HEART - Hybrid ECG Analysis for Recognizing Chagas

Chagas disease, caused by Trypanosoma cruzi, remains a major global public health concern, particularly in endemic regions. Affecting an estimated 6.5 million individuals and causing nearly 10,000 deaths annually, the disease demands effective screening and diagnostic strategies. As part of the 2025 George B. Moody PhysioNet Challenge, we (Team DataNinyas) developed an algorithmic pipeline to prioritize patients for confirmatory testing using ECG data. Our system, HEART (Hybrid ECG Analysis for Recognizing Chagas), is a spectrogram-based convolutional neural network enhanced with a novel lead-aware attention mechanism, designed to detect subtle ECG signatures associated with Chagas disease. Our model achieved a mean Challenge Score of 19.9% during the challenge, placing our team 24th out of 41 participants. This performance demonstrates its effectiveness in ranking Chagas-positive patients among the top cases. These findings highlight the potential of algorithmic approaches to enhance diagnostic accuracy and reduce the morbidity associated with Chagas disease.