Artificial Intelligence Approaches for Diagnosis and Continuous Monitoring of Chronic Liver Disorders

Abstract

Chronic liver disease, or CLD, is becoming a bigger problem in the world's health care system because it is hard to notice when it starts and is fatal in its later stages. It is difficult to detect the disease at an early stage since conventional diagnostic procedures including imaging, biopsies, and liver function tests are intrusive, expensive, or not widely available. This highlights the need for accurate, non-invasive, and data-driven approaches to support timely intervention and reduce clinical risks. Machine learning (ML) has shown promise in processing large, heterogeneous datasets to uncover hidden patterns and enhance predictive performance. Using the Liver Disease Patient Dataset from the UCI Repository, comprising 30,691 records with 11 attributes, this study applied extensive preprocessing, including missing value imputation, outlier removal with the IQR method, Min–Max normalization, and SMOTE for class balancing. Feature selection was employed to improve interpretability and efficiency. A model called Gradient Boosting (GB) was created and tested against many other methods, including SVM, Random Forest, MLP, and XGBoost. Surpassing baseline models, GB attained the best performance with a 98.60% ROC-AUC, 98.50% recall, precision, and F1-score. Based on these findings, ensemble approaches are reliable for predicting early CLD. Improving practicality will be the goal of future studies that investigate clinical validation and the integration of multimodal data