作者: Cheng-Chung Lin , Shuo Zhang , Jens Frahm , Tung-Wu Lu , Chao-Yu Hsu
DOI: 10.1118/1.4820369
关键词: Computer science 、 Knee Joint 、 Artificial intelligence 、 Medical imaging 、 Kinematics 、 Fluoroscopy 、 Computer vision 、 Tibia 、 Knee kinematics 、 Biomechanics 、 Image registration 、 General Medicine
摘要: Purpose: This study developed and assessed a slice-to-volume registration method that integrated three-dimensional (3D) static MRI volumes of the bones with novel single-slice, real-time radial fast low-angle shot for measuring 3D kinematics knee. Methods: Multislice images (for establishing bone models) 2D knee at five positions, during flexion/extension were acquired from three healthy adults. The poses, thus knee, obtained by registering to reformed slice interpolated models according theWEMS similarity measure. ensemble means (biases) standard deviations (precisions) measurement errors proposed method, i.e., differences between registered calculated across all trials subjects. Ensemble repeated registrations dynamic data subjects indicate repeatability method. Results: (standard deviations) femoral poses less than 0.6 (0.6) mm translations −0.2° (1.3°) degrees rotations. corresponding values tibia 0.5 (0.7) −0.4° (1.1°), respectively. joint 0.9 (1.4) −0.3° (1.8°) For tests, 1.2 1.5° Conclusions: With accuracy achieved, without use ionizing radiation multiple repetitive motions, combining MR imaging promises be valuable tool studying noninvasively.