Automatic Normal/Infected Blood-Cell Analysis for Malaria Detection with Lightweight Deep-learning

Malaria is a vector-borne disease and timely recognition and treatment is necessary to cure the patient. Untreated malaria will lead to various health illnesses, including death. Clinical level analysis of the malaria is commonly performed using the blood-teat and this research works proposed a computerized scheme for blood-cell analysis. The proposed research work consists the following phases; labeled image collection and resizing, feature extraction using a chosen Lightweight Deep-learning Model (LDM), performance evaluation using 3-fold cross validation and confirmation. This work also proposed Ensemble of Deep-features (EOD) based classification and achieved better detection accuracy. The merit of proposed tool is verified using; Conventional DeepFeatures (CDF) and the EOD using the SoftMax classifier and the achieved experimental results are compared and the merit of the developed scheme is confirmed. The outcome of this study confirms that the proposed approach helps to get an accuracy of 100% when the EOD based classification is executed. In the future, it can be considered to detect the blood cell infection from the microscopic images collected from hospitals.