Abstract
To develop a comprehensive deep learning and radiomics-based multi-task pipeline for the detection and classification of key fetal anatomical structures in first-trimester ultrasound images, using a diverse multi-center dataset to ensure high variability, reproducibility, and generalizability. A total of 4,532 fetal ultrasound scans (gestational age 11–14 weeks), retrospectively collected from nine medical centers, were included in this study. Two detection models, You Only Look Once version 11 (YOLOv11) and shifted window transformer (Swin Transformer), were trained to localize nine fetal brain and craniofacial structures. From each detected region, 215 radiomic features and 1,792 deep features were extracted. Feature stability was ensured through intra-class correlation coefficient (ICC) filtering (threshold ≥ 0.75), Pearson correlation analysis, and least absolute shrinkage and selection operator (LASSO) regression. The selected features were then used to train a Transformer-based model for tabular data (TabTransformer) to classify fetal anatomical structures into clinically defined categories based on their sonographic appearance. Model performance was evaluated using Accuracy, area under the receiver operating characteristic curve (AUC), and Sensitivity across training, internal validation, and external test datasets. Fusion models integrating radiomic and deep features consistently outperformed single-modality models in both detection and classification. On the external test set, classification accuracy reached 96.1%, with AUCs up to 96.89%, and sensitivity exceeding 95% for key anatomical structures. Swin Transformer achieved superior localization performance compared to YOLOv11, with Intersection over Union (IoU) values up to 0.97 and F1-scores ≥ 0.94. Feature reproducibility remained above 75% across centers. The TabTransformer classifier demonstrated strong generalization and robustness, effectively leveraging the fused feature space for high-precision classification. This study presents the fully integrated, multi-task framework for fetal anatomical structure detection and classification using multi-center ultrasound data. The proposed approach demonstrates high reproducibility and diagnostic performance, offering strong clinical potential for early and objective fetal anomaly screening in the first trimester.
Data availability
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Abbreviations
- AUC:
-
Area under the receiver operating characteristic curve
- CI:
-
Confidence interval
- CM:
-
Cisterna magna
- CNN:
-
Convolutional neural network
- DL:
-
Deep learning
- F1-score:
-
Harmonic mean of precision and recall
- ICC:
-
Intraclass correlation coefficient
- IoU:
-
Intersection over Union
- IT:
-
Intracranial translucency
- LASSO:
-
Least absolute shrinkage and selection operator
- mAP:
-
Mean average precision
- NT:
-
Nuchal translucency
- PCA:
-
Principal component analysis
- ROC:
-
Receiver operating characteristic
- ROI:
-
Region of interest
- SD:
-
Standard deviation
- SE:
-
Standard error
- SMOTE:
-
Synthetic minority over-sampling technique
- ViT:
-
Vision transformer
- YOLOv11:
-
You only look once, Version 11
- Swin transformer:
-
Shifted window transformer
- TabTransformer:
-
Transformer model for tabular data
- IBSI:
-
Image biomarker standardisation initiative
- SGD:
-
Stochastic gradient descent
- GPU:
-
Graphics processing unit
- MRI:
-
Magnetic resonance imaging
- 2D:
-
Two-dimensional
- 3D:
-
Three-dimensional
References
Sriraam, N., Chinta, B., Suresh, S. & Sudharshan, S. Ultrasound imaging-based recognition of prenatal anomalies: a systematic clinical engineering review. Prog Biomed. Eng. 6 (2), 023002 (2024).
Pozza, A. et al. Utility of fetal cardiac resonance imaging in prenatal clinical practice: current state of the art. Diagnostics (Basel). 13 (23), 3523 (2023).
Pelayo-Delgado, I., Gómez-Montes, E. & Álvaro-Navidad, M. Update on second trimester ultrasound scanning in pregnancy. Clínica e Investigación en Ginecología y Obstetricia 52(1), 100997 (2025).
Sepulveda, W., Wong, A. E., Ximenes, R. & Meagher, S. The first-trimester fetal anatomy scan. In: Obstetric Imaging: Fetal Diagnosis and Care-E-Book. 24 (2025).
Puerto, B., Azumendi, P., Corrales, C. & Azumendi, G. How to perform a fetal neurosonography: Key points. Clin. Invest. Ginecol. Obstet. 52, 101050 (2025).
Gupta, N., Hiremath, S. B., Gauthier, I., Wilson, N. & Miller, E. Pediatric neurosonography: Comprehensive review and systematic approach. Can. Assoc. Radiol. J. 76(3), 519–533 (2025).
Kim, R. et al. Artificial intelligence based automatic classification, annotation, and measurement of the fetal heart using HeartAssist. Sci. Rep. 15 (1), 13055 (2025).
Qi, Y. et al. Multi-Center study on deep learning-assisted detection and classification of fetal central nervous system anomalies using ultrasound imaging. https://arXiv.org/abs/250102000. (2025).
Salini, Y., Mohanty, S. N., Ramesh, J. V. N., Yang, M. & Chalapathi, M. M. V. Cardiotocography data analysis for fetal health classification using machine learning models. IEEE Access. 12, 26005–26022 (2024).
