Open Collections

Abstract

Developmental dysplasia of the hip (DDH) is a common pediatric hip dysmorphology that, when undetected or inadequately treated in early life, can lead to long-term morbidity including gait abnormalities, chronic pain, and early-onset osteoarthritis. DDH is typically diagnosed using conventional two-dimensional ultrasound (2DUS), a modality that is highly operator-dependent and prone to substantial inter- and intra-operator variability, leading to inconsistent diagnoses. Automated DDH metric extraction from three-dimensional ultrasound (3DUS) using deep learning has demonstrated improved diagnostic reliability; however, the high cost and limited availability of 3DUS systems restrict their widespread adoption in routine pediatric care. To address this gap, we developed and evaluated an optically tracked 2D ultrasound (OTUS) system, which has previously demonstrated high spatial accuracy and may represent a more accessible alternative for three-dimensional assessment. In this study, we acquired a multimodal dataset comprising repeated OTUS scans and a single reference 3DUS scan per hip from pediatric patients. We found that (1) the repeatability characteristics of OTUS scans differed from those of 3DUS, but exhibited comparable ranges of variability across repeated scans of the same hip, and (2) DDH metrics of consistently clinically acceptable quality could not be robustly reproduced across the full dataset, but in cases where OTUS scan quality was high, we obtained anatomically plausible segmentations and DDH metrics that closely resembled those derived from the corresponding 3DUS scans. These findings indicated that while further methodological refinements, such as improved scan acquisition quality and the incorporation of geometric constraints for more robust metric extraction, will be needed before clinical deployment, OTUS demonstrated promise as a more cost-effective approach for improving the reliability and accessibility of DDH screening.