV. LITERATURE REVIEW

According to the paper, this paper focuses about as obstructive sleep apnea is one of the dangerous diseases in the world. OSA can
cause many varieties of disorders to the patients which includes heart disease etc. In this paper author has used machine learning
algorithms to find out disease.
The results that we obtained show very good ability to identify patients with OSA using data collected during wakefulness and
deployment of machine learning techniques. These promising results can be the seed for a new diagnostic modality that allows rapid
diagnosis and treatment without the need for overnight sleep testing. More experiments and data are needed to optimize this method
and select more discriminatory features, which is subject of future studies [1].
The study in “SmartCare: A Symptoms Based Disease Prediction Model Using Machine Learning Approach” introduces SmartCare,
a machine learning-based disease prediction algorithm based on symptoms. With the help of machine learning algorithms,
SmartCare seeks to reliably detect diseases using symptom data. The algorithms in the model are trained and optimized using a
sizable dataset of symptoms and associated disease diagnoses. The development and assessment of SmartCare utilizing different
machine learning methods, including decision trees, random forests, support vector machines, and neural networks, are covered in
the study. These algorithms examine the symptom data and identify patterns to anticipate outcomes in cases that have not yet
occurred. Metrics including accuracy, precision, recall, and F1 score are used to evaluate the model’s performance. The outcomes
show how well SmartCare does disease prediction based on symptoms. The model achieves excellent accuracy and exhibits the
potential in helping medical practitioners identify disorders early and treat them effectively. SmartCare offers the potential to
increase healthcare decision making and patient outcomes by utilizing machine learning. In conclusion, SmartCare is a machine
learning-based disease prediction model based on symptoms. The model has the potential to help medical personnel identify patients
more quickly and provide better care because it accurately predicts diseases based on symptoms. To guarantee its dependability and
performance in actual healthcare settings, more study and validation are required [2].
The study “Diagnosis of Obstructive Sleep Apnea during Wakefulness Using Upper Airway Negative Pressure and Machine
Learning”, instead of using conventional techniques that call for nightly sleep testing, researchers concentrated on developing a
means to identify obstructive sleep apnea (OSA) in people while they are awake. OSA, a sleep disorder that can have negative
effects on one’s health, is characterized by frequent breathing pauses while one is asleep. The researchers used machine learning
methods and upper airway negative pressure (NP) to accomplish this. To diagnose OSA and gauge its severity, participants
underwent nocturnal polysomnography (PSG), which is regarded as the gold standard in the field. The subjects in separate tests
were then exposed to NP while they were awake. Participants’ airflow waveforms were recorded during these experiments. These
waveforms’ shape properties and other features were retrieved by the researchers as more accurate diagnostic performance.

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 6114
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue V May 2023- Available at www.ijraset.com

