A combination of optimized threshold and deep learning-based approach to improve malaria detection and segmentation on PlasmoID dataset

Malaria is a life-threatening parasitic disease transmitted to humans by infected female Anopheles mosquitoes. Early and accurate diagnosis crucial reduce the high mortality rate of disease, especially in eastern Indonesia, where limited health facilities resources contribute effortless spread disease. In rural areas, lack trained parasitologists presents significant challenge. To address this issue, computer-aided detection (CAD) system for malaria needed support evaluating hundreds blood smear slides every month. This study proposes hybrid automated parasite segmentation method using image processing deep learning techniques. First, an optimized double-Otsu proposed generate patch candidates. Then, approaches are applied recognize segment parasites. The evaluated on PlasmoID dataset, which consists 468 malaria-infected microscopic images containing 691 parasites from Indonesia. results demonstrate that our approach achieved F1-score 0.91 detection. Additionally, it better performance terms sensitivity, specificity, compared original semantic methods. These findings highlight potential be implemented CAD detection, could significantly improve resource-limited areas.