Flow chart of the multiresolution Demons algorithm to register planning CT images with respect to weekly CBCT images.
Illustration of histogram matching in which the intensity histogram of one image set is adjusted to match the intensity histogram of the other image set.
Results of histogram matching: (a) CBCT image slice, (b) corresponding planning CT image slice before histogram matching, and (c) planning CT image slice after histogram matching.
Checkerboard display of CBCT and CT images before (left) and after (right) Demons registration. After registration, the two images match each other well.
Anatomic targets used for TRE calculations (left column: Planning CT; middle column: CBCT; right column: Registered CT). The target points are shown as enlarged dots for illustration. Top row: Bony tissue targets. Bottom row: Soft tissue targets. tip of coronoid mandible process, tip of coronoid mandible process, external auditory canal, and external auditory canal.
Registration accuracy as measured by the mean TRE value averaged over the 9 anatomical points in weekly CBCT images derived from 12 head-and-neck cancer patients. The TRE averages demonstrate registration accuracy near the level of voxel size
TRE sorted by anatomical point for the 12 cases considered. Box-and-whisker plots represent the median TRE (horizontal line), the 25th and 75th percentiles (edges of the box), and total range (extent of whiskers) excluding outliers ( symbols), with outliers defined as greater than 1.5× interquartile range below the first or above the third quartile. The TRE averages demonstrate registration accuracy near the level of voxel size.
Volume overlap index for eight head-and-neck cases.
Article metrics loading...
Full text loading...