Ultrasound-based statistical shape modeling for prognosis in unstable hip dysplasia

Background: Current methods to classify developmental dysplasia of the hip (DDH) on ultrasound (US) images, such as the Graf method, provide limited prognostic information. This study aimed to improve the prediction of the clinical course and outcome at age five of decentered hips, diagnosed on the first US made in the first months after birth, by identifying acetabular shape variants on these US images using a statistical shape model (SSM). Patients and Methods: US images of the hip were retrieved from a single-center retrospective cohort of patients with DDH Graf type D/III/IV. A SSM was created from the US images made at initial diagnosis. The association between the identified acetabular shape variants and an unfavorable outcome (residual DDH at age five and open reduction and/or a pelvic osteotomy before age five) was established with multivariable regression models. Results: 92 decentered dysplastic hips with full history could be retrieved from the database and were included. At age five, 12 patients (13%) had undergone open reduction, 13 (14%) had a pelvic osteotomy, and 32 (35%) patients showed residual DDH. Four shape variants represented 95% of the variance in acetabular shape. Mode 4 was associated with an unfavorable outcome (odds ratio (OR): 1.80 (95% CI 1.12–2.90). Mode 1 was associated with less risk on open reductions or pelvic osteotomies (OR: 0.56 (95% CI 0.33–0.96). Conclusions: A potential new method of analyzing US images for DDH using SSM established four distinct acetabular shapes on neonatal US images with unstable DDH, of which two were associated with outcomes at five years of age. This tool could serve as a basis for a better prediction of outcome and a more personalized and effective guide for treatment.