Abstract
The aim of this study was to develop a deformable image registration (DIR)-based offline ART protocol capable of identifying significant dosimetric changes in the first treatment fractions to determine when adaptive replanning is needed. A total of 240 images (24 planning CT (pCT) and 216 kilovoltage cone-beam CT (CBCT)) were prospectively acquired from 24 patients with prostate adenocarcinoma during the first three weeks of their treatment (76 Gy in 38 fractions). This set of images was used to plan a hypofractionated virtual treatment (57.3 Gy in 15 fractions); correlation with the DIR of pCT and each CBCT allowed to translate planned doses to each CBCT, and finally mapped back to the pCT to compare with those actually administered. In 37.5% of patients, doses administered in 50% of the rectum (D50) would have exceeded the dose limitation to 50% of the rectum (R50). We first observed a significant variation of the planned rectal volume in the CBCTs of fractions 1, 3, and 5. Then, we found a significant relationship between the D50 accumulated in fractions 1, 3, and 5 and the lack of compliance with the R50. Finally, we found that a D50 variation rate [100 × (administered D50 − planned D50/planned D50)] > 1% in fraction three can reliably identify variations in administered doses that will lead to exceeding rectal dose constraint.