Cascading YOLO: Automated Malaria Parasite Detection for Plasmodium Vivax in Thin Blood Smears.
Yang F, Quizon N, Silamut K, Maude RJ, Jaeger S, Antani SK
Proc. SPIE 11314, Medical Imaging 2020: Computer-Aided Diagnosis, 113141Q (16 March 2020); https://doi.org/10.1117/12.2549701.
Abstract:
Purpose: Malaria continues to be a major burden on global health, causing approximately half a millionfatalities each year [1]. However, malaria diagnosis from patient blood smears requires a significant amountof time and expertise. Malaria is caused by Plasmodium parasites including five species, of whichPlasmodium falciparum and Plasmodium vivax pose the greatest health threat [2]. The goal of this work isto develop an automated malaria parasite detection system for Plasmodium vivax. The automated andaccurate detection of Plasmodium vivax is made difficult due to the lower parasitemia levels typicallyobserved in these infections, as compared to Plasmodium falciparum, which tends to require microscopiststo image regions of thin blood smears that are dense with cells and challenging to segment. Contributions: Our main contribution is two-fold: First, in the literature, only a few papers have beenpublished on the automated identification of Plasmodium vivax [3-5], and we are the first to performevaluation on patient level. Second, we propose a cascade of YOLO model and AlexNet classifier for hardnegativemining to reduce the false positive errors of the parasite detection, which improves the meanaverage precision by about 8%. Methods: We develop a rapid and robust diagnosis system for the automated detection of Plasmodiumvivax parasites using a cascaded YOLO model. This system consists of a YOLOv2 model [6] and a classifierfor hard-negative mining. We first split images of 4032×3024×3 pixels into regions of 672×504×3 pixels,and train a YOLOv2 model with the regions and corresponding manual ground-truth annotations. To reducefalse positive errors, we cascade a transferred AlexNet classifier trained with annotated parasites and falsepositives (hard-negatives) generated by the YOLOv2 model. For testing, we also split blood smear imagesinto regions of 672×504×3 pixels, which are then screened for parasites using our cascaded YOLO model.The detected parasite coordinates are then re-projected to the original image space for visualization andevaluation.
Yang F, Quizon N, Silamut K, Maude RJ, Jaeger S, Antani SK. Cascading YOLO: Automated Malaria Parasite Detection for Plasmodium Vivax in Thin Blood Smears.
Proc. SPIE 11314, Medical Imaging 2020: Computer-Aided Diagnosis, 113141Q (16 March 2020); https://doi.org/10.1117/12.2549701.