<scp>MOSQUITO‐NET</scp>: A deep learning based <scp>CADx</scp> system for malaria diagnosis along with model interpretation using <scp>GradCam</scp> and class activation maps

Abstract

Abstract Malaria is considered one of the deadliest diseases in today's world, which causes thousands of deaths per year. The parasites responsible for malaria are scientifically known as Plasmodium, which infects the red blood cells in human beings. Diagnosis of malaria requires identification and manual counting of parasitized cells in microscopic blood smears by medical practitioners. Its diagnostic accuracy is primarily affected by extensive scale screening due to the unavailability of resources. State of the art Computer‐Aided Diagnostic techniques based on deep learning algorithms such as CNNs, which perform an end to end feature extraction and classification, have widely contributed to various image recognition tasks. In this paper, we evaluate the performance of Mosquito‐Net, a custom made convnet to classify the infected and uninfected blood smears for malaria diagnosis. The CADx system can be deployed on IoT and mobile devices due to its fewer parameters and computation power, making it wildly preferable for diagnosis in remote and rural areas that lack medical facilities. Statistical analysis demonstrates that the proposed model achieves greater accuracy than the previous SOTA architectures for malaria diagnosis despite being 10 times lighter in parameters and inference time. Mosquito‐Net achieves an AUC of 99.009% and an F‐1 score of 96.7% on the validation set.