mlsanityData to 14 Oct 2026

ImputeECG: Deep Learning Reconstruction of Complete 12-Lead Electrocardiograms from Incomplete Recordings for Cardiac Assessment

arXiv 2607.05009 ↗First posted 6 Jul 2026cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Complete digital 12-lead electrocardiograms (ECGs) are essential for AI-enabled cardiovascular assessment, yet many clinical ECG records, particularly those digitized from ECG images, remain incomplete because of short display formats, incomplete waveform digitization, lead loss, or signal corruption. We developed ImputeECG, a mask-conditioned one-dimensional Transformer autoencoder that completes 12-lead, 10-s ECGs while retaining all observed samples. The model was trained on PTB-XL and evaluated on PTB-XL and CPSC2018 under simulated incomplete settings, with additional real-world validation in a 43,633-record Kailuan clinical cohort after ECG image digitization. Metrics were computed over originally missing regions, with analyses of morphology and downstream diagnostic utility. On PTB-XL, ImputeECG reduced missing-region MAE by 41.7-51.0% and MSE by 54.0-63.7% versus the strongest baseline, with lower errors in R-peak timing, RR interval, QRS duration, QT interval, and P-wave, QRS-complex, and T-wave reconstruction. On CPSC2018, ImputeECG reduced MAE by 49.7-51.9%, supporting external generalization. In downstream multi-label classification, ImputeECG restored performance to 92.28% AUROC and 33.88% AUPRC in the most incomplete PTB-XL setting, approaching complete-ECG performance. On CPSC2018, completed ECGs achieved 94.75-95.89% AUROC and 78.83-81.86% AUPRC across settings. In Kailuan, ECG completion improved zero-shot sex prediction AUROC from 82.6% to 85.8% and reduced age prediction MAE from 10.72 to 9.87 years after image-based ECG digitization. These findings support ECG completion as a practical strategy for converting incomplete ECG records into AI-ready 12-lead, 10-s digital signals and extending the usable scope of ECG archives for digital cardiac assessment.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 15 of its 31 references that are in the map. First posted on arXiv 6 Jul 2026.

Builds on

15 of 31 references.

References · 15

  1. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
  2. 2018Attention U-Net: Learning Where to Look for the Pancreascited by 1,435
  3. 2021Swin-Unet: Unet-like Pure Transformer for Medical Image Segmentationcited by 1,222
  4. 2021Domain Adaptation for Medical Image Analysis: A Surveycited by 352
  5. 2019Deep Learning for Segmentation using an Open Large-Scale Dataset in 2D Echocardiographycited by 244
  6. 2020DoDNet: Learning to segment multi-organ and tumors from multiple partially labeled datasetscited by 94
  7. 2019Prior-aware Neural Network for Partially-Supervised Multi-Organ Segmentationcited by 88
  8. 2023MultiTalent: A Multi-Dataset Approach to Medical Image Segmentationcited by 45
  9. 2021Label-set Loss Functions for Partial Supervision: Application to Fetal Brain 3D MRI Parcellationcited by 12
  10. 2022Fix-A-Step: Semi-supervised Learning from Uncurated Unlabeled Datacited by 11
  11. 2020Towards Robust Partially Supervised Multi-Structure Medical Image Segmentation on Small-Scale Datacited by 7
  12. 2019Joint Learning of Brain Lesion and Anatomy Segmentation from Heterogeneous Datasetscited by 5
3 more
  1. 2021Learning from Partially Overlapping Labels: Image Segmentation under Annotation Shiftcited by 3
  2. 2022AI-enabled Assessment of Cardiac Systolic and Diastolic Function from Echocardiographycited by 3
  3. 2024Label Dropout: Improved Deep Learning Echocardiography Segmentation Using Multiple Datasets With Domain Shift and Partial Labellingcited by 1

Built on by

Not cited yet.