The quantitative assessment of muscular motion is a fundamental evaluation for rehabilitation biomechanics. We propose two multi-feature block-matching-based (MFBM) tracking methodologies to estimate muscular motion for 3D (2D+t) and 4D (3D+t) ultrasonic image sequences respectively. The proposed MFBM based on image intensity, motion velocity, neighbor relativity and selected features with multilevel framework achieves accuracy and spatial adaptability. To quantify motion of borderless musculature from 4D image sequences, the end of fascia is chosen as target and tracked by multilevel MFBM with drift correction. The tolerance for tissue deformation at low frame rate is effectively increased in the 4D estimation. Although 4D motion evaluation can alleviate the out-of-plane problem in 3D tracking, 3D motion analysis is still commonly used in clinics as the 4D equipments are expensive and unpopular. In this paper, by referring to the most discriminative neighborhood, a multilevel MFBM with Kalman prediction is proposed to resolve the ambiguities caused by region homogeneity or insufficient local texture in 3D motion tracking. The accuracy in both methods is validated from in vivo musculoskeletal ultrasonic image sequences by comparing the results with the doctor-defined ground truth. This study can be applied to clinical diagnosis, such as sport injuries.
關聯:
The 16th International Conference on Mechanics in Medicine And Biology，起迄日：2008/7/23~2008/7/25，地點：Pittsburgh
