Next-Generation intelligent frameworks for accurate and interpretable liver disease detection

Abstract

The liver is an essential organ that performs many vital roles in the body. The malfunction of the liver can be life-threatening. Early detection of liver disease ensures faster treatment of the patients. However, the traditional classification approaches based on etiology have limitations due to the complexity and heterogeneity of liver diseases. However, fortunately, machine learning algorithms can overcome these limitations by identifying patterns and relationships within large data sets of patient information. The main objective of this study is to develop an interpretable soft-computing model that can classify patients as having liver disease or not. The study employed a liver function test (LFT) dataset. Six Support Vector Machine (SVM) models, five Deep Neural Network (DNN) models, and two regularized XGBoost models were trained. In this process, we have achieved the best accuracy of 96.28% in SVM models, 99.61% in NN models, and 99.94% in the L2 regularized XGBoost model. The concepts of Explainable AI like Shapely Values (SHAP) have been applied to each of the models, to make the results interpretable and explain what is going on inside these black box models. The study also conducts a thorough analysis of the dataset. These results demonstrate that the proposed approach is highly accurate and suitable for real-world liver disease detection.