Microscopic Plasmodium classification (MPC) using robust deep learning strategies for malaria detection
Rapti Chaudhuri, Shuvrajeet Das, Suman Deb · 2023
Pathogenic microbes cause harm to human lives. Very specific mention is Plasmodium species which belongs to such category of pathogen, biologically known as malaria parasite, the devastating cause of life loss. The presence of pathogenic Plasmodium microbes are done by optical blood sample analysis through microscope. Manual identification of microscopic pathogen is a challenging and time-consuming task with respect to their similar structural formation. Even through digital microscope, identification of such pathogenic structures depends on several complex factors which is always dominated by human limitation over a long period of sample scanning. Machine-driven identification of such pathogenic microbes will be an additive benefit to speedy identification of probable presence of Plasmodium in human RBC. The machine-driven classification, segmentation and identification by applying deep learning techniques can be incorporated for obtaining near-perfect identification solution. The closed approximation inference has been carried out by applying convolutional neural network models for proper classification followed by precise identification of variant Plasmodium species in a single slide. Classification models such as SE_ResNet, ResNeXt, MobileNet and XceptionNet are studied extensively and applied on taken dataset after data preprocessing, augmentation and regularization, with state of art comparison. Resultant analysis has been done graphically and numerically as well for attaining desired parametric conditions. The aforesaid models are mainly considered here for their confirmed reliability and consistency in producing saturated results relative to the concerned data type and constrained parameterized structure. During experiment the proposed methodology has resulted in the identification of pathogenic Plasmodium microbes in an optimum amount of time and classifying the type of Plasmodium parasite to its exact class, working as a state of decision support system in medical pathology.