The findings of this study suggest that using NP and machine learning techniques has the potential to diagnose OSA during
wakefulness, providing an alternative to conventional overnight sleep tests. This approach could make OSA diagnosis more
accessible, convenient, and less intrusive for individuals, potentially leading to earlier detection and treatment.[3]
The goal of the paper “Sleep Apnea Event Prediction Using Convolutional Neural Networks and Markov Chains” by Rim Haidar,
Irena Koprinska, and Bryn Jeffries is to forecast sleep apnea occurrences using machine learning methods. A sizable portion of
adults suffer from sleep apnea, a breathing disorder that is frequent. The authors suggest three techniques for anticipating sleep
apnea occurrences based on Markov chains and convolutional neural networks (CNNs). analyzing the data from the previous 60
seconds, they anticipate apnea and hypopnea events within a 30-second window by analyzing respiratory signals such as nasal flow,
abdominal, and thoracic signals. On a sizable dataset with 48,000 examples from 1,507 people, the effectiveness of the suggested
strategies is assessed. The accuracy of 80.78% and the F1 score of 80.63% obtained in the results show the CNN method’s efficacy.
The authors also examine the Markov chain laws and offer information on how apnea and regular events change with time. The
linked works discussed above use feature engineering and neural networks to produce accurate predictions. Overall, this research
advances the use of machine learning techniques to predict sleep apnea events and offers insights into the potential of CNNs and
Markov chains in this field.[4]
The study’s main goal is to forecast diseases using symptoms and machine learning methods. Using symptoms to predict diseases
can help in early discovery and prompt treatment, improving patient outcomes. Large datasets of symptoms can be analyzed using
machine learning algorithms to find trends and provide precise disease predictions. The research investigates the use of different
machine learning techniques, including neural networks, support vector machines, decision trees, and random forests, for illness
prediction. These algorithms are trained using a dataset that includes disease classifications and symptoms. The computers can
forecast new, unforeseen cases based on their symptoms by learning from this data. The major objective of the project is to predict
diseases using symptoms and machine learning techniques. In order to improve patient outcomes, early diagnosis and fast treatment
are two benefits of using symptoms to forecast diseases. Machine learning algorithms can be used to examine large datasets of
symptoms to uncover trends and make accurate disease predictions. The study examines the application of various machine learning
methods, such as neural networks, support vector machines, decision trees, and random forests, to the prediction of sickness. A
dataset with disease diagnoses and symptoms is used to train these algorithms. By studying this data, computers can forecast future,
unforeseen cases based on their symptoms [5]
The research paper titled “Obstructive Sleep Apnea Prediction Using Deep Learning Techniques” presents a unique method for
predicting OSA using deep learning. The authors analyze and categorize sleep data using a variety of deep learning approaches,
including convolutional neural networks (CNNs) and recurrent neural networks (RNNs). The suggested approach seeks to precisely
predict OSA, a prevalent sleep disease, by utilizing the power of deep learning. In order to enhance patient outcomes and provide
the right interventions, the paper emphasizes the significance of early OSA detection and prediction. The algorithm can
automatically learn and extract significant elements from sleep data according to the deep learning techniques used in the study,
enabling precise OSA prediction. The outcomes reveal promising OSA prediction ability, highlighting the promise of deep learning
for improving the detection and treatment of sleep disorders. In conclusion, the study offers a novel method for anticipating
obstructive sleep apnea utilizing deep learning approaches. The model efficiently analyses and categorizes sleep data to determine
whether OSA is present using CNNs and RNNs. In terms of prediction accuracy, the suggested method shows encouraging results.
It also has the potential to improve the early identification and treatment of sleep disturbances, which will eventually help people
who are at risk of OSA [6].
In the study “A New Method for Self-Estimation of the Severity of Obstructive Sleep Apnea Using Easily Available Measurements
and Neural Fuzzy Evaluation System”, We describe a unique approach for assessing the severity of one's own OSA using easily
available data and a neural fuzzy evaluation system. The authors provide a doable strategy that enables people to evaluate the
severity of their OSA without specialized equipment or a specialist evaluation. The technique makes use of readily available
measures, including neck circumference, age, gender, and body mass index (BMI), which are entered into a neural fuzzy evaluation
system. This approach offers a thorough assessment of OSA severity by combining fuzzy logic and neural network algorithms.
Using this technique, people can get a rough idea of how severe their OSA is, which can assist in direct subsequent steps like getting
medical advice or looking into treatment alternatives. This method’s ease of use and accessibility make it a useful tool for gauging
one’s own OSA severity. In conclusion, the study report proposes a unique technique for determining the severity of OSA utilizing
metrics that are easily accessible and a neural fuzzy grading system. This method enables people to understand their OSA status,
permitting early intervention and suitable treatment techniques [7]

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 6115
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue V May 2023- Available at www.ijraset.com

