CoChat

Placenta-pulmonary coupling-guided multimodal AI for fetal lung maturity staging and individualized glucocorticoid therapy

Open access

The DOI matches its Crossref record; 21 of the 24 references it cites were found in a registry or an index; 13 of the 28 sources its claims cite were checked for support.

Abstract

Background and aim: Deep learning has improved medical image analysis but often produces opaque decisions and correlation-driven predictions that may diverge from clinical reasoning. We hypothesize that a physiology-informed hybrid framework, which explicitly models placenta-pulmonary interactions and integrates multimodal data, could provide interpretable and reliable guidance for assessing fetal lung maturity (FLM) and optimizing antenatal glucocorticoids (GCs). Materials and methods: In a prospective cohort study involving 320 pregnancies-including 160 with hypertensive disorders of pregnancy (HDP)-each with weekly acquisitions from 28 to 36 weeks, we combined 2D/3D ultrasound, shear-wave elastography, Doppler, and maternal plasma metabolomics. A biophysical placenta-pulmonary coupling model used the umbilical artery pulsatility index (PI) and a metabolomic hypoxia-steroid score to represent placental reserve, while backscatter integrals and elastography were used to characterize fetal lung properties. Constrained by this model, a dual-branch network was developed: (i) a cross-modal attention Transformer with self-supervised contrastive learning to generate unsupervised FLM stages from fused representations and (ii) a spatiotemporal convolution-LSTM network to predict individualized GC dosing and the optimal administration window. A composite loss penalized both projected respiratory distress syndrome (RDS) risk and the biomarker-derived neurotoxicity index. Results: The cross-modal representations clustered into four distinct maturity stages matching biochemical benchmarks, with an inter-stage silhouette score of 0.72. A downstream classifier achieved 92.3% accuracy in discriminating early from late maturity. The dosing branch predicted the GC dose within ±0.5 mg of clinically prescribed regimens and reduced projected RDS risk by 27% compared to standard dosing, while maintaining the biomarker-derived neurotoxicity index below the prespecified threshold. Conclusion: A mechanism-guided, multimodal AI framework constrained by placenta-pulmonary physiology transforms imaging features into traceable decision pathways that align with clinical cognition. This interpretable framework may enable non-invasive FLM staging and individualized GC therapy, providing hypothesis-generating decision support that warrants external validation and prospective trials.

Overview

CoChat Summary

## TL;DR

This prospective cohort study developed a physiology-guided multimodal AI system that used ultrasound, Doppler, elastography, and maternal metabolomics to stage fetal lung maturity and recommend individualized antenatal glucocorticoid dosing. In 320 pregnancies, the model formed four maturity stages, reached 92.3% accuracy for early vs late maturity, and was associated with a 27% reduction in model-projected RDS risk versus standard dosing.

## Key findings

  • In 320 singleton pregnancies with weekly scans from 28 to 36 weeks, multimodal self-supervised representations clustered into 4 maturity stages with an inter-stage silhouette score of 0.72 (95% CI 0.69–0.75).
  • The downstream classifier achieved 92.3% accuracy (95% CI 89.5–94.6%) for distinguishing early (Stages 1–2) from late (Stages 3–4) maturity, with 91.7% sensitivity and 92.9% specificity.
  • The dosing branch predicted glucocorticoid dose with MAE 0.31 ± 0.17 mg; 80% of predictions were within ±0.5 mg of prescribed doses, and predicted administration windows had a median deviation of 0.4 days.
  • Model-guided dosing reduced projected RDS risk from 30.8% to 22.5% in the total cohort, a 27% relative reduction; the reduction was 33% in HDP pregnancies (41.2% to 27.6%) and 21% in normotensive pregnancies (20.5% to 16.2%).

## Methods

  • Design/sample: Prospective cohort study of 320 singleton pregnancies enrolled at two tertiary perinatal centers from January 2021 to December 2022; 160 had hypertensive disorders of pregnancy and 160 were normotensive controls, with weekly data collected from 28+0 to 36+0 weeks.
  • Data/modalities: 2D and 3D ultrasound, shear-wave elastography, umbilical artery Doppler, and maternal plasma metabolomics (a 45-metabolite panel) were collected longitudinally.
  • Analysis/modeling: A biophysical placenta–pulmonary coupling model was combined with a dual-branch deep network: a Transformer-based cross-modal staging branch trained with contrastive learning and a spatiotemporal convolution–LSTM dosing branch; evaluation used patient-level splits, Gaussian mixture clustering, silhouette score, accuracy, MAE, and risk/safety comparisons.

