Multiple anatomical landmark calibration for optimal bone pose estimation

Abstract Bone motion estimation by means of photogrammetric, non-invasive methods can be severely corrupted by experimental errors. The largest fraction of such errors is associated with the relative movement between externally located markers and the underlying bone, due to the interposition of both passive and active soft tissues. The errors affecting the estimates of anatomical landmarks trajectories in the laboratory frame can be considerably reduced by following the Calibrated Anatomical System Technique protocol which entails: (i) a static calibration of the anatomical landmarks in a technical reference frame defined by the cluster of skin markers, and (ii) the use of of a rigid model of the cluster. This paper illustrates how a modification of the above-mentioned protocol, involving a multiple calibration of the anatomical landmarks in different postures, and the use of a deformable model of the cluster, can effectively enhance bone motion estimation. In order to validate the new protocol a cycling test on a patient wearing a femoral external fixator was performed. The root mean square reconstruction error (RMSE) on an anatomical landmark (greater trochanter) trajectory drops from over 15 mm to less than 10 mm while the RMSEs of the bone (femur) orientation and position decrease respectively from about 5 deg and 7 mm with our previous protocol to less than 4 deg and 4.5 mm. The improvements are even more significant when movement components relative to the main planes and axes of motion are considered.