The influence of anatomical shape variations of wrist bones on kinematic parameter extraction in CT scans

Four-Dimensional Computed Tomography (4DCT) shows promise in diagnosing wrist pathologies such as scapholunate ligament lesions (SLL). Two parameters indicative of SLL lesions in these scans, the scapholunate angle (SLA) and the capitolunate angle (CLA), might be affected by anatomical shape variations. Therefore, this study characterizes the impact of anatomical shape variations on SLA and CLA estimates from 4DCT scans. 3D CT images retrieved from 4DCT scans of healthy wrists were used. Scans were automatically segmented using an Artificial Intelligence-based algorithm. A statistical shape model (SSM) was created for each bone separately. The quality of each SSM was assessed using a leave-one-out approach by calculating the root mean squared error (RMSE). Next, local coordinate systems (LCSs) were assigned to the SSMs while shape variations were introduced. The rotational deviation of the LCSs was calculated by a combined rotation around the three axes. Finally, the SLAs and CLAs were calculated for all possible bone shape combinations. SSMs of the carpal bones and radius were created using 106 and 99 scans, respectively, with a maximum RMSE across all SSMs of 0.65 mm. The 95th percentile of the rotational deviations of the LCSs of the carpal bones and radius were below 3°. The resulting uncertainty due to anatomical variations in the calculation of the SLA and CLA was 2-3°. Anatomical shape variations hardly influence the SLA, CLA, and LCSs of the carpal bones and radius so the methods to estimate the LCSs can be used to estimate relevant kinematic parameters in 4DCT scans.