The study “Explainable machine learning for sleep apnea prediction”, the Author links an open overlay panel The use of explainable
machine learning approaches for sleep apnea prediction is covered in the research article. A common sleep disorder called sleep
apnea is characterized by breathing interruptions while you’re asleep. The capacity of machine learning models to predict sleep
apnea occurrences has shown promise, but its interpretability restricts both their practical application and our ability to comprehend
the underlying processes that influence the predictions. The use of explainable machine learning techniques to increase the
interpretability of sleep apnea prediction models is explored in this work. These techniques seek to provide light on the traits and
tendencies that the models employ when making predictions. Clinicians and researchers can learn crucial information for making
decisions about diagnosis and treatment by comprehending the crucial aspects that affect sleep apnea prediction. The work
underlines the significance of interpretability in healthcare applications and shows the efficacy of explainable machine learning
algorithms in sleep apnea prediction. The results imply that transparent and interpretable outcomes can be produced by explainable
models, assisting in the comprehension and acceptance of machine learning models in the medical industry. In order to improve
clinical decision-making and patient care, the research emphasizes the necessity for explainable methodologies in sleep apnea
prediction [8].
The study “Sleep Apnea Events Detection Using Deep Learning Techniques” provides a unique method for detecting sleep apnea
events. To analyze and categorize sleep signals, the authors use a variety of deep learning approaches, including Convolutional
Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks. The suggested model intends to precisely identify sleep
apnea events, which are essential for diagnosing and treating sleep disorders, by utilizing the power of deep learning. The
importance of reliable sleep apnea event detection in enhancing patient outcomes and directing treatment choices is emphasized in
the paper. The model can automatically learn and extract useful features from sleep signals according to the deep learning
techniques used in the study, which facilitates efficient event classification. The findings show the potential of deep learning for
improving the diagnosis and management of sleep disorders by demonstrating promising performance in recognizing sleep apnea
occurrences. In conclusion, the study proposes a novel method for employing deep learning to identify sleep apnea occurrences. The
model efficiently analyses and categorizes sleep signals to detect sleep apnea occurrences by utilizing CNNs and LSTM networks.
The suggested approach gives encouraging accurate findings and has the potential to enhance the identification and treatment of
sleep disorders, thereby enhancing patient care [9].
The study paper “Diagnostic Accuracy of Obstructive Airway Adult Test for Diagnosis of Obstructive Sleep Apnea” focuses on
assessing the diagnostic precision of the Obstructive Airway Adult Test (OAAT) in identifying OSA. The common sleep disorder
OSA is characterized by recurrent bouts of whole or partial obstruction of the upper airway while sleeping, which can cause
breathing problems and pose health hazards. For the illness to be properly managed and treated, an accurate OSA diagnosis is
essential. The goal of the study is to evaluate how well the OAAT performs as a screening tool for OSA by contrasting its findings
with those of the gold-standard diagnostic technique, polysomnography (PSG). They then contrasted the findings to ascertain the
OAAT’s diagnostic precision. The study’s conclusions show that the OAAT has good diagnostic precision for detecting OSA. The
test has a relatively high degree of specificity and is helpful in identifying upper airway blockage, a major aspect of OSA. This
implies that OAAT can be used as a trustworthy and affordable screening tool, enabling medical practitioners to recognize people at
risk for OSA and to facilitate early intervention and treatment. The authors place a strong emphasis on the use of readily available,
non-invasive diagnostic instruments like the OAAT in advancing OSA diagnosis and treatment [10].

V. SYSTEM REQUIREMENT
A. Software Requirements
1) Operating System: Windows 10
2) Front End: HTML, CSS, Bootstrap, JavaScript
3) Back End: Python, Flask
4) IDE: VS Code, PyCharm

B. Hardware Requirements
1) Processor: Intel Core I3 or Higher
2) RAM: Minimum 2GB and recommended 8GB or More
3) Operating System: Windows, IOS, Android, Linux
4) Bandwidth: 2 Mbps or more

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 6116
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue V May 2023- Available at www.ijraset.com

