Demographic-Aware Self-Supervised Anomaly Detection Pretraining for Equitable Rare Cardiac Diagnosis

Chaoqin Huang; Zi Zeng; Aofan Jiang; Yuchen Xu; Qing Cao; Kang Chen; Chenfei Chi; Yanfeng Wang

cs.CVMar 20, 2026

Demographic-Aware Self-Supervised Anomaly Detection Pretraining for Equitable Rare Cardiac Diagnosis

Authors: Chaoqin Huang, Zi Zeng, Aofan Jiang, Yuchen Xu, Qing Cao, Kang Chen, Chenfei Chi, Yanfeng Wang, +1 more

Organizations: Shanghai Jiao Tong University, Shanghai, China · Shanghai AI Laboratory, Shanghai, China · National University of Defense Technology, Wuhan, China · Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China · Renji Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China

Abstract

Rare cardiac anomalies are difficult to detect from electrocardiograms (ECGs) due to their long-tailed distribution with extremely limited case counts and demographic disparities in diagnostic performance. These limitations contribute to delayed recognition and uneven quality of care, creating an urgent need for a generalizable framework that enhances sensitivity while ensuring equity across diverse populations. In this study, we developed an AI-assisted two-stage ECG framework integrating self-supervised anomaly detection with demographic-aware representation learning. The first stage performs self-supervised anomaly detection pretraining by reconstructing masked global and local ECG signals, modeling signal trends, and predicting patient attributes to learn robust ECG representations without diagnostic labels. The pretrained model is then fine-tuned for multi-label ECG classification using asymmetric loss to better handle long-tail cardiac abnormalities, and additionally produces anomaly score maps for localization, with CPU-based optimization enabling practical deployment. Evaluated on a longitudinal cohort of over one million clinical ECGs, our method achieves an AUROC of 94.7% for rare anomalies and reduces the common-rare performance gap by 73%, while maintaining consistent diagnostic accuracy across age and sex groups. In conclusion, the proposed equity-aware AI framework demonstrates strong clinical utility, interpretable anomaly localization, and scalable performance across multiple cohorts, highlighting its potential to mitigate diagnostic disparities and advance equitable anomaly detection in biomedical signals and digital health. Source code is available at https://github.com/MediaBrain-SJTU/Rare-ECG.

Figures & tables

Figure 1: The proposed two-stage ECG diagnosis framework. Step 1: Self-supervised pretraining for ECG anomaly detection, training a model to detect abnormal patterns using global and local ECG features. Step 2: Fine-tuning the classifier based on the pretrained anomaly detection model for detailed diagnosis. This method enhances classification performance, particularly for less frequent cardiac conditions, by leveraging anomaly detection pretraining.

Figure 2: The details of the multi-scale cross-restoration framework for ECG anomaly detection pretraining.

Figure 3: The long-tailed (extremely imbalanced) distribution of cardiac disease types across the ECG-LT dataset, with cardiac disease types divided into common set, uncommon set, and rare set based on their frequency of occurrence. The red box highlights the expanded view of part of the Rare Set, where cardiac disease types occur fewer than 80 times.

Figure 4: Comprehensive analysis of the novel ECG-LT dataset. a. Hierarchical architecture of cardiac types. b. Comparison of the number of cardiac types in the ECG-LT dataset to those in existing ECG databases. c. Age distribution across the training, internal validation, and external validation sets. d. Gender distribution across the training, internal validation, and external validation sets.

