Learnijoy
ResearchAI ResearchAI Engineering & DevTools

AI Translates Fetal-Maternal ECG to Fetal Doppler Waveforms

Tongli Su, Alireza Rafiei, Marly van Assen, Reza Sameni, Gari D. Clifford, Faezeh Marzbanrad, Nasim Katebi· July 10, 2026 View original

Key takeaways

  • AI framework synthesizes fetal Doppler waveforms from fetal and maternal ECGs.
  • It quantifies mechanical contributions to fetal circulation beyond electrical signals.
  • The model uses cross-modal and self-attention for high accuracy.
  • This advances non-invasive fetal cardiovascular assessment and clinical decision-making.

Who benefits

HealthcareMedical DevicesAI/ML DevelopmentPharmaceuticals

Summary

Researchers developed a cross-modal generative AI framework that synthesizes fetal Doppler waveforms from fetal and maternal electrocardiograms (ECGs). This model helps quantify the mechanical contributions to fetal circulation not captured by electrical signals, offering deeper insights into fetal cardiovascular health.

A novel cross-modal generative framework has been developed to translate electrical signals from fetal and maternal electrocardiograms (ECGs) into mechanical fetal Doppler waveforms. This innovation aims to provide a more comprehensive understanding of fetal cardiovascular function by distinguishing between electrical activity and mechanical hemodynamics, which are influenced by factors like placental resistance. The framework utilizes dilated convolutions combined with cross-modal attention to selectively incorporate maternal ECG data, alongside self-attention to capture long-range temporal dependencies. This design allows the model to computationally map the complex relationships between maternal-fetal cardiac coupling and fetal hemodynamics. Trained on 885 synchronized ECG and Doppler segments from 39 pregnancies, the model demonstrated high accuracy. It synthesized Doppler envelopes with significantly lower power spectral density mean squared error (49.9 +/- 15.8 dB^2, 51% lower than baseline) and improved heart-rate error (4.71 +/- 0.77 bpm). The 39% PSD MSE reduction attributed to cross-modal attention quantifies the valuable contribution of maternal-fetal coupling, advancing computational modeling for fetal assessment.

Why it matters

This technology offers a non-invasive way to gain deeper insights into fetal cardiovascular health, potentially enabling earlier detection of complications and more informed clinical decisions for obstetricians and neonatologists.

How to implement this in your domain

  1. 1Collaborate with medical researchers to validate the model's clinical utility in diverse patient populations.
  2. 2Integrate this signal translation capability into new or existing fetal monitoring devices.
  3. 3Develop diagnostic tools that leverage the synthesized Doppler waveforms for comprehensive fetal assessment.
  4. 4Explore the "residual Doppler components" to identify purely mechanical factors indicative of specific fetal conditions.

Original post by Tongli Su, Alireza Rafiei, Marly van Assen, Reza Sameni, Gari D. Clifford, Faezeh Marzbanrad, Nasim Katebi

"arXiv:2607.08073v1 Announce Type: new Abstract: Fetal electrocardiogram (fECG) and Doppler ultrasound provide complementary views of fetal cardiovascular function: fECG captures electrical activity while Doppler reflects mechanical hemodynamics shaped by factors such as placental…"

View on X

Originally posted by Tongli Su, Alireza Rafiei, Marly van Assen, Reza Sameni, Gari D. Clifford, Faezeh Marzbanrad, Nasim Katebi on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses

More in AI Research

AI ResearchAI Engineering & DevTools

Children Outperform AI in Language Acquisition, Mystery Remains

Human children still learn language with perfect fluency more efficiently than advanced AI models, a phenomenon scientists do not yet fully understand. This highlights a significant gap in current artificial intelligence capabilities compared to biological learning.

Elise CuttsAug 24, 2026
AI ResearchAI Engineering & DevTools

Harmony Improves Protein-Ligand Flexible Docking with Torsional Diffusion

Researchers introduce Harmony, a harmonic torsional diffusion framework for flexible protein-ligand docking that explicitly accounts for the periodic geometry of angular variables. This method improves ligand pose accuracy and pocket all-atom reconstruction on benchmarks like PDBBind and enhances the physical validity of generated complexes on PoseBusters.

Maksim Zhdanov, Pavel Strashnov, Vladislav KurenkovAug 24, 2026
AI Engineering & DevToolsAI Research

Multilingual Verifier Bias Impacts RLVR in LLM Mathematical Reasoning

A study reveals that exact-match verifiers in Reinforcement Learning with Verifiable Rewards (RLVR) for Large Language Models (LLMs) exhibit significant language-dependent false-negative reward noise in multilingual mathematical reasoning. This bias, particularly pronounced in Japanese, stems from format and script variations, highlighting a cross-lingual selection bottleneck that impedes effective multilingual LLM training.

Chenyu Zhou, Qiliang Jiang, Xu ZhouAug 24, 2026