mlsanityData to 14 Oct 2026

TRACE: Tractable Routing Autoencoder for Clinical ECG

arXiv 2609.23460 ↗First posted 20 Sep 2026cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

Deep learning has advanced automated electrocardiogram (ECG) diagnosis, but the field's most accurate models, foundation models pretrained on millions of recordings, are not decision-pathway auditable: a clinician cannot trace a diagnosis to a physiological pathway or intervene on one. We propose TRACE, a Tractable Routing Autoencoder for Clinical ECG, whose 32-dimensional clinical latent space is specified in advance from domain knowledge rather than discovered by optimization. TRACE partitions this space into perfusion, structure, and conduction subspaces, routes each to its own diagnostic head by design, regularizes the partition with an orthogonality penalty, and reconstructs the ECG through a decoder that permits latent perturbation. On PTB-XL and Georgia, TRACE exceeds unconstrained classifiers and stays ahead of an ECG foundation model pretrained on ten million recordings, evaluated by linear probe on frozen features, at roughly an eighth of the parameter count. On the nine-label CPSC2018 cohort, which carries no structural class, the framework transfers with only the routing table re-specified to a perfusion/rhythm/conduction partition. Joint probe, erasure, and perturbation analyses verify the routing contract, and perturbing the depolarization and repolarization pathways modulates the reconstructed waveform. Removing the specified partition and its orthogonality penalty costs 1.70 AUC and 11.30 macro-F1 points on PTB-XL, and 2.76 AUC and 16.92 macro-F1 points on Georgia. A capacity-matched permutation control places arbitrary assignments within 0.34 AUC points of the ontology routing and leaves macro-F1 statistically level (p=0.619): the ontology supplies decision-pathway auditability at no macro-F1 cost.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 32 of its 73 references that are in the map. First posted on arXiv 20 Sep 2026.

Builds on

32 of 73 references.

References · 32

  1. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 12,049
  2. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,403
  3. 2017On Calibration of Modern Neural Networkscited by 4,960
  4. 2020Concept Bottleneck Modelscited by 1,032
  5. 2018Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representationscited by 941
  6. 2018Disentangling by Factorisingcited by 616
  7. 2019Variational Autoencoders and Nonlinear ICA: A Unifying Frameworkcited by 503
  8. 2019InceptionTime: Finding AlexNet for Time Series Classificationcited by 387
  9. 2016Time Series Classification from Scratch with Deep Neural Networks: A Strong Baselinecited by 360
  10. 2023Label-Free Concept Bottleneck Modelscited by 285
  11. 2020Logic Tensor Networkscited by 199
  12. 2020CLOCS: Contrastive Learning of Cardiac Signals Across Space, Time, and Patientscited by 171
20 more
  1. 2019Automatic diagnosis of the 12-lead ECG using a deep neural networkcited by 164
  2. 2020Deep Learning for ECG Analysis: Benchmarks and Insights from PTB-XLcited by 158
  3. 2022Weakly supervised causal representation learningcited by 140
  4. 2021Do Concept Bottleneck Models Learn as Intended?cited by 116
  5. 2023Large-scale Training of Foundation Models for Wearable Biosignalscited by 95
  6. 2020On Disentangled Representations Learned From Correlated Datacited by 94
  7. 2024ECG-FM: An Open Electrocardiogram Foundation Modelcited by 90
  8. 2024Guiding Masked Representation Learning to Capture Spatio-Temporal Relationship of Electrocardiogramcited by 84
  9. 2021Self-supervised representation learning from 12-lead ECG datacited by 80
  10. 2024Causal machine learning for predicting treatment outcomescited by 74
  11. 2020Conformal prediction for time seriescited by 53
  12. 2021Comprehensive Comparative Study of Multi-Label Classification Methodscited by 38
  13. 2021How I failed machine learning in medical imaging -- shortcomings and recommendationscited by 37
  14. 2025Reading Your Heart: Learning ECG Words and Sentences via Pre-training ECG Language Modelcited by 36
  15. 2024Energy-Based Concept Bottleneck Models: Unifying Prediction, Concept Intervention, and Probabilistic Interpretationscited by 33
  16. 2024Boosting Masked ECG-Text Auto-Encoders as Discriminative Learnerscited by 17
  17. 2024Knowledge-guided Machine Learning: Current Trends and Future Prospectscited by 14
  18. 2024Multilevel functional distributional models with application to continuous glucose monitoring in diabetes clinical trialscited by 7
  19. 2025ProtoECGNet: Case-Based Interpretable Deep Learning for Multi-Label ECG Classification with Contrastive Learningcited by 4
  20. 2026Physics-Informed Deep Learning for False Ventricular Tachycardia Alarm Reduction in the ICUcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026TRACE: Expert-Aligned ECG Representation Learning with Rigorous Benchmarking and Real-World Validation in Acute Cardiac Care
  2. 2026ECG-InterpBench: Benchmarking the Interpretability of ECG Foundation Models with Matched-Scale Sparse Autoencoders
  3. 2026CADENCE: A Cardiac Atom Dictionary for Interpretable Neural Concept Extraction from ECG Foundation Models
  4. 2025FlowECG: Using Flow Matching to Create a More Efficient ECG Signal Generator
  5. 2026Pretraining Strategies and Scaling for ECG Foundation Models: A Systematic Study
  6. 2023Towards quantitative precision for ECG analysis: Leveraging state space models, self-supervision and patient metadata
  7. 2026LSTrans: Efficient Knowledge Transfer for Lightweight and Automated ECG Classification
  8. 2026From Foundation ECG Models to NISQ Learners: Distilling ECGFounder into a VQC Student