					Pretraining + Finetuning
Dataset	Evaluation Data	ECG Sample Count	Metrics	Train from Scratch	Contrastive Learning	MoCo	AD:TSL	AD:Ours Finetuning	AD:Ours Joint-train
Internal Validation (ECG-LT)	All Test Data	189,440	AUROC	0.910 ±0.053	0.923 ±0.050	0.906 ±0.054	0.925 ±0.049	0.948 ±0.041	0.964 ±0.035
			Sensitivity	0.879 ${}_{\pm 0.061}^{*}$	0.882 ${}_{\pm 0.060}^{*}$	0.844 ${}_{\pm 0.068}^{*}$	0.878 ${}_{\pm 0.061}^{*}$	0.925 ${}_{\pm 0.049}^{*}$	0.931 ±0.047
			Specificity	0.830 ${}_{\pm 0.070}^{*}$	0.888 ${}_{\pm 0.059}^{*}$	0.876 ${}_{\pm 0.062}^{*}$	0.879 ${}_{\pm 0.061}^{*}$	0.889 ${}_{\pm 0.059}^{*}$	0.925 ±0.049
	Common Test Data	183,810	AUROC	0.940 ±0.077	0.937 ±0.078	0.964 ±0.060	0.932 ±0.081	0.962 ±0.062	0.969 ±0.056
			Sensitivity	0.887 ${}_{\pm 0.102}^{*}$	0.886 ${}_{\pm 0.102}^{*}$	0.920 ${}_{\pm 0.087}^{*}$	0.890 ${}_{\pm 0.100}^{*}$	0.919 ${}_{\pm 0.088}^{*}$	0.935 ±0.079

Table 1: Performance of ECG diagnosis across multiple evaluation datasets, including internal validation on ECG-LT and external validation on PTB-XL and Renji. The best-performing method is shown in bold , and the second-best is underlined .

Figure 5: Performance of ECG diagnosis. (a) Diagnosis performance on tail classes. Comparison of ECG diagnosis for each type in the Rare Set, where the proposed method outperforms, particularly for anomalies with fewer samples. (b) Diagnosis fairness across sex. The model shows consistent performance between male and female subjects, ensuring balanced accuracy. (c) Diagnosis fairness across age. Performance is evaluated across age groups in ten-year intervals, with the model demonstrating stable and equitable results. (d) Visualization of anomaly localization. Comparison of the proposed method (Ours) with a leading baseline (BeatGAN), against ground truth (pink boxes). Color-coded scores (0–1) indicate anomaly likelihood, with red highlighting the most probable locations.

Task	Evaluation	Metrics	State-of-the-art Methods				Ours
Task	Data	Metrics	TranAD	AnoTran	BeatGAN	TSL	Ours
Anomaly Detection	All Test Data	AUROC	0.566 ±0.002	0.601 ±0.002	0.653 ±0.002	0.824 ±0.002	0.912 ±0.001
		F1 Score	0.667 ${}_{\pm 0.002}^{**}$	0.673 ${}_{\pm 0.002}^{**}$	0.669 ${}_{\pm 0.002}^{**}$	0.746 ${}_{\pm 0.002}^{**}$	0.837 ±0.002
		Sensitivity	0.511 ${}_{\pm 0.002}^{**}$	0.579 ${}_{\pm 0.002}^{**}$	0.522 ${}_{\pm 0.002}^{**}$	0.793 ${}_{\pm 0.002}^{**}$	0.842 ±0.002
		Specificity	0.514 ±0.002	0.564 ${}_{\pm 0.002}^{**}$	0.698 ${}_{\pm 0.002}^{**}$	0.668 ${}_{\pm 0.002}^{**}$	0.830 ±0.002
		Pre@90	0.501 ${}_{\pm 0.002}^{**}$	0.530 ${}_{\pm 0.002}^{**}$	0.528 ${}_{\pm 0.002}^{**}$	0.613 ${}_{\pm 0.002}^{**}$	0.756 ±0.002

Table 2: Anomaly detection and localization results across multiple evaluation data, compared to state-of-the-art methods. The most effective method is highlighted in bold , and the second-best is underlined .

Figure 6: Evaluation of the attribute prediction module. a. Comparison of predicted ages across different age groups for normal and abnormal ECGs. b. Age classification accuracy for normal ECGs depicted across age ranges, showing predicted versus actual age. c. Heart rate prediction deviation compared to a standard reference range (y-axis) across varying anomaly rarity (x-axis). d-g. Analysis of deviations in PR interval, QT interval, corrected QT interval, and QRS complex predictions.

