Artificial intelligence algorithm for real-time detection and counting of Trypanosoma cruzi parasites using smartphone microscopy.

Chagas disease affects 6-7 million people worldwide and causes approximately 12,000 deaths annually. Diagnostic methods vary by disease stage, with serological tests commonly used in the chronic phase, while microscopy and molecular techniques like PCR and LAMP are employed in the acute phase. While microscopy remains the most accessible tool in resource constrained settings, its effectiveness depends on skilled personnel, creating diagnostic bottlenecks. To overcome these limitations, we developed a portable, smartphone-integrated AI system for real-time Trypanosoma cruzi detection in microscopy images. The platform combines a 3D-printed microscope adapter which aligns the smartphone camera with the microscope ocular to digitize images, with telemedicine-enabled annotation workflows, and lightweight AI models (SSD-MobileNetV2, YOLOv8) deployed on smartphone for real-time analysis. Trained on a diverse dataset of human samples (478 images from 20 samples), including thick/thin blood smears and cerebrospinal fluid) and murine thin smears (570 images from 33 samples), the SSD-MobileNetV2 model achieved 86% precision, 87% recall, and 86.5% F1-score on human samples, demonstrating robust performance across variable imaging conditions. We additionally piloted a real-world experiment with the proposed system. Three thin blood smears were scanned by a user operating the smartphone-based system, with predictions generated in real time. Model outputs were benchmarked against expert annotations as the ground truth. At the object level, the algorithm achieved a precision of 67.1%, a recall of 96.4%, and an F1-score of 79.1%, showing high sensitivity under operational conditions with a configuration possibly suitable for screening. This system could enable rapid, accurate parasite detection in field settings without advanced infrastructure, addressing critical gaps in early diagnosis and monitoring. Its modular design allows adaptation to other pathogens and cellular structures, offering a scalable solution for neglected tropical disease diagnostics. By bridging AI innovation with microscopy, this approach holds promise for advancing equitable healthcare delivery in endemic regions and aligning with global health priorities.