A Novel Approach for Enhancing Malaria Detection Accuracy Through Deep Learning With C3TR and BiFPN Architectures

Each  year,  millions  of  people  experience  the  terrible  effects  of  malaria  in different parts of the world. Humans get sick with malaria through infected female Anopheles mosquitoes. Despite recent advancements in the field, microscopy remains the most common method for diagnosing malaria. Recent strides in the domains of signal processing, image processing, and sensing, through the application of artificial intelligence, especially machine learning and deep learning have yielded cost-effective solutions that improve diagnostic accuracy. In this letter, we propose a tweaked version of the You Only Look Once (YOLOv5) framework, incorporating image signal processing techniques for screening malaria-infected cells and comparing it with established methods that highlight the pivotal role of sensing technologies. Our innovation revolves around two pivotal modifications. The conventional C3 module in YOLOv5 is replaced by the innovative C3TR structure, leveraging signal processing to facilitate the extraction of precise object properties while minimizing the interference of the image background. The PANet structure in the neck has been upgraded to the state-of-the-art Bi-directional Feature Pyramid Network architecture harnessing image processing techniques to further extract the features of objects. The outcomes of our model are remarkable with an accuracy of 99.2%, a precision of 98.7%, and a recall of 98.5%. With these results, our model outperformed many of the existing methodologies which highlighted the crucial importance of image signal analysis and sensing technologies in transforming malaria detection and classification.