A Hybrid EPO-SVM Model for Efficient Anaemia Classification

Abstract

Anemia, a prevalent blood disorder affecting approximately 1.62 billion people worldwide, represents a significant global health challenge requiring accurate and efficient diagnostic methods. Traditional manual analysis of peripheral blood smear (PBS) images for anemia classification is time-consuming, error-prone, and requires specialized expertise. This paper presents a novel hybrid approach that integrates the Emperor Penguin Optimizer (EPO) algorithm for feature selection with Support Vector Machine (SVM) classifier to achieve efficient anemia classification from microscopic red blood cell (RBC) images. The proposed methodology addresses the critica challenge of high-dimensional feature spaces in medical image analysis by employing EPO's bio-inspired optimization capabilities to reduce feature dimensionality from 121 extracted features to approximately 64 optimal features. The hybrid EPO-SVM mode demonstrates superior performance in terms of classification accuracy, reduced computational complexity, and enhanced diagnostic efficiency. Experimental results show significant improvements in accuracy (93.2%) while substantially reducing training time (51.3% reduction) compared to traditional approaches using full feature sets. The integration of EPO's huddling behavior-inspired optimization with SVM's robust classification capabilities provides a promising solution for automated anemia diagnosis, contributing to more accessible and reliable healthcare diagnostics.