Biomarker-Based Pretraining for Chagas Disease Screening in Electrocardiograms

Elias Stenhede; Arian Ranjbar

cs.CVApr 10, 2026

Biomarker-Based Pretraining for Chagas Disease Screening in Electrocardiograms

Authors: Elias Stenhede, Arian Ranjbar

Organizations: Medical Technology & E-health, Akershus University Hospital, Lørenskog, Norway · Faculty of Medicine, University of Oslo, Oslo, Norway

Abstract

Chagas disease screening via ECGs is limited by scarce and noisy labels in existing datasets. We propose a biomarker-based pretraining approach, where an ECG feature extractor is first trained to predict percentile-binned blood biomarkers from the MIMIC-IV-ECG dataset. The pretrained model is then fine-tuned on Brazilian datasets for Chagas detection. Our 5-model ensemble, developed by the Ahus AIM team, achieved a challenge score of 0.269 on the hidden test set, ranking 5th in Detection of Chagas Disease from the ECG: The George B. Moody PhysioNet Challenge 2025. Source code and the model are shared on GitHub: github.com/Ahus-AIM/physionet-challenge-2025

Figures & tables

Biomarker	Clinical domain
Albumin	Liver function
Calcium, Total	Electrolytes
Creatinine	Renal
Hematocrit	Hematology
Hemoglobin	Hematology
INR(PT)	Coagulation

Table 1: Biomarkers included for model pretraining (alphabetical).

Dataset	# ECGs	# Patients	Step
MIMIC-IV-ECG	523,275	102,511	Pretraining
CODE15%	345,779	233,770	Fine-tuning
SaMi-Trop	1,959	1,959	Fine-tuning

Table 2: Datasets used for model development. In pretraining, not all ECGs were included, as each ECG had to occur within 24 hours of at least one included test.

Component	Specification
System	Debian 12
CPU	Intel Core i9-14900KF
RAM	2 $\times$ 48 GB; 4800 MT/s
GPU	NVIDIA GeForce RTX 5090
CUDA version	12.9
Programming language	Python 3.12

Table 3: Development environments and hardware used in model training.

Parameter	Value
Batch size	64
Optimizer	Muon and Adam
Muon momentum	0.95
Learning rate	0.0037
Loss function	Cross-entropy

Table 4: Training parameters for biomarker-based pretraining.

Parameter	Value
Batch size	128
Optimizer	Muon and Adam
Muon momentum	0.95
Learning rate	0.001
Loss function	Binary cross-entropy

Table 5: Parameters used during fine-tuning.

Parameter	Value
Number of blocks	6
Kernel sizes	9, 19, 39
Number of filters	32
Bottleneck channels	32

Table 6: Hyperparameters for the InceptionTime network.

Panel of Figure 1 — Figure 1: Shows the predicted probability distributions over blood test percentiles for four ECGs taken from four different patients (one colour per patient). The vertical lines represent the actual values for the same patients, measured within 24 hours of ECG recording. The selected biomarkers represent a subset of the biomarkers used in pretraining.

Team	Challenge score	Rank
Ahus AIM	0.269	5

Table 7: Challenge score on the hidden test set.

Electrocardiogram Classification

Electrocardiogram classification uses machine learning to identify heart rhythms and cardiac conditions from ECG waveforms for screening, diagnosis, and monitoring.

See all 106 papers

Leveraging Cardiac Imaging to Improve ECG-Based Detection of Chagas Disease in Resource-Constrained Settings
Sep 8, 2026Laura Alvarez-Florez, Daniel Uyterlinde, Samuel Ruipérez-Campillo +3Cross-Modal Representation Learning Medical Diagnosis
Department of Biomedical Engineering and Physics, Amsterdam University Medical Center, The Netherlands · Quantitative Healthcare Analysis Group, University of Amsterdam, The2026 Netherlands · Department of Clinical and Experimental Cardiology, Amsterdam University Medical Center, The Netherlands +2
Do ECG Foundation Models Transfer to Rare Cardiac Diseases? Evidence from Brugada Syndrome Detection
Jul 3, 2026Beatrice Zanchi, Giuliana Monachino, Alvise Dei Rossi +4Domain Generalization Electrocardiogram Foundation Models
University of Applied Sciences and Arts of Southern Switzerland, Department of Innovative Technologies, Via La Santa 1, Lugano, 6900, Ticino, Switzerland · University of Zurich, Department of Quantitative Biomedicine, Winterthurerstrasse 190, Zurich, 8057, Zurich, Switzerland · Neurocenter of Southern Switzerland, Ente Ospedaliero Cantonale, Movement Disorder Research Group, Neurology Department, Via Tesserete 46, Lugano, 6900, Ticino, Switzerland +5
Domain-Adapted Fine-Tuning of ECG Foundation Models for Multi-Label Structural Heart Disease Screening
Apr 25, 2026Duc N. Do, Minh N. Do, Dang Nguyen +14Unsupervised Domain Adaptation Echocardiography
1PASSIO Laboratory, North Carolina A&T State University, Greensboro, NC, USA · Department of Computer Science, University of Manitoba, Winnipeg, MB, Canada · 3Harvard T.H. Chan School of Public Health, Harvard University, Boston, MA, USA +4

Sep 8, 2026J/K move · Enter open · S save

Leveraging Cardiac Imaging to Improve ECG-Based Detection of Chagas Disease in Resource-Constrained Settings

Laura Alvarez-Florez, Daniel Uyterlinde, Samuel Ruipérez-Campillo, Lukas P. A. Arts+2

Department of Biomedical Engineering and Physics, Amsterdam University Medical Center, The Netherlands · Quantitative Healthcare Analysis Group, University of Amsterdam, The2026 Netherlands · Department of Clinical and Experimental Cardiology, Amsterdam University Medical Center, The Netherlands+2

Cross-Modal Representation Learning Medical Diagnosis Electrocardiogram Classification Electrocardiogram Interpretation

Open paper AskPDF arXiv

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.

Abstract

Electrocardiogram Classification

Explore similar work

Leveraging Cardiac Imaging to Improve ECG-Based Detection of Chagas Disease in Resource-Constrained Settings

Do ECG Foundation Models Transfer to Rare Cardiac Diseases? Evidence from Brugada Syndrome Detection

Domain-Adapted Fine-Tuning of ECG Foundation Models for Multi-Label Structural Heart Disease Screening