VI. METHODOLOGY
1) Data Gathering: This entails gathering clinical and physiological data from sleep examinations, such as polysomnography,
from people who may have obstructive sleep apnea. Clinical data include demographics and medical history.
2) Data Preprocessing: This involves cleaning and preprocessing the collected data, which may include data normalization,
feature extraction, and feature selection.
3) Development of a Machine Learning Model: Based on the preprocessed data, a machine learning model is created to predict
the existence and severity of obstructive sleep apnea. The model may employ a number of techniques, including deep neural
networks, support vector machines, logistic regression, and random forests.
4) Model Evaluation: In this step, the effectiveness of the machine learning model is assessed using the right metrics, including
accuracy, AUC, sensitivity, specificity, and F1 score. For this, hold-out validation techniques as well as cross-validation
techniques may be employed.
5) Model Optimization: Model optimization entails enhancing the performance of the machine learning model by modifying its
hyperparameters or by employing methods like ensemble learning, feature engineering, or transfer learning.
6) Implementation and interpretation: This integrating the machine learning model into a program or system that can forecast
obstructive sleep apnea. Clinicians should be able to input patient data and obtain the model's output through the system's
interface. In order to promote openness and interpretability, the system should also be able to explain how the model arrived at
its forecast.
7) Validation and Monitoring: Monitoring the performance of the deployed model over time to make sure it continues to be
accurate and dependable is part of the validation and monitoring process. To ensure its efficacy, the model should be
periodically reviewed and changed as required.

A. Architecture Diagram

Fig 5.1: Architecture Diagram

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 6117
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue V May 2023- Available at www.ijraset.com

In the above diagram, we have experimented on obstructive sleep apnea prediction. From first step, user want to enter the medical
attributes then based on inputs, the machine learning-based OSA prediction to determine if a patient has four states of obstructive
sleep apnea or not.
In next step we have imported the dataset. Once we have imported the dataset then visualization of each inputted data takes place.
After visualization the next step pre-processing of data takes place where we check outliers, missing values and also scale the
dataset then on the updated dataset we split the data into training and testing. Next is on the training dataset we had applied various
like random forest algorithm. Then we build a pickle file for disease. The pickle file is integrated with the Flask framework for the
on the webpage.

VII. IMPLEMENTATION
A. Algorithm Used
Random Forest Algorithm:
An effective supervised learning method is Random Forest, a well-known machine learning algorithm. Both Classification and
Regression issues in ML can be solved with it.
It is based on the idea of ensemble learning, which is the act of integrating different classifiers to address a complicated issue and
enhance the performance of the model.
As the name suggests, "A random forest is a classifier that takes a set of decision trees for different subsets of a given dataset and
averages them to improve the prediction accuracy of that dataset."
Instead of depending on a single decision tree, the random forest takes the prediction from each tree and guesses the result based on
the predictions that have received the most votes.
First, N decision trees are combined to generate the random forest, and then predictions are made for each tree that was produced in
the first phase.

The stages and graphic below can be used to demonstrate the working process:

Step 1: Pick K data points at random from the training set.

Step 2: Create the decision trees linked to the subsets of data that have been chosen.

Step 3: Select N for the size of the decision trees you wish to construct.

Step 4: Repeat the steps 1 and 2

Step 5: Assign new data points to the category that receives the majority of votes by looking up each decision tree's
predictions for the new data points.

B. Performances Metrics Formula

 = ( + )/( + + + )
 Prec = ( )/( + )
 = ( )/( + )
 1 = (2 + + )/( + )
 = − ∑ log( ) ∗ ( ) + (1 − ) ∗ log(1 − )

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 6118
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue V May 2023- Available at www.ijraset.com

Fig 6.1: Confusion Matrix

VIII. RESULT
In the system, we have used Machine learning especially random forest algorithm for the prediction. When the patient will input the
values in system according to that it will show whether the patient has a disease or not. The parameters will show the range of the
values needed. The entered value is not between the ranges or is not valid or is empty it will show the warning sign that add a
correct value.

