Scribd
Upload
18 views

Final Year Project

Uploaded by

danielrukundo312
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
18 views

Final Year Project

Uploaded by

danielrukundo312
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Explainable Machine Learning Approaches for

Liver Cirrhosis Diagnosis: Integrating Prediction


Models, Computer Vision, and a
Retrieval-Augmented Generation (RAG)
Application
BWAYO CHRISPUS BEN KYOMUGASO JOVIN
22/U/5909 22/U/6223
2200705909 2200706223

SENYONJO HENRY NAMUWAYA WINFRED


22/U/22641 22/U/5152/PS
2200722641 2200705152

1 Introduction/Background and Motivation


Liver cirrhosis is a chronic, progressive condition characterised by fibrosis (scar-
ring) of the liver, often resulting from chronic liver diseases such as viral hepati-
tis, alcoholism, and non-alcoholic fatty liver disease (NAFLD) [6]. Cirrhosis is a
leading cause of morbidity and mortality worldwide, with millions of deaths at-
tributed to liver complications each year. Early detection and intervention can
significantly improve patient outcomes, yet current diagnostic methods, such as
liver biopsies, remain invasive and expensive, limiting their accessibility, partic-
ularly in resource-constrained environments.
In recent years, advancements in artificial intelligence (AI), machine learn-
ing (ML) [1][2], and computer vision [5] have shown promise in tackling these
challenges by enabling accurate, non-invasive diagnostic methods. ML-driven
image analysis tools, for instance, have been used to detect liver abnormalities
through the examination of medical images, offering clinicians an alternative to
traditional diagnostic procedures. However, despite the potential of these tech-
nologies, many models remain ”black boxes”—their decision-making processes
are not easily interpretable or explainable, which undermines trust in ML-driven
clinical decisions.

1
This project aims to address both the diagnostic and explainability gaps
in liver cirrhosis diagnosis through a novel approach that integrates explain-
able machine learning models, computer vision techniques [5], and a Retrieval-
Augmented Generation (RAG) application. The machine learning and com-
puter vision models will analyse medical images and diagnostic data to accu-
rately predict liver cirrhosis, while the RAG application will provide real-time,
evidence-based information on liver cirrhosis, extracted from medical literature
and other reliable sources. The RAG system will also help in explaining ML-
generated diagnostic predictions, offering clinicians a deeper understanding of
both the disease and the ML model’s reasoning behind the diagnosis.

2 Problem Statement
Liver cirrhosis, marked by irreversible liver damage and fibrosis, poses a severe
health risk. Traditional diagnostic methods, such as liver biopsies, are invasive,
costly, and often unavailable, particularly in resource-limited settings. While
machine learning (ML) and computer vision [5] offer promising non-invasive
diagnostic tools, their effectiveness is limited by a lack of explainability, which
can undermine clinical trust and application.
This project seeks to address the need for a reliable, non-invasive method
for diagnosing liver cirrhosis and the challenge of ensuring transparency in
ML models. The proposed solution involves developing an integrated system
that combines explainable ML models, computer vision techniques [5], and a
Retrieval-Augmented Generation (RAG) application. The RAG application
will enhance diagnostic accuracy by leveraging predictive analytics to provide
detailed, evidence-based information on liver cirrhosis.

3 Main Objective
To develop an explainable, non-invasive diagnostic tool for liver cirrhosis related
to chronic alcohol consumption by integrating machine learning models, com-
puter vision techniques, and a Retrieval-Augmented Generation (RAG) appli-
cation to enhance diagnostic accuracy, transparency, and accessibility in clinical
settings.

4 Specific Objectives
• Design and develop explainable machine learning models for predicting
liver cirrhosis, with a focus on cases linked to chronic alcoholism.
• Apply computer vision techniques to analyse medical imaging and diag-
nostic data for liver cirrhosis detection.
• Integrate a Retrieval-Augmented Generation (RAG) application to pro-
vide evidence-based information during the diagnostic process.

2
• Evaluate the diagnostic accuracy, transparency, and usability of the sys-
tem in real-world healthcare settings.
• Ensure accessibility and applicability of the diagnostic tool in resource-
limited environments.

5 Methodology
This project utilises machine learning models, computer vision techniques, a
Retrieval-Augmented Generation (RAG) [3] application, and a web interface to
create an explainable diagnostic tool for liver cirrhosis, focusing on accuracy,
transparency, and ease of use for clinicians.

5.1 Data Collection and Preprocessing


Patient records, including clinical and health metrics, will be gathered and pre-
processed, addressing missing data and normalising values. Medical imaging
data, such as ultrasound and CT scans, will undergo preprocessing steps like
noise reduction, segmentation, and feature extraction, preparing the data for
analysis by the computer vision models.

5.2 Machine Learning Models for Numerical Data


Various machine learning models (e.g., logistic regression, random forests, or
neural networks) will be developed to predict liver cirrhosis based on clinical
data. Explainability techniques such as SHAP or LIME will be incorporated to
provide clear, interpretable insights into the model’s decision-making process.

5.3 Computer Vision for Medical Imaging


Multiple computer vision models will be implemented to analyse liver images
and detect cirrhosis. Explainability methods such as Grad-CAM will be applied
to highlight critical areas of the images that influence the models’ decisions,
making the results more interpretable for clinicians.

5.4 Retrieval-Augmented Generation (RAG) Application


The RAG [3] application will serve as a knowledge retrieval tool, providing rel-
evant medical literature and case studies on liver cirrhosis to assist in diagnosis.
This system will complement the machine learning and computer vision models
by offering evidence-based information to support clinical decision-making.

6 Expected Results
The project is expected to yield a highly accurate, non-invasive diagnostic tool
for liver cirrhosis. By utilising machine learning models and computer vision

3
techniques, the tool will improve diagnostic precision. Explainability will be a
key outcome, as methods such as SHAP, LIME, and Grad-CAM will make the
models more transparent.
The integration of a Retrieval-Augmented Generation (RAG) application is
expected to provide clinicians with relevant medical literature, improving the
overall clinical workflow.

7 References
References
[1] A.K.M Sazzadur Rahman, F. M. Javed Mehedi Shamrat, Zarrin Tasnim,
Joy Roy, Syed Akhter Hossain, A Comparative Study On Liver Disease
Prediction Using Supervised Machine Learning Algorithms, International
Journal of Scientific & Technology Research, Volume 8, Issue 11, November
2019.

[2] Vyshali J Gogi, Dr. Vijayalakshmi M.N, Prognosis of Liver Disease: Us-
ing Machine Learning Algorithms, International Conference on Recent
Innovations in Electrical, Electronics & Communication Engineering -
(ICRIEECE), 2018.

[3] Ge, Jin et al., Development of a liver disease–specific large language model
chat interface using retrieval-augmented generation, Hepatology, 2024.
[4] Linardatos, P. et al., Explainable AI: A Review of Machine Learning Inter-
pretability Methods, Entropy, 2021.
[5] Lu, Henry Horng-Shing et al., Computer-aided diagnosis of liver cirrhosis
by simultaneous comparisons of the ultrasound images of liver and spleen,
Journal of Data Science, 2008.
[6] Guo A, Mazumder NR, Ladner DP, Foraker RE, Predicting mortality among
patients with liver cirrhosis in electronic health records with machine learn-
ing, PLoS ONE, 2021.

You might also like