Automated system for characterization and classification of malaria-infected stages using light microscopic images of thin blood smears.

摘要: Summary In this paper, we propose a comprehensive image characterization cum classification framework for malaria-infected stage detection using microscopic images of thin blood smears. The methodology mainly includes imaging Leishman stained slides, noise reduction and illumination correction, erythrocyte segmentation, feature selection followed by machine classification. Amongst three-image segmentation algorithms (namely, rule-based, Chan–Vese-based marker-controlled watershed methods), technique provides better boundary erythrocytes specially in overlapping situations. Microscopic features at intensity, texture morphology levels are extracted to discriminate infected noninfected erythrocytes. In order achieve subgroup potential features, techniques, namely, F-statistic information gain criteria considered here ranking. Finally, five different classifiers, Naive Bayes, multilayer perceptron neural network, logistic regression, regression tree (CART), RBF network have been trained tested 888 (infected noninfected) each features’ subset. Performance evaluation the proposed shows that higher accuracy recognition Results show top 90 ranked (specificity: 98.64%, sensitivity: 100%, PPV: 99.73% overall accuracy: 96.84%) 60 results 97.29%, 99.46% 96.73%) classification. Lay Description Malaria, being mosquito-borne disease, is triggered parasite. general, it causes fever, chills, flu-like illness. It transmitted from one patient healthy individual through bite female Anopheles mosquito. Plasmodium parasite species solely responsible malaria attack. reality, there four types parasites (Plasmodium falciparum, vivax, malariae, ovale). falciparum (P. falciparum) vivax vivax) common infection observed world population. P. more deadly than vivax. as an epidemic disease due its outbreak over or populations. situation, if late diagnosed, left undiagnosed untreated well, may develop severe complications even die. view this, efficient rapid diagnostic method essential. most frequent test manual peripheral It, fact, involves lot ambiguity inter-observer variability leading error. addition, sometimes clinicians not available time. aim automate process smears without human intervention under medical analytics framework. From images, (parasite non-infected) automatically segmented quantified based characteristics. Mathematical measures adopted select good (discriminative nature) accelerate system's performance. Our establish achieves highest minimum number comparison with other classifiers combinations.