Multihipposenet: multi-task deep learning for automated DDH assessment in infant hip ultrasound
4 affiliations
- 1Department of Ultrasound, Haidian District Maternal and Child Health Hospital , ,
- 2Department of Prenatal Diagnosis, Haidian District Maternal and Child Health Hospital , ,
- 3Senior Department of Orthopaedics, Fourth Medical Center of PLA General Hospital , ,
- 4Director of the Department of Child Health and the Division of Comprehensive Early Childhood Development, Haidian District Maternal and Child Health Hospital , ,
Abstract
Introduction: Developmental dysplasia of the hip (DDH) is a common skeletal condition in infants and young children. For infants under six months, diagnosis relies on Graf ultrasonography to measure the α and β angles, which assess hip-joint development. However, this method requires experienced sonographers with expertise in complex anatomy.
Methods: In this study, we propose an end-to-end multi-task learning network that improves DDH assessment by jointly exploiting complementary anatomical information. First, we use ResNet-50 with a feature pyramid network (FPN) and a contextual semantic fusion module to extract multi-scale features from hip anatomy. Second, we employ a multi-expert gated channel transformation (ME-GCT) unit that explicitly models the relationships between tasks and channels to balance collaboration and conflict across tasks. Finally, a multi-task graph adaptive learning mechanism (MT-GALM) mines the hidden dynamic dependencies between adjacent channels across tasks. Two datasets comprising 1,355 annotated hip ultrasound images from 566 infants were constructed for this work.
Results: Across the patient-level cross-validation test folds, 97.97% and 96.81% of the estimated α and β angles fall within 5 degrees of the expert reference.
Discussion: Because validation was restricted to Graf type I and type II hips, these results should be interpreted as evidence of technical feasibility rather than of established clinical utility; evaluation on type III and type IV hips is required before clinical deployment.
Related
- Multidimensional institutional learning curve of ultrasound-guided percutaneous transthoracic lung biopsy: a cumulative sum analysis with case-mix assessment
- Quantitative Glandular Tissue Component Assessment on Breast Ultrasound Using an Anatomically Constrained Deep Learning Framework
- A novel morphological assessment approach for abdominal aortic aneurysms: automated segmentation and measurement using a semi-supervised deep learning algorithm
- A deep learning model for automated assessment of neonates at risk of impaired breast crawl: a proof-of-concept video-based study
- A Semi-Automated Interactive Annotation Framework for Intraoperative Neurosurgical Ultrasound: A Proof-of-Concept Study
BibTeX
@article{chi2026multihipposenet,
title = {{Multihipposenet: multi-task deep learning for automated DDH assessment in infant hip ultrasound}},
author = {Wenjing Chi and Hong Qi and Huawei Zhao and Hui Cheng and Quanlian Yuan and Shuai Zheng},
journal = {Frontiers in pediatrics},
year = {2026},
month = oct,
doi = {10.3389/fped.2026.1939157},
url = {https://europepmc.org/article/PMC/PMC13719586},
note = {CC BY},
}
RIS
TY - JOUR
TI - Multihipposenet: multi-task deep learning for automated DDH assessment in infant hip ultrasound
AU - Wenjing Chi
AU - Hong Qi
AU - Huawei Zhao
AU - Hui Cheng
AU - Quanlian Yuan
AU - Shuai Zheng
PY - 2026
DA - 2026/10/16
JO - Frontiers in pediatrics
DO - 10.3389/fped.2026.1939157
UR - https://europepmc.org/article/PMC/PMC13719586
AB - Introduction: Developmental dysplasia of the hip (DDH) is a common skeletal condition in infants and young children. For infants under six months, diagnosis relies on Graf ultrasonography to measure the α and β angles, which assess hip-joint development. However, this method requires experienced sonographers with expertise in complex anatomy. Methods: In this study, we propose an end-to-end multi-task learning network that improves DDH assessment by jointly exploiting complementary anatomical information. First, we use ResNet-50 with a feature pyramid network (FPN) and a contextual semantic fusion module to extract multi-scale features from hip anatomy. Second, we employ a multi-expert gated channel transformation (ME-GCT) unit that explicitly models the relationships between tasks and channels to balance collaboration and conflict across tasks. Finally, a multi-task graph adaptive learning mechanism (MT-GALM) mines the hidden dynamic dependencies between adjacent channels across tasks. Two datasets comprising 1,355 annotated hip ultrasound images from 566 infants were constructed for this work. Results: Across the patient-level cross-validation test folds, 97.97% and 96.81% of the estimated α and β angles fall within 5 degrees of the expert reference. Discussion: Because validation was restricted to Graf type I and type II hips, these results should be interpreted as evidence of technical feasibility rather than of established clinical utility; evaluation on type III and type IV hips is required before clinical deployment.
SN - 2296-2360
ER -