AUTOMATING 3D ORTHOPAEDIC MEASUREMENTS USING WEIGHTBEARING CT IMAGING

Foot and ankle alignment is a fundamental aspect of human biomechanics, playing a crucial role in gait, posture, and overall mobility. However, deformities in foot and ankle alignment can lead to pain, impaired mobility, and a reduced quality of life. Accurate assessment and precise measurement of alignment are critical for proper diagnosis, treatment planning, and monitoring of therapeutic outcomes. Traditionally, foot and ankle alignment evaluation has relied on manual measurements performed on 2D standing radiographs. While these methods provide valuable anatomical insights, they have inherent limitations, including the inability to capture the complex three-dimensional nature of foot alignment under physiological loading conditions. Additionally, manual measurements are susceptible to observer error, leading to variability in results and potential discrepancies in clinical decision-making. Weight-bearing CT (WBCT) enables a more precise 3D assessment under natural loading conditions. However, its clinical adoption is hindered by the time-intensive nature of manual segmentation and landmark selection. Deep learning (DL) and statistical shape modeling (SSM) offer the potential to automate 3D image segmentation and accurately identify anatomical landmarks. This work aims to integrate WBCT with advanced DL algorithms to automate and enhance 3D foot and ankle alignment quantification, reducing analysis time and improving measurement consistency for clinical applications. Thirty-two patients who underwent WBCT of the foot and ankle were retrospectively included. A 3D nnU-Net model was trained and validated on 45 cases to automate the segmentation of bony structures. Following the automated segmentation, 35 clinically relevant 3D measurements were automatically computed. Automated measurements were assessed for accuracy against manual measurements, while the latter were analyzed for inter-observer reliability. DL-based segmentation achieved a mean Dice coefficient of 0.95 and a mean Hausdorff distance of 1.41 mm. A good to excellent reliability was observed, with a mean prediction error of under 2 degrees for all angles except the talonavicular coverage angle and the distal metatarsal articular angle. This study introduces a fully automated framework for quantifying foot and ankle alignment, demonstrating reliability comparable to current clinical measurement practices. This tool holds promise for implementation in clinical settings, benefiting both radiologists and surgeons. Future studies are encouraged to assess the tool's impact on streamlining image assessment workflows in a clinical environment.