Deep learning-based early prediction of gestational diabetes mellitus through first-trimester placental texture analysis

Clinical Snapshot

Key Findings

• P Value: P < 0.05 for Combined Model vs. each single-modality model; Hosmer-Lemeshow calibration P > 0.05 (good calibration)

• Effect Size: Combined Model AUC = 0.879 (internal validation); statistically superior to all single-modality models (P < 0.05)

• Primary Outcome: Prediction of gestational diabetes mellitus using a multimodal fusion model applied to first-trimester placental ultrasound

• Nnt Or Sensitivity: Sensitivity, specificity, PPV, NPV, and likelihood ratios not reported; DCA demonstrates net clinical benefit over treat-all and treat-none strategies across a range of threshold probabilities; external validation AUC not disclosed

• Confidence Interval: Not reported in the abstract

Clinical Application

Currently low feasibility for routine clinical implementation. The model requires: (1) standardised placental ultrasound image acquisition protocols; (2) radiomic feature extraction software; (3) deep learning inference infrastructure; (4) integration with clinical data systems. These requirements exceed the capacity of most antenatal care settings globally. Prospective clinical trials with pre-specified thresholds and workflow integration are required before adoption. In Australia, GDM screening is recommended at 24–28 weeks via the 75g OGTT per ADIPS (Australasian Diabetes in Pregnancy Society) and RANZCOG guidelines. First-trimester prediction tools could theoretically enable earlier lifestyle intervention and monitoring for high-risk women, consistent with RACGP preventive care frameworks. However, this model has not been validated in Australian or ethnically diverse populations — a significant limitation given the higher GDM prevalence among South Asian, East Asian, and Aboriginal and Torres Strait Islander women in Australia. The TGA would require prospective Australian validation data before regulatory clearance of any such diagnostic software as a Class IIb medical device. PBS listing of associated monitoring or intervention programs would depend on demonstrated clinical and cost-effectiveness in local populations. RANZCOG and ADIPS would need to review and endorse any change to screening pathways. Pregnant women in the first trimester (11–14 weeks) undergoing routine nuchal translucency ultrasound, particularly those with risk factors for GDM (obesity, family history, prior GDM, polycystic ovary syndrome)

Abstract

INTRODUCTION: This study aimed to develop a multi-parameter fusion model for early GDM risk prediction and validate its performance through external multicenter testing. METHODS: A total of 628 pregnant women at 11+0-13+6 weeks were enrolled from two medical centers. The Center I cohort was divided into training (n = 356) and testing sets (n = 153). Radiomic features (1,289) and deep learning features (2,048) were extracted from placental ultrasound images. Feature-level fusion resulted in 3337 features, which were selected using Spearman correlation, mRMR, and LASSO. Five models were built: Rad Model, DTL Model, DLR Model, Clinic Model, and Combined Model. Performance was assessed using ROC analysis, DCA, and calibration curves. RESULTS: The Combined Model achieved the best overall performance, with an area under the ROC curve (AUC) of 0.879 in the internal validation, significantly outperforming any single-modality model (P < 0.05). DCA demonstrated that the fusion-based model provided higher net clinical benefit across a wide range of threshold probabilities compared with both "treat-all" and "treat-none" strategies. The calibration curve showed excellent agreement between predicted and observed probabilities (Hosmer-Lemeshow test, P > 0.05). DISCUSSION: The multimodal fusion model enhanced early GDM prediction by detecting subtle placental changes in first-trimester, enabling timely intervention and personalized decision-making.

References

1. 1.Peng, Y., Zhang, X., Fang, M., He, Y., Yuan, H., Du, W., Feng, W., Gan, L., & Zhang, J. (2026). Deep learning-based early prediction of gestational diabetes mellitus through first-trimester placental texture analysis. Placenta. https://doi.org/10.1016/j.placenta.2026.04.016