Learning from VAE Errors to support ECG-based Differential Diagnosis of Myocardial Scar

Shayan Sharifi; Riccardo Treu; Ilaria Gandin; Federico Garoia; Marco Merlo; Giulia Cisotto

cs.LGSep 4, 2026

Learning from VAE Errors to support ECG-based Differential Diagnosis of Myocardial Scar

Authors: Shayan Sharifi, Riccardo Treu, Ilaria Gandin, Federico Garoia, Marco Merlo, Giulia Cisotto

Organizations: Department of Mathematics, Informatics, and Geosciences, University of Trieste, Italy · Centre of Diagnosis and Management of Cardiomyopathies; Azienda Sanitaria Universitaria Giuliano Isontina; University of Trieste, Italy; Member of Ern Guard-Heart · Department of Medical, Surgical and Health Sciences, University of Trieste, Italy

Abstract

Late Gadolinium Enhancement (LGE) on cardiac magnetic resonance is a key marker of myocardial scar, but its limited accessibility motivates routine ECG-based screening. We evaluated whether $β$ -variational autoencoder (VAE)-derived ECG representations can discriminate LGE+ from LGE- cardiomyopathic patients in a local cohort of 300 subjects. We compared 32-dimensional features from the foundation ECGx.AI model with those from a shallower $β$ -VAE trained on normal PTB-XL ECGs, evaluating downstream classification and Dynamic Time Warping (DTW)-based reconstruction errors. ECGx.AI reached an area under ROC of 0.686 with Random Forest, while the proposed $β$ -VAE reached 0.577 with sensitivity of 0.775 with Gradient Boosting. Notably, DTW-reconstruction errors significantly differed between classes in 10 out of 12 leads according to Mann-Whitney U test and help in classification, leading to an area under ROC of 0.643 with Logistic Regression, supporting their potential as markers of scar-related ECG alterations.

Figures & tables

Panel of Figure 1 — Figure 1: Comparison of original (black) and reconstructed (red) ECG signals, showing a median beat extracted from Lead II. (a) A normal ECG sample from the PTB-XL validation set (MSE = 0.0007, DTW = 0.2872, $r$ = 0.9846). (b) A LGE- sample (MSE = 0.5280, DTW = 1.4314, $r$ = 0.8734) and (c) a LGE+ sample (MSE = 0.6374, DTW = 11.6183, $r$ = 0.6194).

Figure 2: Study processing pipeline overview

Model (CV)	Normalization	Sensitivity	Specificity	Accuracy	AUROC
Random Forest (5-Fold CV)	Factor-wise	0.852 $\pm$ 0.049	0.389 $\pm$ 0.074	0.657 $\pm$ 0.036	0.686 $\pm$ 0.053
Random Forest (80/20 Split)	Factor-wise	0.858 $\pm$ 0.048	0.394 $\pm$ 0.086	0.664 $\pm$ 0.042	0.676 $\pm$ 0.047
Random Forest (LOSO)	None	0.873 $\pm$ 0.024	0.380 $\pm$ 0.044	0.666 $\pm$ 0.028	0.672 $\pm$ 0.032
Gradient Boosting (LOSO)	Patient-wise	0.868 $\pm$ 0.026	0.383 $\pm$ 0.046	0.664 $\pm$ 0.029	0.657 $\pm$ 0.033
Logistic Regression (LOSO)	Patient-wise	0.880 $\pm$ 0.025	0.340 $\pm$ 0.042	0.653 $\pm$ 0.028	0.644 $\pm$ 0.032

Table 1:

\mathrm{ECGx.AI}

: best classification models (ranked by AUROC).

Model (CV, input)	Normalization	Sensitivity	Specificity	Accuracy	AUROC
Logistic Regression (80/20, DTW errors)	Lead-wise	0.885 $\pm$ 0.078	0.191 $\pm$ 0.115	0.596 $\pm$ 0.030	0.643 $\pm$ 0.068
Extra Trees (80/20, DTW errors)	Lead-wise/None	0.815 $\pm$ 0.067	0.383 $\pm$ 0.089	0.635 $\pm$ 0.052	0.638 $\pm$ 0.066
Logistic Regression (5-Fold CV, DTW errors)	Lead-wise	0.895 $\pm$ 0.078	0.188 $\pm$ 0.108	0.597 $\pm$ 0.038	0.637 $\pm$ 0.081
Extra Trees (5-Fold CV, DTW errors)	Lead-wise/None	0.801 $\pm$ 0.067	0.373 $\pm$ 0.100	0.621 $\pm$ 0.057	0.626 $\pm$ 0.088
Logistic Regression (LOSO, DTW errors)	None	0.753 $\pm$ 0.032	0.431 $\pm$ 0.042	0.618 $\pm$ 0.027	0.624 $\pm$ 0.032
Gradient Boosting (LOSO, latents)	Patient-wise	0.775 $\pm$ 0.033	0.308 $\pm$ 0.041	0.579 $\pm$ 0.029	0.577 $\pm$ 0.034

Table 2: Shallow

\beta

- VAE : best classification models (ranked by AUROC).

