3-D Ultrasonographic Quantification of Hand and Calf Muscle Volume: Statistical Shape Modeling Approach

Accurate acquisition and segmentation of muscles are essential in 3-D freehand ultrasonography (US) to estimate in vivo muscle volume, but the source of segmentation inaccuracy in shape variation has never been the focus. This study was aimed at investigating reliability of 3-D US in the acquisition and segmentation for muscle volume of two muscles of different sizes and in identifying a primary source of measurement difference. The lateral gastrocnemius and flexor pollicis brevis of 12 healthy adults were assessed using freehand 3-D US scans. The motion-tracking data of the probe were synchronized with the B-mode ultrasound scan to reconstruct 3-D muscle volume. Statistical shape modeling was used to provide a spatial segmentation volume difference that further explains the variation around segmentation repeatability. The absolute difference of the flexor pollicis brevis was 3.5 percentage points greater than that for the lateral gastrocnemius. The highest measurement differences were observed when for inter-acquirer analysis. Statistical shape modeling revealed that the primary segmentation volume differences were at the muscle ends and edges, where the muscle interfaces with the surrounding muscles. Three-dimensional US is a reliable tool in the clinical setting, but care must be taken to ensure that acquisition and segmentation are consistent, particularly in a small muscle that interfaces with tendons and other soft tissues.