Enhancing Liver Cirrhosis Diagnosis Using Machine Learning With Explainable AI and Cross-Validated Hyperparameter Tuning Techniques

Abstract

Cirrhosis is a form of Chronic Liver Disease (CLD) resulting from sustained liver damage from several causes, including viral infection, autoimmune disorders, cholestatic and metabolic disease Non-Alcoholic Fatty Liver Disease (NAFLD) or heavy alcohol use. Testing for the presence of Liver Cirrhosis (LC) is one of the most critical diagnostics and prognostic assessments for patients. Detecting liver illness in its early stages is challenging as initial symptoms are often mild and difficult to diagnose. Cirrhosis may not present symptoms until an advanced stage, posing a challenge for identification and diagnosis. The lack of explanations for a complex model’s decision-making process creates clinician distrust by concealing the reasoning behind its choices. This research sought to design and assess a novel and non-invasive proposed model by utilizing the XGBoost classifier with Explainable Artificial Intelligence (XAI) for diagnosing liver cirrhosis. In this work XAI framework enhances transparency in complex models, offering valuable insights into decision-making, thereby guiding treatment strategies, and improving cirrhosis diagnosis. The experimental results have shown superiority for the proposed method compared with various machine learning models. The proposed method has achieved significantly 91.19% using XGBoost for cirrhosis dataset with K-fold validation(K=10) approach.