Figure 3: Lead-wise DTW -based reconstruction errors in LGE- (blue bars) and LGE+ (red bars) classes, after averaging over patients. Statistical significance between classes was evaluated via two-tailed Mann-Whitney U test (* $p<0.05$ , ** $p<0.01$ , *** $p<0.001$ , n.s.: not significant). Error bars indicate standard error of the mean.

Explore similar work

Progressive Learning with Anatomical Priors for Reliable Left Atrial Scar Segmentation from Late Gadolinium Enhancement MRI
Mar 27, 2026Jing Zhang, Bastien Bergere, Emilie Bollache +4Medical Image Segmentation Cardiac MRI
Laboratoire d’Imagerie Biomédicale (LIB), Sorbonne Université, INSERM, CNRS, Paris, France · APHP, Institut de Cardiologie, Hôpital Pitié-Salpêtrière, Paris, France · APHP, Imagerie Cardiovasculaire et thoracique, Institut de Cardiologie, Hôpital Pitié Salpêtrière, Paris, France
Curriculum-Guided Myocardial Scar Segmentation for Ischemic and Non-ischemic Cardiomyopathy
Mar 30, 2026Nivetha Jayakumar, Jonathan Pan, Shuo Wang +6Medical Image Segmentation Cardiac MRI
Department of Electrical and Computer Engineering, University of Virginia, USA · Division of Cardiovascular Medicine, University of Virginia Health System, USA · Department of Computer Science, University of Virginia, USA
MSAIC-Net: A Multi-Scale Attention and Imbalance-Aware Contrastive Network for ECG-Based Myocardial Substrate Abnormality Detection
Jun 4, 2026Canyu Lei, Fenglin Zhang, Derek Bivona +5Class-Imbalanced Learning Electrocardiogram Classification

Mar 27, 2026J/K move · Enter open · S save

Progressive Learning with Anatomical Priors for Reliable Left Atrial Scar Segmentation from Late Gadolinium Enhancement MRI

Jing Zhang, Bastien Bergere, Emilie Bollache, Jonas Leite+3

Laboratoire d’Imagerie Biomédicale (LIB), Sorbonne Université, INSERM, CNRS, Paris, France · APHP, Institut de Cardiologie, Hôpital Pitié-Salpêtrière, Paris, France · APHP, Imagerie Cardiovasculaire et thoracique, Institut de Cardiologie, Hôpital Pitié Salpêtrière, Paris, France

Medical Image Segmentation Cardiac MRI

Open paper AskPDF arXiv

Cardiac MRI late gadolinium enhancement (LGE) enables non-invasive identification of left atrial (LA) scar, whose spatial distribution is strongly associated with atrial fibrillation (AF) severity and recurrence. However, automatic LA scar segmentation remains challenging due to low contrast, annotation variability, and the lack of anatomical constraints, often leading to non-reliable predictions. Accordingly, our aim was to propose a progressive learning strategy to segment LA scar from LGE images inspired from a clinical workflow. A 3-stage framework based on SwinUNETR was implemented, comprising: 1) a first LA cavity pre-learning model, 2) dual-task model which further learns spatial relationship between LA geometry and scar patterns, and 3) fine-tuning on precise segmentation of the scar. Furthermore, we introduced an anatomy-aware spatially weighted loss that incorporates prior clinical knowledge by constraining scar predictions to anatomically plausible LA wall regions while mitigating annotation bias. Our preliminary results obtained on validation LGE volumes from LASCARQS public dataset after 5-fold cross validation, LA segmentation had Dice score of 0.94, LA scar segmentation achieved Dice score of 0.50, Hausdorff Distance of 11.84 mm, Average Surface Distance of 1.80 mm, outperforming only a one-stage scar segmentation with 0.49, 13.02 mm, 1.96 mm, repectively. By explicitly embedding clinical anatomical priors and diagnostic reasoning into deep learning, the proposed approach improved the accuracy and reliability of LA scar segmentation from LGE, revealing the importance of clinically informed model design.

Abstract

Explore similar work

Progressive Learning with Anatomical Priors for Reliable Left Atrial Scar Segmentation from Late Gadolinium Enhancement MRI

Curriculum-Guided Myocardial Scar Segmentation for Ischemic and Non-ischemic Cardiomyopathy

MSAIC-Net: A Multi-Scale Attention and Imbalance-Aware Contrastive Network for ECG-Based Myocardial Substrate Abnormality Detection