A Modified Deep Learning Architecture For Detecting MalariaInfected Cells

Malaria is a major public health problem because it is prevalent in tropical and subtropical regions with inadequate healthcare systems and limited resources. A correct diagnosis is necessary for prompt malaria intervention and treatment. Expert pathologists manually review blood smear images to obtain a microscopic diagnosis. Alternatively, they use a $50 \%$ accurate Rapid antigen malaria test. Advancements in Deep Learning(DL) have provided promising features in automating malaria diagnosis. Generating predictions based on patterns seen in large-scaledatasets. the DL approach, which is recommended work, provides a more reliable diagnosis by reducing the cost of diagnosis, especially in areas with limited resources. The approach we propose combines Image transformation, Feature extraction, Feature Engineering, and data augmentation using the Convolutional Neural Network(CNN) architecture with an accuracy of $94.44 \%$ in classification, and by incorporating a Support Vector Machine (SVM) function as a method of activation in the final layer of the CNN design, it influences the layers. Our objective is to improve deep neural networks’ performance in classifying cells that are affected by malaria parasite.