Figure 7: Validation results from Ruijin Hospital’s simulated clinical environment, comparing diagnostic timing, completeness, and accuracy between cardiologists and AI-assisted diagnosis. The figure shows diagnosis time and accuracy rates for: (i) cardiologists under time constraints, (ii) cardiologists without time constraints, (iii) AI alone, and (iv) cardiologists with AI assistance. Metrics include correct identification of comprehensive diagnoses (key diseases and detailed signal characteristics), correct identification of key diseases, and frequency of missed key diseases. AI-assisted diagnosis outperformed others in speed and accuracy, notably reducing diagnostic time under time constraints.

Evaluation Data	Metrics	Cardiologist (time constraints)	Cardiologist (no time constraints)	AI	Cardiologist (AI assisted)
Common data	Sensitivity	0.626 ±0.165	0.747 ±0.148	0.837 ±0.126	0.820 ±0.131
Common data	Specificity	0.992 ±0.030	0.973 ±0.055	0.916 ±0.095	0.989 ±0.036
Long-tail data	Sensitivity	0.347 ±0.226	0.469 ±0.237	0.592 ±0.234	0.714 ±0.215
Long-tail data	Specificity	0.997 ±0.026	0.997 ±0.026	0.986 ±0.056	0.997 ±0.026

Table 3: Validation results from Ruijin Hospital’s simulated clinical environment: Sensitivity and specificity analysis for common and long-tail anomalies.

Data	Condition	Cardiac Conditions	# All data	# Test data
Type	Types	Cardiac Conditions	# All data	# Test data
Common	37	Normal electrocardiogram	539,607	94,304
		T wave changes	182,756	33,907
		Sinus bradycardia	99,275	17,463
		ST-T segment changes	80,334	11,953
		Sinus tachycardia	61,287	11,272

Table S1: Detailed information about common data on our general hospital ECG-LT dataset.

Data	Condition	Cardiac Conditions	# All data	# Test data
Type	Types	Cardiac Conditions	# All data	# Test data
Uncommon	36	Prolonged QT interval	2,486	410
		Right axis deviation	2,476	450
		Old anterior myocardial infarction	2,403	418
		Old inferior myocardial infarction	2,344	362
		Left axis deviation	1,856	188

Table S2: Detailed information about uncommon data on our general hospital ECG-LT dataset.

Data	Condition	Cardiac Conditions	# All data	# Test data
Type	Types	Cardiac Conditions	# All data	# Test data
Rare	43	High lateral myocardial infarction	394	30
		Upsloping ST segment depression	368	30
		Downsloping ST segment depression	311	33
		Second-degree sinoatrial block	257	50
		High-degree atrioventricular block	247	20

Table S3: Detailed information about rare data on our general hospital ECG-LT dataset.

Dataset	Condition	Cardiac Conditions	# Test data
Dataset	Types	Cardiac Conditions	# Test data
Renji	26	Paced electrocardiogram	398
		Myocardial infarction	223
		Complete left bundle branch block	214
		Complete right bundle branch block	209
		Atrial fibrillation	144

Table S4: Detailed information about Renji dataset.

Data	Condition	Cardiac Conditions	#Data	Rarity
Source	Types	Cardiac Conditions	#Data	Rarity
	50	Normal electrocardiogram	40	Common
		Sinus tachycardia	35	Common
		Atrial fibrillation	26	Common
		ST-T segment changes	26	Common
		T wave changes	24	Common

Table S5: Detailed information about data in the simulated clinical validation.

