Cross-Modal Generative Framework for Signal Translation from Fetal-Maternal Electrocardiograms to Fetal Doppler Waveforms

2026-07-09 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Medical Devices & Health Technology · Depth: Expert, quick

Summary

A novel cross-modal generative framework has been developed to translate fetal-maternal electrocardiograms (fECG) into fetal Doppler waveforms, offering insights into fetal cardiovascular function. This framework integrates dilated convolutions with cross-modal attention, specifically incorporating maternal ECG data, and self-attention to capture long-range temporal dependencies. Trained on 885 synchronized fECG/mECG and Doppler segments from 39 pregnancies, the model synthesizes Doppler envelopes with a power spectral density mean squared error (PSD MSE) of 49.9 +/- 15.8 dB^2, representing a 51% reduction compared to a two-channel baseline. It also achieves a heart-rate error of 4.71 +/- 0.77 bpm, which is 1.5% better than baseline and negligible within the 110-160 bpm physiological range. The framework's cross-modal attention component alone contributes a 39% PSD MSE reduction, quantifying the impact of maternal-fetal coupling. This approach advances computational modeling by enabling Doppler synthesis from dual-lead ECG, facilitating a more comprehensive fetal assessment.

Key takeaway

For research scientists developing advanced fetal monitoring systems, this generative framework offers a robust method to synthesize fetal Doppler waveforms from ECG data. You can utilize its cross-modal attention to quantify maternal-fetal coupling, providing deeper insights into mechanical contributions to fetal circulation. Consider integrating this approach to enhance the comprehensiveness of your fetal assessment tools and improve diagnostic capabilities.

Key insights

The framework translates fECG/mECG to Doppler, quantifying electrical vs. mechanical contributions to fetal circulation.

Principles

• Fetal-maternal cardiac coupling informs fetal hemodynamics.
• Cross-modal attention quantifies coupling contributions.
• Complementary signals offer comprehensive fetal assessment.

Method

A generative framework combines dilated convolutions, cross-modal attention for maternal ECG, and self-attention for temporal dependencies.

In practice

• Synthesize Doppler envelopes from dual-lead ECG.
• Quantify mechanical contributions to Doppler waveforms.
• Inform clinical decisions on fetal circulation.

Topics

• Fetal Monitoring
• Electrocardiogram
• Doppler Ultrasound
• Generative Models
• Cross-Modal Attention
• Cardiovascular Modeling

Best for: AI Scientist, Research Scientist

Related on AIssential

• Knowledge-Guided Cross-Modal Fusion for Adult-to-Pediatric ECG Transfer via Label-Conditioned Contrastive Alignment — Label-conditioned knowledge alignment is key for effective adult-to-pediatric ECG model transfer, especially with scarce pediatric data.
• Structure-Augmented Standard Plane Detection with Temporal Aggregation in Blind-Sweep Fetal Ultrasound — Structure-augmented temporal aggregation improves fetal abdomen plane detection in blind-sweep ultrasound for reliable biometry.
• FHRFormer: A Self-Supervised Masked Transformer Framework for Fetal Heart Rate Time-Series Inpainting and Forecasting — A masked transformer autoencoder can robustly reconstruct missing fetal heart rate data for improved AI analysis.
• MVSegNet: A Lightweight Boundary-Aware Network for Fetal Lateral Ventricle Segmentation and Atrial Width Estimation in Prenatal Ultrasound — MVSegNet accurately segments fetal lateral ventricles for atrial width estimation despite ultrasound challenges.
• WING: A Window-Prior-Based Generative Network with Gated Inception for Cross-Modality CT Synthesis — WING synthesizes CT from MRI/CBCT using windowed representations and a multi-component generative network for improved radiotherapy planning.
• Transformer-Guided Graph Attention for Direct Cardiac Mesh Reconstruction: A Structural Digital Twin Framework — The framework directly generates simulation-ready cardiac meshes from medical images, bypassing traditional multi-stage, manual post-processing.

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.