Malaria detection and classification using machine learning algorithms

Abstract

Malaria is one of the most infectious diseases, specifically in tropical areas where it affects millions of lives each year. Manual laboratory diagnosis of Malaria needs careful examination to distinguish infected and healthy Red Blood Cells (RBCs). However, it is time consuming, needs experience, and may face inaccurate lab results due to human errors. As a result, doctors and specialists are likely to provide improper prescriptions. With the current technological advancement, the whole diagnosis process can be automated. Hence, automating the process needs analysis of the infected blood smear images so as to provide reliable, objective result, rapid, accurate, low cost and easily interpretable outcome. In this paper comparison of conventional image segmentation techniques for extracting Malaria infected RBC are presented. In addition, Scale Invariant Feature Transform (SIFT) for extraction of features and Support Vector Machine (SVM) for classification are also discussed. SVM is used to classify the features which are extracted using SIFT. The overall performance measures of the experimentation are, accuracy (78.89%), sensitivity (80%) and specificity (76.67%). As the dataset used for training and testing is increased, the performance measures can also be increased. This technique facilitates and translates microscopy diagnosis of Malaria to a computer platform so that reliability of the treatment and lack of medical expertise can be solved wherever the technique is employed.