Loading
Life sciences · Preprint
arXiv · September 8, 2026
Early or partial results. Treat as a signal, not a conclusion.
This is a machine learning model development study that uses contrastive pre-training on paired ECG-cardiac MR data from a non-Chagas population (UK Biobank) to improve ECG-based detection of Chagas disease in endemic-region cohorts. The model shows improved performance over baseline ECG models in cross-validation and challenge benchmarks, but the work remains at an early methodological stage without prospective clinical validation or direct comparison to clinical decision-making.
Retrospective model development and validation using pre-trained contrastive learning. Pre-training: UK Biobank participants with paired ECG and cardiac MR examinations, no Chagas disease cases. Validation: Chagas-endemic cohorts CODE-15%, SaMi-Trop, SaMi-Trop-3, and ELSA-Brasil; specific eligibility criteria and clinical presentation not stated.. Intervention: Contrastive pre-trained ECG encoder aligned to CMR structural embedding space using asymmetric InfoNCE objective; frozen linear probe applied at validation stage.. Compared with: Unaligned ECG-FM (foundation model) baseline without CMR-derived structural knowledge.. UK Biobank (pre-training); endemic regions implied for validation cohorts (CODE-15%, SaMi-Trop, ELSA-Brasil) but not explicitly specified in text..
Aligned ECG-CMR model achieved AUROC of 0.851 on CODE-15% and SaMi-Trop cohorts in five-fold cross-validation, versus 0.827 for unaligned baseline Top 5% sensitivity (Top5%-TPR) of 0.427 with aligned model versus 0.377 for unaligned baseline Model obtained highest AUROC on SaMi-Trop-3 and best ELSA-Brasil challenge score among three top-performing methods on PhysioNet/CinC 2025 test set
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
If validated prospectively, this approach could improve ECG-based Chagas detection in resource-limited settings where MR imaging is unavailable. However, current evidence is limited to algorithm benchmarking; clinical utility, diagnostic accuracy in real-world workflows, and patient outcomes remain undemonstrated.
A machine learning method development study using pre-trained representations from unaffected populations (UK Biobank) to detect Chagas disease in external datasets; shows promising cross-validation and challenge performance but lacks prospective clinical validation, independent test cohort assessment, and clinical outcome data.
As stated by the source record.
Quoted from the source exactly as published.
If validated prospectively, this approach could improve ECG-based Chagas detection in resource-limited settings where MR imaging is unavailable. However, current evidence is limited to algorithm benchmarking; clinical utility, diagnostic accuracy in real-world workflows, and patient outcomes remain undemonstrated.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
Chagas disease is a major cause of cardiomyopathy in Latin America. Cardiac magnetic resonance (CMR) imaging can characterize its structural abnormalities, but scanners and expert readers remain scarce in endemic regions. Electrocardiography (ECG) is inexpensive and widely available, yet structural disease must be inferred indirectly from electrical signals. We propose to transfer CMR-derived structural knowledge to ECG through contrastive pre-training. Using 63,193 paired ECG-CMR examinations from the UK Biobank, we align an ECG encoder with a clinically grounded CMR embedding space using an asymmetric InfoNCE objective. Despite seeing no Chagas cases during pre-training, the resulting representation improves ECG-based Chagas detection. Across CODE-15% and SaMi-Trop, a frozen linear probe achieves an AUROC of 0.851 and sensitivity at the top 5% of predicted risk (Top5%-TPR) of 0.427 in five-fold cross-validation, compared with 0.827 and 0.377 for an unaligned ECG-FM baseline. On the PhysioNet/CinC 2025 Challenge test set, our model obtains the highest AUROC on SaMi-Trop-3 and the best ELSA-Brasil challenge score among the three top-performing methods, indicating that imaging-supervised ECG representations can generalize to populations and settings beyond the pre-training distribution.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.