Evolutionary-based intelligent decision model to optimize the liver fibrosis stadialization

Abstract

ABSTRACT. This paper presents a novel approach to build an intelligent decision system (IDS) inspired by the evolutionary paradigm in order to solve the automatic liver fibrosis stadialization by optimizing the decision-making process.  The evolutionary paradigm was used to answer the basic question: how to distinguish between machine learning algorithms facing a medical decision issue, and how to integrate the most effective of them into IDS, able to provide an optimum decision? In the proposed IDS, a set of well-known neural networks are regarded as the initial population of solutions, and an appropriate hierarchy of algorithms is established fitness-proportionally based on a statistically built fitness measure. Then, the IDS framework is built using the best algorithms and the paradigm of a weighted voting system. In a concrete application, the degrees of liver fibrosis, ranging from F0 (no fibrosis) to F4 (cirrhosis), have been automatically identified in 722 patients with chronic hepatitis C infection using 25 main medical attributes. The decision performance proved significantly superior to the classical approach using standalone algorithms. This approach showed a way to directly and easy optimize the medical decision-making by using the evolutionary paradigm.