Applying Machine Learning to Evaluate for Fibrosis in Chronic Hepatitis C

Machine learning (ML) is proving to be an appealing analytical tool as modern medicine progresses towards preventative care. Clinical risk prediction models built using ML offer the potential for more effective diagnostic modalities without the need for invasive procedures. With such models, healthcare practitioners may be empowered towards a more preventative approach in management, thus improving clinical outcomes. The management of patients with chronic Hepatitis C is incomplete without considering the presence and extent of liver fibrosis, which is traditionally assessed with biopsy of liver tissue. Although non-invasive testing alternatives to liver biopsy are gaining popularity, they are considered limited due to inadequate accuracy and were designed to be complementary to liver biopsy. In this study, our aim is to build clinical risk models to predict the extent of fibrosis in patients with chronic Hepatitis C using ML algorithms. We developed nine ML algorithms based on an Egyptian cohort dataset, relying only on patient demographics and commonly-obtained serum laboratory values. One of our models was able to evaluate for fibrosis with an accuracy of 0.81, sensitivity of 0.95, and specificity of 0.73. Furthermore, most of our models outperformed many current diagnostic alternatives to liver biopsy for the evaluation of fibrosis in this patient population.