Yin, Y. & Bingi, Y. Using machine learning to classify human fetal health and analyze feature importance. BioMedInformatics 3 (2), 280–298 (2023).
Mushtaq, G. & Veningston, K. AI driven interpretable deep learning based fetal health classification. SLAS Technol. 29 (6), 100206 (2024).
Montin, E. et al. Radiomics for precision diagnosis of FAI: How close are we to clinical translation? A Multi-Center Validation of a Single-Center Trained Model. J. Clin. Med. 14(12), 4042 (2025).
Krishna, N. S. et al. Generalizability of AI-based image segmentation and centering estimation algorithm: A multi-region, multi-center, and multi-scanner study. Radiat. Prot. Dosimetry 201(6), 441–449 (2025).
Yin, S., Ming, J., Chen, H., Sun, Y. & Jiang, C. Integrating deep learning and radiomics for preoperative glioma grading using multi-center MRI data. Sci. Rep. 15(1), 36756 (2025).
Richter, M. et al. Generalizability of clinical prediction models in mental health. Mol. Psychiatry 1–8. (2025).
Degtiar, I. & Rose, S. A review of generalizability and transportability. Annu. Rev. Stat. Its Appl. 10 (1), 501–524 (2023).
Maleki, F. et al. Generalizability of machine learning models: Quantitative evaluation of three methodological pitfalls. Radiol. Artif. Intell. 5(1), e220028 (2022).
Ambsdorf, J. et al. General methods make great domain-specific foundation models: A case-study on fetal ultrasound. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. 271–281 (Springer, 2025).
Fiorentino, M. C., Migliorelli, G., Villani, F. P., Frontoni, E. & Moccia, S. Contrastive prototype federated learning against noisy labels in fetal standard plane detection. Int. J. Comput. Assist. Radiol. Surg. https://doi.org/10.1007/s11548-025-03400-6 (2025).
Wang, F. et al. Fusing radiomic features with deep representations for gestational age estimation in fetal ultrasound images. In: International Conference on Medical Image Computing and Computer-Assisted Intervention. 230–240 (Springer, 2025).
Lasala, A., Fiorentino, M. C., Micera, S., Bandini, A. & Moccia, S. Exploiting class activation mappings as prior to generate fetal brain ultrasound images with GANs. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. 2023, 1–4 (2023).
Huang, S. et al. Semi-supervised fetal brain parcellation via hierarchical learning framework. Med. Image Anal. 103835. (2025).
Islam, U. et al. Fetal-Net: enhancing Maternal-Fetal ultrasound interpretation through Multi-Scale convolutional neural networks and Transformers. Sci. Rep. 15 (1), 25665 (2025).
Khanam, R. & Hussain, M. YOLOv11: An overview of the key architectural enhancements. https://arXiv.org/abs/241017725 (2024).
Wei, W. et al. YOLOv11-based multi-task learning for enhanced bone fracture detection and classification in X-ray images. J. Radiat. Res. Appl. Sci. 18(1), 101309 (2025).
Cao, H. et al. Swin-UNet: Unet-like pure transformer for medical image segmentation. In: European Conference on Computer Vision. 205–218 (Springer, 2022).
Liu, Z. et al. Swin Transformer: Hierarchical vision transformer using shifted windows. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. 10012–10022 (2021).
Scapicchio, C. et al. A deep look into radiomics. Radiol. Med. 126(10), 1296–1311 (2021).
Yip, S. S. F. & Aerts, H. J. W. L. Applications and limitations of radiomics. Phys. Med. Biol. 61 (13), R150 (2016).
Rezaeijo, S. M. et al. Neighboring tissues as diagnostic windows: Neighborhood effects in radiomic detection of pancreatic ductal adenocarcinoma. Comput. Methods Programs Biomed. https://doi.org/10.1016/j.cmpb.2025.109056 (2025).
Khan, A. et al. A survey of the vision transformers and their CNN-transformer based variants. Artif. Intell. Rev. 56(Suppl 3), 2917–2970 (2023).
Pereira, G. A. & Hussain, M. A review of transformer-based models for computer vision tasks: Capturing global context and spatial relationships. https://arXiv.org/abs/240815178 (2024).
Wang, J. et al. A machine learning model based on placental magnetic resonance imaging and clinical factors to predict fetal growth restriction. BMC Pregnancy Childbirth. 25 (1), 325 (2025).
Lai, H. et al. Radiomics-based correlation analysis of fetal brain MRI features and children’s neurodevelopmental outcomes in monochorionic twins. BMC Pregnancy Childbirth. 25 (1), 1040 (2025).
Zuo, M. et al. A nomogram based on MR radiomics and MR sign score for prenatal diagnosis of placenta accreta spectrum disorders and risk assessment of adverse clinical outcomes. Abdom. Radiol. 1–12. (2025).
Xu, F. et al. Prediction of clinical pregnancy after frozen embryo transfer based on ultrasound radiomics: An analysis based on the optimal periendometrial zone. BMC Pregnancy Childbirth 25(1), 391 (2025).
Koo, T. K. & Li, M. Y. A guideline of selecting and reporting intraclass correlation coefficients for reliability research. J. Chiropr. Med. 15(2), 155–163 (2016).