## Limitations

  • The paper states that external generalizability remains to be established and that performance may vary across vendors, transducers, acquisition settings, and clinical workflows; external multi-vendor validation is warranted.
  • The neurotoxicity index is a surrogate biomarker and “not a validated proxy for long-term neurodevelopmental outcomes,” so safety conclusions are preliminary.
  • Respiratory benefit was based on model-projected RDS risk rather than adjudicated neonatal RDS outcomes; the paper says prospective studies with neonatal endpoints are needed.
  • The text also notes that higher-order nonlinearities (e.g., cytokine-mediated feedback) were not incorporated, which is a limitation evident from the model description.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3389/fmed.2026.1791481

Cites

21 of 24 references verified2 citation errors · 1 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Patient inclusion and exclusion criteria, along with the data collection workflow in our study.
Figure 1 Patient inclusion and exclusion criteria, along with the data collection workflow in our study.CC BY · PubMed Central
Figure 2 Alignment with the placenta–pulmonary coupling model. (A) Scatter plot of predicted L(t) versus the coupling model L(t), with regression line (r = 0.84). (B) Coupling coefficient k across maturity stages, showing declining placental influence with advancing maturity; p < 0.001.
Figure 2 Alignment with the placenta–pulmonary coupling model. (A) Scatter plot of predicted L(t) versus the coupling model L(t), with regression line (r = 0.84). (B) Coupling coefficient k across maturity stages, showing declining placental influence with advancing maturity; p < 0.001.CC BY · PubMed Central
Figure 3 The workflow.
Figure 3 The workflow.CC BY · PubMed Central
Figure 4 GC dose prediction performance. (A) Bland–Altman plot comparing predicted versus prescribed GC doses. The solid line represents the mean difference (0.03 mg), and the dashed lines indicate the 95% limits of agreement. (B) Distribution of deviations between predicted and clinical administration windows (days).
Figure 4 GC dose prediction performance. (A) Bland–Altman plot comparing predicted versus prescribed GC doses. The solid line represents the mean difference (0.03 mg), and the dashed lines indicate the 95% limits of agreement. (B) Distribution of deviations between predicted and clinical administration windows (days).CC BY · PubMed Central
Figure 5 Model-projected respiratory risk and biomarker-derived neurotoxicity indices under model-guided versus standard GC dosing. (A) Model-projected RDS risk in the total cohort and subgroups. (B) Distribution of biomarker-derived neurotoxicity index values, with the dashed line indicating the prespecified threshold (1.0). p < 0.001, p < 0.01.
Figure 5 Model-projected respiratory risk and biomarker-derived neurotoxicity indices under model-guided versus standard GC dosing. (A) Model-projected RDS risk in the total cohort and subgroups. (B) Distribution of biomarker-derived neurotoxicity index values, with the dashed line indicating the prespecified threshold (1.0). p < 0.001, p < 0.01.CC BY · PubMed Central

Verification

Claims the paper makes

7 Valid · 4 Not valid · 12 Not checked · 68 Own results and interpretations

31 more claims

Create a free account to see all 91 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 21
  • It may state two separate claims as one. 44

Create a free account to see these claims.

Citations

Cites

Show all 24 references

Research works printed but not linked: 3

  1. Vaswani, 2017, Advances in Neural Information Processing Systems 30 (NeurIPS 2017), 5998
  2. Grill, 2020, NeurIPS, 21271
  3. Grill, 2020, Advances in Neural Information Processing Systems 33 (NeurIPS 2020), 21271

Metadata

Authors
Ma, B., Ran, J., Pan, T., Yan, F., Yue, Z., Yao, Y., Li, Y., & Nie, F.
Affiliations
  • Gansu Provincial Maternal and Child Health Hospital (CN)
  • Lanzhou University Second Hospital (CN)
Corresponding author
Nie, Fang
Venue
Frontiers in Medicine
Published
2026-03-12
Volume
13
Pages
1791481-1791481
ISSN
2296-858X (ISSN-L)
OpenAlex
W7135129954
Cited by
0
References
20
FWCI
0.00
Type
journal article
Language
en
Open access
Locations
  • Frontiers in Medicine (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • Gansu Province Clinical Research Center for Ultrasonography (2020-0411-SFC-0021)
  • The Gansu Provincial Natural Science Foundation (23JRRA1383)
Topics
  • Health Sciences › Medicine › Pulmonary and Respiratory Medicine › Neonatal Respiratory Health Research
  • Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Radiomics and Machine Learning in Medical Imaging
Keywords
fetal lung maturity, antenatal glucocorticoids, hypertensive disorders of pregnancy, respiratory distress syndrome, shear wave elastography, umbilical artery pulsatility index, fetal lung ultrasound, plasma metabolomics, explainable AI, decision support

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Placenta-pulmonary coupling-guided multimodal AI for fetal lung maturity staging and individualized glucocorticoid therapy. https://app.cochat.ai/papers/10.3389/fmed.2026.1791481
The paper
Ma, B., Ran, J., Pan, T., Yan, F., Yue, Z., Yao, Y., Li, Y., & Nie, F. (2026). Placenta-pulmonary coupling-guided multimodal AI for fetal lung maturity staging and individualized glucocorticoid therapy. Frontiers in Medicine. https://doi.org/10.3389/fmed.2026.1791481

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.