Comparison of AI Techniques for Prediction of Liver Fibrosis in Hepatitis Patients

Abstract

Globally one in twelve people have the Hepatitis B or Hepatitis C virus. Diagnosis and treatment of this disease is guided by liver biopsies where a small amount of tissue is removed by a surgeon and examined by a pathologist to determine the fibrosis stage from F₀ (no damage) to F₄ (cirrhosis). Biopsies are costly and carry some risk for the patient. Non-invasive techniques for determining fibrosis stage have been developed and evaluated since 2003. Non-invasive methods have utilized serum markers, imaging test, and genetic studies. The accuracy of these non-invasive techniques has not achieved sufficient acceptance and so the invasive biopsy is still considered the gold standard.

Clinical decision support systems (CDSS) use decision support system theory and technology to assist clinicians in the evaluation and treatment process. Using historical clinical data and the relationship processed by Artificial Intelligence (AI) techniques to aid physicians in their decision making process is the goal of CDSS. The CDSS provides a large number of medical support functions to help clinicians make the most reasonable diagnosis and choose the best treatment measures.

This paper applies four artificial intelligence predictive techniques to publicly available data on 424 Hepatitis B and Hepatitis C patients. Demographic and standard serum markers are utilized to predict fibrosis stage and compare these predictions to known biopsy results. A final decision tree evaluation is applied to make a final prediction. We have also developed a publically available web application that can be used as a prototype for presenting AI predictive results in a CDSS environment based on these models. This technique along with others could mitigate the need for some liver biopsies in the more than 500 million Hepatitis B and C patients worldwide with additional validation and verification.