Title Development of a tool for assessing hip abduction during hip-scipa casting in patients with developmental dysplasia of the hip
Translation of Title Įrankio, skirto klubo sąnario displazija sergančių pacientų klubo sąnario atvedimo įvertinimui gipsavimo metu, kūrimas.
Authors Garšvaitė, Inga
Full Text Download
Pages 101
Keywords [eng] hip abduction angle, MediaPipe BlazePose, One Euro filter, pose estimation, pediatric orthopedics, Developmental Dysplasia of the Hip, avascular necrosis
Abstract [eng] Developmental Dysplasia of the Hip (DDH) is a spectrum of hip abnormalities in infants ranging from joint instability to complete hip dislocation. In the surgical management of DDH, the correct hip abduction angle is essential for treatment success: insufficient abduction increases the risk of hip redislocation, while excessive abduction increases the risk of avascular necrosis. Despite the clinical importance of maintaining a ``safe zone," casting decisions are often based on subjective visual estimation. This thesis presents the development and validation of a Technological Readiness Level (TRL) 4 prototype for estimating hip abduction angle from monocular video and pose estimation. The prototype was evaluated using data collected from eleven healthy adults under a standardized protocol. Results showed a systematic underestimation of the angle, and the mean absolute error increased at higher abduction values, suggesting range compression. One Euro filtering reduced frame-to-frame jitter by approximately 83%, but did not improve abduction angle estimation. Zone-based decision support demonstrated high performance, with a Support Vector Machine (SVM) classification model providing the best Leave-one-patient-out generalization accuracy (0.909) in the reported experiments. The Support Vector Regressor performed best among the regression models, with R^2 = 0.713, suggesting that the current prototype may be more useful for zone-based decision support than for precise angle estimation.
Dissertation Institution Vilniaus universitetas.
Type Master thesis
Language English
Publication date 2026