Malaria parasite detection and classification using CNN and YOLOv5 architectures

Abstract

Malaria is a severe global public health issue that is caused by the bite of an infected mosquito. It is curable, but only with early detection and prompt, effective treatment. It may lead to severe conditions if it is not properly diagnosed and treated in its early stages. In the worst-case scenario, it can result in death. The preferred methods for diagnosing malaria are microscopic analysis of blood samples or rapid diagnostic tests. This method is time-consuming and requires qualified medical personnel, who are currently in short supply and quite expensive. Detecting malaria from blood smears using this method is therefore extremely difficult. As a result of the rapid development of deep learning, numerous experts have attempted to implement it in the medical sector due to its ability to overcome the limitations of conventional methods. Therefore, the purpose of this paper is to design and develop a model for the detection of malaria parasites. Implementing a Convolutional Neural Network (CNN) and YOLOv5 algorithm to detect and classify malaria with the selected dataset, respectively, is the proposed method. This project will utilize a publicly accessible dataset containing all images of the malaria parasite. After training, the CNN model's accuracy in detecting infected blood images is 96.21 percent, and its performance will be evaluated and compared to the opinion of experts.