Detection and classification of plasmodium parasites in human blood smear images using Darknet with YOLO
Donata D. Acula, J. A. P. Carlos, M. M. Lumacad, J. C. L. O. Minano, J. K. R. Reodica · IET conference proceedings. · 2023
Plasmodium parasites cause malaria, a deadly, contagious, and potentially fatal blood disease spread by mosquitoes. Under a qualified technician's microscope, the traditional and most common method of diagnosing malaria is visually examining human blood smears for parasite-infected red blood cells. The method, as mentioned earlier, is time-consuming and inefficient, and the diagnosis depends on the examiner's experience and knowledge. Automated image recognition techniques based on image processing were previously applied to malaria blood smears for diagnosis. However, the practical performance still needs to be improved. Hence, it encouraged researchers to develop malaria detection and diagnostics quickly, easily, and efficiently. The study's primary goal is to detect cells from images of multiple cells in human blood smears on microscope slides, detect whether a particular cell is infected, and identify parasite species and life cycle stages. YOLOv4 and Darknet53 models were trained to do as such, and were able to achieve a 98.71% for detection accuracy and 97.08%, and 97.36% for parasite type accuracy and life-cycle stage accuracy, respectively. The proposed model also achieved an average runtime of 237.55ms per image in detection, 66.5ms in type classification, and 58.75ms in life-cycle stage classification.