Parasitized Malarial Infected Cell Detection through Variance Scaling Kernel Initializer Conv2D

Medical pathologists physically examine Plasmodium falciparum parasites infected blood cells that have been infected through a magnifying glass as part of the traditional diagnostic process, which is time-consuming and liable to mistakes. CNN and advanced technology have recently been used to conduct research on optimizing this procedure to identify orientation and geometrical information from blood sample particle images. With this view, this research recommends Variance Scaling Kernel Initializer Conv2D (VSKIC) Sequential model for predicting the infected malaria cell. The Malaria Cell Images Dataset from KAGGLE, which contains 27,558 visuals of plasmodium parasites has been used for execution in the suggested VSKIC model. For the purpose of predicting a malarial cell infection, images are preprocessed and fitted using the VSKIC model. Three convolutional layers merged with max pooling layer, and three dense layers that are weight initialized with variance scaling kernel initializer are used as preprocessing steps on the malaria cell images for the purpose of identifying the presence of affected malarial tissue. The picture files were separated into training and evaluation images, and the training Malaria Cell images were used to evaluate the effectiveness of the VSKIC model along with the standard conV2D paradigms. While comparing the suggested model with the available CNN methods, the execution observations made with Python on a V100 Nvidia GPU workstation using 64 and 250 training intervals demonstrates that the proposed VSKIC model predicts with the higher accuracy of 98.96% and Area under curve of 97.8%.