Malaria parasite detection in thick blood smear microscopic images using modified YOLOV3 and YOLOV4 models

Abstract Background Manual microscopic examination of Leishman/Giemsa stained thin and thick blood smear is still the “gold standard” for malaria diagnosis. One drawbacks this method that its accuracy, consistency, diagnosis speed depend on microscopists’ diagnostic technical skills. It difficult to get highly skilled microscopists in remote areas developing countries. To alleviate problem, paper, we propose investigate state-of-the-art one-stage two-stage object detection algorithms automated parasite screening from image slides. Results YOLOV3 YOLOV4 models, which are detectors accuracy speed, not optimized detecting small objects such as parasites images. We modify these models by increasing feature scale adding more layers enhance their capability without notably decreasing speed. one modified model, called YOLOV4-MOD two YOLOV3, YOLOV3-MOD1 YOLOV3-MOD2. Besides, new anchor box sizes generated using K-means clustering algorithm exploit potential detection. The performance were evaluated a publicly available dataset. These have achieved exceeding original versions, Faster R-CNN, SSD terms mean average precision (mAP), recall, precision, F1 score, IOU. has best among all other with mAP 96.32%. YOLOV3-MOD2 96.14% 95.46%, respectively. Conclusions experimental results study demonstrate promising images captured smartphone camera over microscope eyepiece. proposed system suitable deployment low-resource setting areas.