Evaluation	Metrics	None	+MR	+MR+MC	+MR+MC	+MR+MC
Data	Metrics	None	+MR	+MR+MC	+TAR	+TAR+APM
All Test Data	AUROC	0.779	0.882	0.905	0.909	0.912
	F1 Score	0.720	0.802	0.828	0.833	0.837
	Sensitivity	0.823	0.810	0.825	0.841	0.842
	Specificity	0.537	0.790	0.833	0.823	0.830
	Pre@90	0.585	0.684	0.737	0.749	0.756

Table S6: Ablation studies evaluating the impact of masking and restoring (MR), multi-scale cross-attention (MC), trend assisted restoration (TAR), and the attribute prediction module (APM) for anomaly detection on various test sets. Results are shown in the patient-level AUC. The best-performing method is in bold .

Attribute	Reference Range	Normal	Abnormal
Attribute	Reference Range	Normal	All	Common	Uncommon	Rare
Gender	0 (male) or 1 (female)	86.5% Acc	75.9% Acc	76.2% Acc	69.2% Acc	70.9% Acc
Age	0 $\sim$ 100 years	$\pm$ 12.6	$\pm$ 14.0	$\pm$ 13.8	$\pm$ 16.8	$\pm$ 18.1
Heart Rate (HR)	60 $\sim$ 100 bpm	$\pm$ 3.64	$\pm$ 6.94	$\pm$ 6.86	$\pm$ 8.20	$\pm$ 8.62
PR Interval (PR)	120 $\sim$ 200 ms	$\pm$ 18.5	$\pm$ 26.7	$\pm$ 26.2	$\pm$ 35.2	$\pm$ 37.8
QT Interval (QT)	320 $\sim$ 440 ms	$\pm$ 20.7	$\pm$ 33.0	$\pm$ 32.4	$\pm$ 43.2	$\pm$ 44.0

Table S7: Attribute prediction results for various attributes and data types on general hospital ECG dataset. Results are shown in accuracy for binary gender and averaged deviation for other attributes.

Explore similar work

May 25, 2026

A Signal-Language Foundation Model for Broad-Spectrum Cardiovascular Assessment from Routine Electrocardiography

Electrocardiography (ECG) is central to cardiovascular care, but conventional AI models are often restricted to common arrhythmias and may generalize poorly across populations or clinically subtle diseases. We developed ECG Contrastive Language-Image Pre-training (ECGCLIP), a signal-language contrastive learning framework that aligns ECG waveforms with expert diagnostic reports. ECGCLIP was pre-trained on 2,837,962 ECG studies from 1,324,856 patients and evaluated on a held-out internal test set plus nine independent external cohorts comprising about 1.5 million ECGs. Evaluation covered 89 downstream tasks, including 45 ECG diagnoses, 39 echocardiographic targets, and 5 rare cardiac diseases, using PRAUC as the primary metric. ECGCLIP consistently improved performance over random initialization and Merl-R18 baselines. On the internal test set, ECGCLIP-R34 achieved strong performance for atrial fibrillation (PRAUC 0.900) and ST-segment elevation myocardial infarction (PRAUC 0.383), with robust generalization across all external cohorts. It also improved low-prevalence and diagnostically elusive diseases, including Ebstein anomaly, constrictive pericarditis, dextrocardia, and cardiac amyloidosis, with internal PRAUC values of 0.253, 0.175, 0.121, and 0.201, respectively. ECGCLIP was data efficient, matching or exceeding full-dataset baseline performance with only 10% of training data. Feature visualization and saliency analysis suggested clinically meaningful representations aligned with established electrocardiographic criteria. These findings indicate that large-scale ECG-report contrastive pre-training can expand routine ECG interpretation beyond common arrhythmias toward broad cardiovascular assessment and opportunistic screening of echocardiographic and rare conditions.

