AI-Powered Diagnosis of Cirrhosis Through Medical Data Analytics and Machine Learning
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Liver cirrhosis remains a silent killer, which often progresses without symptoms until it develops into severe stages, and in that case, there are minimal treatment options, and prognosis is poor. The failure to detect early intervention is caused by the conventional diagnosis method that relies on the invasive liver biopsy or late emergence of the symptoms. This paper introduces a new AI-based diagnostic system that processes regular clinical and laboratory data and allows diagnosing liver cirrhosis at the early non-invasive stage. We use advanced data processing algorithms: iterative imputation, precise label encoding and SMOTE to balance classes and state-of-the-art ensemble learning with the help of the CatBoost algorithm. The use of the Explainable AI (XAI) methods like SHAP and LIME is an excellent development that provides transparent and understandable predictions, which are essential to implement in clinical practice. We achieve results of 97.4 percent, 96.8 percent and 98.1 percent accuracy, sensitivity, and specificity respectively when our system is evaluated on a large scale (614 patients) of 18 clinical features. The model can detect the years beforehand by identifying the subtle patterns of the common biomarkers that normally go undetected by the human clinicians. This is particularly useful in resource-limited settings because the solution offers a scalable tool of screening, which can revolutionize the control of cirrhosis and save many lives by intervening in the disease early.
Fields of Study & Subjects
Artificial Intelligence in Healthcare
Liver Disease and Transplantation
Liver Disease Diagnosis and Treatment
Publication details
- Journal / source
- International Conference on Computing, Communication, Control and Cyber-Physical Systems (I5CPS)
- Pages
- 1–6
- Publication date
- 26 February 2026
- Language
- EN