Pronation-Supination Standardization Using a Data-Driven Statistical Pose Model

Image-based clinical measurements of the forearm can be biased when pronation-supination (PS) is not properly accounted for when repositioning during surgery. We therefore aim to build and validate an original data-driven joint statistical pose model (SPM) combining both radius and ulna, that supports a linear angle-to-pose PS model, for computer-assisted PS standardization (CAPSS). We built an SPM from 88 forearm CT scans by registering ulnas to a common template and performing PCA on the coupled radius-ulna pose. We fitted the SPM to 25 PS sweeps acquired on optically tracked cadavers (4 specimens) and compared it with a fixed-axis rotation baseline, using translational error (TE), rotational error (RE), and mesh RMSE versus optical tracking. Spearman correlations between SPM modes and PS angle were used to identify PS-related modes, and further derive a linear angle-to-pose model evaluated by specimen-wise leave-one-out. The SPM reproduced cadaveric PS trajectories with 0.12 mm mean mesh RMSE, 0.12 mm TE, and 0.08 $_{}^{\circ }$ RE, versus 2.5 mm RMSE, 1.9 mm TE, and 5.3 $_{}^{\circ }$ RE for the fixed-axis baseline. The linear angle-to-pose model using only strongly PS-related modes reconstructed unseen sweeps with 1.9 mm RMSE, 1.6 mm TE, and 1.0 $_{}^{\circ }$ RE. Forearm PS is better captured as a combination of coupled radius-ulna pose modes than as a single fixed-axis rotation. In this preclinical validation on intact forearms, these modes supported a linear angle-to-pose model enabling automated pose standardization without subject-specific calibration. These findings support the feasibility of computer-assisted PS standardization for pose-consistent 3D morphometric analysis, while direct application to pathological anatomy and clinical surgical planning requires dedicated validation.