Ziqing Yu, Yuhui Tao, Jiayu Huo +19

Department of Cardiology, Zhongshan Hospital of Fudan University, Shanghai Institute of Cardiovascular Diseases, National Clinical Research Centre for Interventional Medicine, Shanghai, China · Digital Medical Research Center, School of Basic Medical Sciences, Fudan University, Shanghai, 200032, China. · Shanghai Key Laboratory of Medical Imaging Computing and Computer Assisted Intervention, Shanghai, 200032, China. +8

Electrocardiography Electrocardiogram Foundation Models Electrocardiogram Classification Multimodal Contrastive Learning Time-Series Foundation Models

Ask PDF

May 23, 2026

HeartBeatAI: An Interpretable and Robust Deep Learning Framework for Multi-Label ECG Arrhythmia Detection

While Deep Learning (DL) enhances automated electrocardiogram (ECG) analysis, clinical deployment is hindered by class imbalance and the generalization gap. This paper presents HeartBeatAI, a deep learning framework combining domain generalization, multi-scale feature aggregation, and clinical explainability for robust 12-lead ECG classification. Moving beyond image-based paradigms, HeartBeatAI integrates a Squeeze-and-Excitation (SE) ResNet to isolate diagnostic leads alongside a Multi-Layer Concentration Pipeline to capture macro-rhythm and micro-morphological anomalies. To mitigate domain shift, the framework employs MixStyle regularization and Label Smoothing. Rigorous benchmarking across four large-scale datasets using intra-source and Leave-One-Domain-Out (LODO) protocols demonstrates high performance (98% Macro F1-score) under intra-source conditions. However, LODO evaluations reveal significant degradation in detecting rare anomalies, highlighting a persistent challenge in cross-institutional deployment.

Shubham Gupta, Nikhil Panwar, Partha Pratim Roy

Department of Computer Science and Engineering, Indian Institute of Technology (ISM) Dhanbad, Dhanbad, Jharkhand, India. · Department of Computer Science and Engineering, Indian Institute of Technology Roorkee, Roorkee, Uttarakhand, India.

Electrocardiography Domain Generalization Electrocardiogram Classification

Ask PDF

Jul 16, 2026

Angular Gaussian Supervised Contrastive Learning for Long-Tailed Electrocardiogram Arrhythmia Diagnosis

Long-tailed label distributions reduce the reliability of deep learning for electrocardiogram (ECG) arrhythmia diagnosis, particularly for clinically important but rare abnormalities. Existing rebalancing and logit adjustment methods mainly address class frequency while overlooking direction-dependent morphological variability across ECG classes. This study proposes Angular Gaussian Supervised Contrastive Learning (AG-SCL) for long-tailed multi-label ECG diagnosis. AG-SCL integrates three components into a unified framework: an Angular Gaussian contrastive branch that models full-covariance class uncertainty on unit-normalized embeddings, Adaptive Logit Adjustment that learns bounded label-state-specific prior corrections instead of fixed frequency-based margins, and tail-aware augmentation that generates morphology-preserving views while protecting the 7-25 Hz QRS-dominant band. The method was evaluated on the public PTB-XL benchmark and a nocturnal ECG dataset comprising 1317 hours of recordings from 141 subjects. AG-SCL achieved the best macro-level performance on both datasets. On PTB-XL, it obtained a balanced accuracy of 0.838, sensitivity of 0.709, specificity of 0.968, mean average precision of 0.495, and TPR at 5% FPR of 0.778. On Noc-ECG, the corresponding values were 0.918, 0.889, 0.947, 0.488, and 0.900. The largest gains occurred in rare or morphologically unstable rhythm classes, while ablation studies confirmed the contributions of full-covariance modelling, Adaptive Logit Adjustment, and tail-aware augmentation. AG-SCL improves long-tailed ECG diagnosis by combining prior calibration with anisotropic representation learning, enhancing sensitivity to rare arrhythmias while maintaining clinically relevant specificity. Our code is available at: https://github.com/Open-EXG/AG-SCL-for-Long-Tailed-ECG.

Jin Dai, Qiuzhen Zhang, Chenyun Dai +2

School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, 200030, China · Department of Neurology and Neurological Rehabilitation, Shanghai Yangzhi Rehabilitation Hospital, School of Medicine, Tongji University, Shanghai, 201619, China

Class-Imbalanced Learning Multi-Label Classification Electrocardiogram Classification Supervised Contrastive Learning

Ask PDF