Decision Support Technique for Prediction of Acute Lymphoblastic Leukemia Subtypes Based on Artificial Neural Network and Adaptive Neuro-Fuzzy Inference System

Abstract

Leukemia is usually characterized as a potentially fatal disease with different types and subtypes. It pretends as one of the major reason for death in developed and developing nations like Bangladesh, almost a hundred thousand people around the world are losing the fight against leukemia in every decade. An early prediction and diagnosis of a leukemia type have become inexorable in treatment, as it can reduce deadly risk to the valuable life of affected patients. This paper's objective is to design and development a smart decision support prediction system of Acute Lymphoblastic Leukemia Subtypes using Artificial Neural Network (i.e. ANN) and Adaptive Neuro-Fuzzy Inference System (i.e. ANFIS). For this prediction system, 500 datasets (main database) of different subtypes of Acute Lymphoblastic Leukemia patients are used. Training and testing are the two main steps of ANN and ANFIS, where 445 and 55 datasets are elected for training and testing respectively. According to the experimental result, it is apparent that the Adaptive Neuro-Fuzzy Inference System (i.e. ANFIS) approach performs better than the Artificial Neural Network (i.e. ANN) in both training and testing. Analysis reveals that a simple and well ergonomic back-propagation learning technique of Artificial Neural Network (i.e. ANN) and a hybrid optimization learning technique of the Adaptive Neuro-Fuzzy Inference System (i.e. ANFIS) can easily predict blood disorder efficiently and effectively in terms of hematological parameters.