Improved Bare-Bones Artificial Bee Colony Optimization (IBB-ABC) and Enhanced Ensemble Classifier for Leukemia Diagnosis

Abstract

In medical image processing, an emerging and noticeable area of research is leukaemia detection, which is a malignant neoplastic disorder.  Recent works proposes an edge and ensemble classifier techniques to classify Leukaemia Diagnosis. Highly discriminative characteristics are identified using Bare-bones with Adaptive Bat Optimization (BBABO) and classification is also done by using ensemble based techniques. However, classifier accuracy is not enhanced up to the required level in the enhancing results of segmentation, the new classifiers are developed. An intelligent decision support system is proposed using microscopic images. Highly significant discriminative characteristics of blast and health cells are identified using Bare-bones with Adaptive Bat Optimization (IBB-ABC) for enabling effective ALL classification. Accelerated differential evaluation functions mechanisms of bee position update is incorporated in IBB-ABC variant for expanding search and original BBABC algorithm’s premature convergence is eased. Online Learning with Enhanced Support Vector Machine (OLERSVM), RBF network (RBFNets) and Improved Convolutional Neural Network (ICNN) are used for classifying blasted and healthy cells. Various Average rule is used for combining ensemble of classifiers. Proposed classifier’s experimentation results are compared with various leukaemia detection methods. On leukaemia image collections, with respect to f-measure, accuracy, recall and precision, better performance is shown by proposed method as demonstrated in qualitative and quantitative analysis.