Kim, Y. & Kim, J. Gradient LASSO for feature selection. In: Proceedings of the Twenty-First International Conference on Machine Learning, p. 60. (2004).
Gomes, R., Pham, T., He, N., Kamrowski, C. & Wildenberg, J. Analysis of Swin-UNet vision transformer for Inferior Vena Cava filter segmentation from CT scans. Artif. Intell. Life Sci. 4, 100084 (2023).
Liang, J. et al. SwinIR: Image restoration using Swin transformer. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. 1833–44 (2021).
Liu, Z. et al. Swin Transformer V2: Scaling up capacity and resolution. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 12009–19 (2022).
Hatamizadeh, A. et al. Swin UNETR: Swin transformers for semantic segmentation of brain tumors in MRI images. In: International MICCAI Brainlesion Workshop. 272–284 (Springer, 2021).
Kotthapalli, M., Ravipati, D. & Bhatia, R. YOLOv1 to YOLOv11: A Comprehensive Survey of Real-Time Object Detection Innovations and Challenges. https://arXiv.org/abs/250802067 (2025).
Tan, M., Pang, R. & Le, Q. V. EfficientDet: Scalable and efficient object detection. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 10781–90 (2020).
Powers, D. M. W. Evaluation: from precision, recall and F-measure to ROC, informedness, markedness and correlation. https://arXiv.org/abs/201016061 (2020).
Wang, X. et al. Advanced network intrusion detection with TabTransformer. J. Theory Pract. Eng. Sci. 4(03), 191–8 (2024).
Huang, X., Khetan, A., Cvitkovic, M. & Karnin, Z. TabTransformer: Tabular data modeling using contextual embeddings. https://arXiv.org/abs/201206678 (2020).
Insalata, B., Schmidt, F. & Vlassov, V. Multimodal survival prediction using TabTransformer and BioClinicalBERT on MIMIC-III. In: 2024 IEEE International Conference on Big Data (BigData). 1986–92 (IEEE, 2024).
Cichosz, P. Assessing the quality of classification models: Performance measures and evaluation procedures. Cent. Eur. J. Eng. 1 (2), 132–158 (2011).
Du, Y. et al. Ultrasound-based radiomics technology in fetal lung texture analysis prediction of neonatal respiratory morbidity. Sci. Rep. 12(1), 12747 (2022).
Drukker, L. et al. Clinical workflow of sonographers performing fetal anomaly ultrasound scans: Deep-learning‐based analysis. Ultrasound Obstet. Gynecol. 60(6), 759–65 (2022).
Krishna, T. B. & Kokil, P. Standard fetal ultrasound plane classification based on stacked ensemble of deep learning models. Expert Syst. Appl. 238, 122153 (2024).
Gofer, S., Haik, O., Bardin, R., Gilboa, Y. & Perlman, S. Machine learning algorithms for classification of first-trimester fetal brain ultrasound images. J. Ultrasound Med. 41 (7), 1773–1779 (2022).
Hesse, L. S. et al. Subcortical segmentation of the fetal brain in 3D ultrasound using deep learning. Neuroimage 254, 119117 (2022).
Coronado-Gutiérrez, D. et al. Automatic deep learning-based pipeline for automatic delineation and measurement of fetal brain structures in routine mid-trimester ultrasound images. Fetal Diagn. Ther. 50 (6), 480–490 (2023).
Prieto, J. C. et al. An automated framework for image classification and segmentation of fetal ultrasound images for gestational age estimation. In Medical Imaging 2021: Image Processing 453–462 (SPIE, 2021).
Ghabri, H. et al. Transfer learning for accurate fetal organ classification from ultrasound images: A potential tool for maternal healthcare providers. Sci Rep 13(1), 17904 (2023).
Author information
Authors and Affiliations
Contributions
Conceptualization: Xuan Zhou, Jie Wan, and Pin Li. Data Curation: Fengjie Sun and Ruxin Wang. Formal Analysis: Xuan Zhou, Jie Wan, and Yafei Yan. Methodology: Pin Li, Cuihua Wang, and Xuan Zhou. Resources: Cuihua Wang and Pin Li. Supervision: Cuihua Wang and Pin Li. Validation: Jie Wan, Fengjie Sun, and Yafei Yan. Visualization: Xuan Zhou and Jie Wan. Writing—Original Draft: Xuan Zhou and Jie Wan. Writing—Review & Editing: Pin Li and Cuihua Wang.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Ethics approval and consent to participate
This study was conducted in accordance with the Declaration of Helsinki. Ethical approval and the requirement for informed consent were waived by the Institutional Review Board of the Affiliated Hospital of Hebei University (Baoding, China) because the study involved retrospective analysis of fully anonymized data and posed no risk to participants.
Additional information
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Below is the link to the electronic supplementary material.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Zhou, X., Wan, J., Sun, F. et al. A multi-task deep learning and radiomics framework for fetal anatomical structure detection and classification in ultrasound imaging. Sci Rep (2026). https://doi.org/10.1038/s41598-026-41635-8
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-026-41635-8
