Knowledge-Guided Cross-Modal Fusion for Adult-to-Pediatric ECG Transfer via Label-Conditioned Contrastive Alignment

2026-07-17 · Source: Machine Learning · Field: Science & Research — Artificial Intelligence & Machine Learning, Health & Medical Research · Depth: Expert, quick

Summary

The Pediatric-Adult ECG Alignment via Cross-modal Enhancement (PEACE) framework improves ECG interpretation model transfer from adult to pediatric populations. This is crucial when pediatric labels are scarce, as adult-pretrained models often perform poorly. Pretrained on the adult MIMIC-IV ECG corpus, PEACE uses a knowledge-guided approach. It describes diagnoses via rhythm, morphology, and ST-T axes, composing positive-label descriptors into axis tokens and a fused embedding. Key components include a Label Query Network (LQN) for cross-attention and Label Set Aware Bidirectional Contrastive Learning (LSBC) for feature alignment. Curriculum Adaptive Fusion (CAF) manages alignment strength during training. Inference uses only ECG signals. On ZZU-pECG, PEACE achieved macro average AUCs of 59.39% (zero-shot), 81.74% (50-shot), and 91.56% (full fine-tuning). It showed clear gains under limited supervision. Fine-tuning on PTB-XL yielded a 96.90% macro average AUC over nine harmonized labels.

Key takeaway

For Machine Learning Engineers developing medical AI for ECG interpretation, prioritize label-conditioned knowledge alignment. This is crucial when adapting models from adult to pediatric populations, especially with limited pediatric data. This approach significantly improves performance, as PEACE achieved a 59.39% zero-shot AUC on ZZU-pECG. Integrate diagnosis-specific axis tokens and curriculum adaptive fusion into your transfer learning strategies. This will enhance model robustness and accuracy across diverse age groups.

Key insights

Label-conditioned knowledge alignment is key for effective adult-to-pediatric ECG model transfer, especially with scarce pediatric data.

Principles

• Age-sensitive criteria hinder ECG model transfer.
• Aligning specific diagnostic labels improves transfer.
• Global text fusion is less effective than label-conditioned alignment.

Method

PEACE uses a Label Query Network (LQN) for cross-attention and Label Set Aware Bidirectional Contrastive Learning (LSBC) to align ECG features with fused, diagnosis-specific embeddings. Curriculum Adaptive Fusion (CAF) gates alignment strength.

In practice

• Pretrain on large adult ECG corpora like MIMIC-IV.
• Focus on label-conditioned alignment for transfer learning.
• Use axis tokens for rhythm, morphology, ST-T descriptions.

Topics

• ECG Interpretation
• Pediatric Cardiology
• Transfer Learning
• Contrastive Learning
• Cross-Modal Fusion
• MIMIC-IV

Best for: AI Scientist, Research Scientist, Machine Learning Engineer

Related on AIssential

• CAP: Towards PPG Universal Representation Learning with Patient-level Supervision — CAP improves PPG representation learning by anchoring signal data to patient-level clinical semantics via cross-modal contrastive pretraining.
• Echo2ECG: Enhancing ECG Representations with Cardiac Morphology from Multi-View Echos — Aligning global ECGs with multi-view Echos resolves representational mismatch, yielding powerful, lightweight cardiac morphology features.
• Language Models as Semantic Teachers: Post-Training Alignment for Medical Audio Understanding — Aligning audio encoders with medical LLMs via a "semantic teacher" paradigm significantly enhances clinical diagnostic performance.
• LSTrans: Efficient Knowledge Transfer for Lightweight and Automated ECG Classification — LSTrans efficiently classifies ECGs on wearables by combining a hybrid neural architecture with knowledge distillation and low-rank adaptation.
• Cross-Modal Generative Framework for Signal Translation from Fetal-Maternal Electrocardiograms to Fetal Doppler Waveforms — The framework translates fECG/mECG to Doppler, quantifying electrical vs. mechanical contributions to fetal circulation.
• UniECG: Understanding and Generating ECG in One Unified Model — UniECG unifies evidence-based ECG interpretation and text-conditioned signal generation in a single model.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.