Comprehensive Study of Liver Disease Prediction using Machine Learning

Liver disease ranks as the eleventh leading cause of global mortality. A significant 75% of all liver cancer cases are reported in Asia, with Vietnam having the highest incidence rate of liver cancer. Diagnosing liver disease typically involves a combination of medical history assessment, physical examinations, and various diagnostic tests. Among these blood tests seems to be one of the most achievable way. This paper presents a framework for investigating the feasibility of applying machine learning for the automatic prediction of liver diseases from blood tests and demographic information of the patient. Furthermore, we determined the importance of features for each of the studied models and combined them with the ones recommended by doctors. We conduct extensive experiments to evaluate the selected features and machine learning models to classify liver diseases on three datasets (two benchmark datasets (ILPR and HCC) and one self-collected dataset vdLiver-small). Experiments show promising liver disease classification results with the highest sensitivity reaching 98.1 % on the vdLiver -small dataset. This framework could serve as a valuable tool for assigning doctors, especially young physicians, to help prevent missed or erroneous detection.