Accuracy Precision Recall F1-Score Time

Random Forest 91 91 91 91 0.82

Decision Tree 83.1 86 90 86 0.04

Extra Trees 90.3 90 86 90 0.57

Logistic Regression 52.2 52 52 51 0.38

SVC 56.1 58 56 55 3.87

KNN 83.1 83 83 82 0.11

Gaussian NB 57.5 58 58 57 0.57

Linear SVC 31.3 30 31 21 1.32

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 6119
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue V May 2023- Available at www.ijraset.com

Fig. 8.1: Welcome Page

Fig. 8.2: Disease Information field with payment field

Fig. 8.3: Disease Input Data field

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 6120
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue V May 2023- Available at www.ijraset.com

Fig. 8.4: Disease Prediction Normal Stage Output

Fig. 8.5: Disease Prediction Mild Stage Output

Fig. 8.6: Disease Prediction Moderate Stage Output

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 6121
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue V May 2023- Available at www.ijraset.com

Fig. 8.7: Disease Prediction Severe Stage Output

Fig. 8.8 Disease Prediction Feedback form

IX. CONCLUSION
Predicting the disease earlier can improve the patients’ health. The aim of this project is to predict obstructive sleep apnea diseases
based on patient inputs. The project is built in such a way that the system takes the patients measurements as input and generates an
output, which is nothing but the disease prediction. This model can help to reduce the cost required in dealing with this disease and
also help to improve the recovery process. By using this system patient can reduce the money required for treatment and can save
the time. In this review, we addressed the different machine learning (ML) methods that have been applied to OSA prediction,
including the logistic regression, random forest, convolutional neural network, and deep neural network. These algorithms provide
tailored prediction models for specific individuals based on a number of physiological and clinical variables, including demographic
information, anthropometric measurements, and polysomnography results.
In conclusion, ML algorithms have the potential to increase the precision and effectiveness of OSA diagnosis. To verify these
models in larger and more diverse populations and to address the difficulties in interpreting and applying these models in clinical
practice, additional research is required. The field of sleep medicine could be completely changed by machine learning algorithms,
which would also benefit patients with OSA.

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 6122
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.538
Volume 11 Issue V May 2023- Available at www.ijraset.com

