Betti-Number Based Machine-Learning Classifier Frame-work for Predicting the Hepatic Decompensation in Patients with Primary Sclerosing Cholangitis

This paper proposes a computationally efficient method for estimating the topology of manifold data in the context of medical applications. Betti numbers computed with persistent homology tools can be more useful for hepatic decompensation prediction in patients with Primary Sclerosing Cholangitis. We propose an alternative method that uses Betti numbers to estimate hepatic decompensation status. The results show that the proposed methodology is capable of distinguishing between hepatic decompensation and non-hepatic decompensation status. We discovered that using a Betti number-based machine-learning approach, we can make accurate predictions from small datasets, such as predicting who is likely to have hepatic decompensation and those who do not.