Purpose: The inter‐fraction organ motion/deformation can be conveniently modeled using Bayesian theory with Normal‐gamma conjugate prior if signed distance from any fixed point in space to surface of the organ of interest obeys normal distribution. In this study, we investigated whether the inter‐fraction motion/deformation of bladder and rectum observed from clinical prostatectomy patients satisfy this normality condition. Method and Materials: 285 treatment planning CT and daily CT‐on‐rails scans from 7 prostatectomy patients were used in this study. Both bladder and rectum were contoured on all scans. Each patient's daily CT‐on‐rails scans were registered to his treatment planning CT and the bladder/rectum contours were mapped into treatment planning CT space for analysis. A cubic box with orientations along treatment planning CT image axes is defined to contain all bladders/rectums with 2cm margin. For each voxel inside this box (size: 2mm×2mm×1.5mm), its distance to the bladder/rectum surfaces was measured. Sign is added to the distance to indicate whether a point is inside or outside of an organ of interest. Now the inter‐fraction motion/deformation of bladder/rectum can be characterized by the distance variation from the voxels to the bladder/rectum surface. Jarque‐Bera normality statistical test was employed to examine whether the signed distances obey normal distribution. Results: For each patient, the signed distance to bladder or rectum from at least 99.99% of the voxels passed the Jarque‐Bera test with p‐value 0.05. Conclusions: For prostatectomy patients, their bladder or rectum inter‐fraction organ motion/deformation can be statistically described by a Gaussian signed distance field. This makes it possible to use Bayesian statistics model with Normal‐gamma conjugate prior to predict bladder or rectum daily location and shape during a prostatectomy patient fractionated radiotherapy.
ASJC Scopus subject areas
- Radiology Nuclear Medicine and imaging