Overall, these papers demonstrate the potential of machine learning algorithms in predicting OSA, and highlight the importance of
incorporating both physiological and clinical features in the models. The latest research is showing that the combination of different
machine learning algorithms and the use of novel features can improve the performance of OSA prediction models, which could
have significant implications for the diagnosis and treatment of this sleep disorder. This model can help to reduce the cost required
in dealing with this disease and also help to improve the recovery process. By using this system patient can reduce the money
required for treatment and can save the time.
X. FUTURE SCOPE
The future scope of obstructive sleep apnea (OSA) prediction using machine learning is quite promising, and there are several
potential directions for further research and development in this field. Possible future directions include the following, among
others.
1) Customized OSA Prediction: Presently, OSA prediction models are based on data from the general population and may not take
into account physiologic variations between individuals and other OSA-influencing variables. It might be possible to create
tailored OSA prediction algorithms in the future that use patient-specific data to raise the predictive accuracy.
2) Real-time Monitoring: OSA prediction models can be used to monitor patients and alert medical personnel to any sudden
changes in OSA risk or severity. These real-time monitoring devices can aid clinicians in the early detection of complications
and the avoidance of those complications.
3) Wearable Tech: Wearable tech can be used to gather physiological data in a non-invasive manner and includes smartwatches,
fitness bands, and other sensors. Future wearable technology may incorporate OSA prediction algorithms, allowing for real-
time feedback and continuous monitoring.
4) Integration with Telemedicine: OSA prediction models can be integrated into telemedicine systems to provide remote OSA
monitoring and diagnosis. Telemedicine is quickly expanding. In places where access to sleep clinics is limited, this can be
extremely useful.
5) Prediction Models that Combine Many Factors: Obesity, cardiovascular disease, and diabetes are frequently co-occurring
conditions with OSA. Future prediction models may incorporate information from several medical problems to offer a more
thorough evaluation of OSA risk and severity.
In future, we can add more features in exiting prediction system. We can try to increase the accuracy of prediction of disease to
reduce the morality. Also, we can try to make the system more user friendly by adding new features.
REFERENCES
[1] Lim J, Alshaer H, Khan SS, Pandya A, Ryan CM, Haleem A, Sivakulam N, Sahak H, Haq AU, Macarthur K. Diagnosis of Obstructive Sleep Apnea during
Wakefulness Using Upper Airway Negative Pressure and Machine Learning. Annu Int Conf IEEE Eng Med Biol Soc. 2019 Jul; 2019:1605-1608. doi:
10.1109/EMBC.2019.8856754. PMID: 31946203.
[2] Stepheny Lucas1, Mitali Desai, Amisha Khot1, Sincee Harriet, Nilambari Narkar,” SmartCare: A Symptoms Based Disease Prediction Model Using Machine
Learning Approach”, IJRASET, 10(11) - 11 Nov 2022.
[3] P. Hema, N. Sunny, R. Venkata Naganjani and A. Darbha,” Disease Prediction using Symptoms based on Machine Learning Algorithms,” 2022 International
Conference on Breakthrough in Heuristics and Reciprocation of Advanced Technologies (BHARAT), Visakhapatnam, India, 2022, pp. 49-54, doi:
10.1109/BHARAT53139.2022.00021
[4] Rim Haidar, Irena Koprinska, Bryn Jeffries,” Sleep Apnea Event Prediction Using Convolutional Neural Networks and Markov Chains”, 2020 International
Joint Conference on Neural Networks (IJCNN) IEEE - 28 September 2020.
[5] Rinkal Keniya, Aman Khakharia, Vruddhi Shah, Vrushabh Gada, Ruchi Manjalkar, Tirth Thaker, Mahesh Warang, Ninad Mehendale,” Disease prediction from
various symptoms using machine learning”, 2018 International Conference on Machine Learning and Data Engineering (iCMLDE) IEEE January 2019.
[6] A. Pavani1, Kalyanapu Srinivas,” Obstructive Sleep Apnea Prediction Using Deep Learning Techniques”, International Journal for Research in Engineering
Application Management (IJREAM), ISSN: 2454-9150, Vol-07, Issue-02 MAY 2021.
[7] Ming-Feng Wu, Wei-Chang Huang, Chia-Feng Juang, Kai ‘Ming Chang, Chih-Yu Wen, Yu-Hsuan Chen, Ching-Yi Lin, Yi-Chan Chen, and Ching-Cheng
Lin.,”A new method for self-estimation of the severity of obstructive sleep apnea using easily available measurements and neural fuzzy evaluation system”,
IEEE journal of biomedical and health informatics, 21(6):1524–1532, 2017 A.R. Troncoso-Garc´ıa, M. Mart´ınez-Ballesteros,F. Mart´ınez-Alvarez, ´ A.
Troncoso, ”Explainable machine learning for sleep apnea prediction”, 26th International Conference on Knowledge-Based and Intelligent Information
Engineering Systems (KES 2022).
[8] A.R. Troncoso-Garc´ıa, M. Mart´ınez-Ballesteros,F. Mart´ınez-Alvarez, ´ A. Troncoso,” Explainable machine learning for sleep apnea prediction”, 26th
International Conference on Knowledge-Based and Intelligent Information Engineering Systems (KES 2022).
[9] Sleep Mahmood Abed, Turgay Ibrikci,” Apnea Events Detection Using Deep Learning Techniques”, Journal of Sleep Disorders: Treatment and Care, 10
February 2023, DOI: 10.4172/2325-9639.23.12.101
[10] Gasparini, G., Vicini, C., De Benedetto, M., Salamanca, F., Sorrenti, G., Romandini, M., Bosi, M., Saponaro, G., Foresta, E., Lafor`ı, A., Meccariello, G.,
Bianchi, A., Toraldo, D. M., Campanini, A., Montevecchi, F., Rizzotto, G., Cervelli, D., Moro, A., Arigliani, M., Gobbi, R., Pelo, S., ”Diagnostic Accuracy of
Obstructive Airway Adult Test for Diagnosis of Obstructive Sleep Apnea” , Volume 2015 - Article ID 915185 - https://doi.org/10.1155/2015/915185

©IJRASET: All Rights are Reserved | SJ Impact Factor 7.538 | ISRA Journal Impact Factor 7.894 | 6123