A comparative approach for Parasite Detection via ML Algorithm

This study focuses on identifying the most effective feature selection technique for the enhanced and automated detection of malaria parasites using machine learning in wholeslide images of peripheral blood smears. Our approach involves extracting samples from these wholeslide images and conducting feature extraction. Various feature selection techniques are evaluated to determine their impact on the performance of the Support Vector Machine (SVM) when used as a binary classifier. For each method, the SVM is trained with the critical features identified. We utilize cross-validation and grid-search techniques to optimize SVM parameters. The trained SVM then classifies randomly selected but unknown "Normal" and "Infected" samples from the wholeslide images. Confusion matrices are created for each classification task, and the performance metrics for each task are documented. Our findings reveal that the SVM classifier achieves an outstanding accuracy rate of 95.5% when feature selection is based on the Kullback-Leibler distance between the two classes.