作者: Michael Brehler , Gaurav Thawait , Jonathan Kaplan , John Ramsay , Miho J. Tanaka
关键词: Set (abstract data type) 、 Intraclass correlation 、 Image registration 、 Medicine 、 Atlas (anatomy) 、 Metric (mathematics) 、 Algorithm 、 Reliability (statistics) 、 Mutual information 、 Image segmentation
摘要: We present an algorithm for automatic anatomical measurements in tomographic datasets of the knee. The uses a set atlases, each consisting knee image, surface segmentations bones, and locations landmarks required by metrics. A multistage volume-to-volume surface-to-volume registration is performed to transfer from atlases target volume. Manual segmentation volume not this approach. Metrics were computed transferred best-matching atlas member (different bone), identified based on mutual information criterion. Leave-one-out validation was 24 scans obtained using extremity cone-beam tomography. Intraclass correlation (ICC) between expert who generated above 0.95 all This compares favorably inter-reader ICC, which varied 0.19 0.95, depending metric. Absolute agreement with also good, median errors below 0.25 deg tibial slope static alignment, 0.2 mm tuberosity-trochlear groove distance medial depth. approach anticipated improve measurement workflow mitigate effects operator experience training reliability