AI-Based Real-Time Detection of Leishmania Parasites in Microscopic Images

Abstract

Leishmaniasis is a parasitic disease transmitted by sandflies and remains a major health burden in tropical regions. Conventional diagnosis relies on microscopic examination of stained tissue samples, which requires expert interpretation and is subject to observer variability. To overcome these limitations, we developed and evaluated a deep learning–based real-time detection system for identifying Leishmania parasites in brightfield microscopy images. Tissue samples from PCR-confirmed cutaneous leishmaniasis cases were imaged using both a high-resolution laboratory microscope and a low-cost mobile device to ensure variability in image quality. Parasites were annotated with bounding boxes to train a YOLOv8 model. The trained model achieved 84% accuracy and sensitivity, 82% specificity, and 87% precision. A graphical user interface enables real-time analysis of live microscope streams, supporting practical use in both laboratory and resource-limited settings.