Room: AAPM ePoster Library
Purpose: To validate the efficacy of a localized Pearson’s Correlation Coefficient (PCC) to detect dosimetrically significant anatomic variations from CBCT imaging as a predictive tool for adaptive prostate radiotherapy.
Methods: A retrospective sample of 15 low-to-intermediate risk prostate cancer patients treated with a standard fractionation regimen and daily cone-beam CT imaging was selected. An automated scripting interface in the treatment planning system (Raystation 8A) was used to perform CBCT dose calculations. The clinical target volume (CTV), planning target volume (PTV), and the rectum were outlined on each CBCT. PCC metrics were computed within the PTV between the first CBCT and all subsequent CBCT images to detect voxel density changes. Observed changes in PCC were compared to dosimetric trends to determine predictive thresholds for significant dose deviations: >5% deviation for target D90 or D95, >15% increase in rectum V70.
Results: The automatic dose calculation and PCC tracking framework was analyzed for approximately 540 CBCT datasets. The PCC varied greatly between patients and CBCT imaging systems (0.4–0.9). Target dose deviations were negligible (±3% for CTV and PTV), but rectal dose deviations were found to correlate well with PCC deviations (>15% increase in V70 for >20% change in PCC).
Conclusion: Since PCC is image noise dependent and varies with patient size, imaging hardware, and software parameters, relative PCC deviations were more reflective of anatomic variations. Minimal deviations in target coverage were likely due to generous planning margins (8–10 mm) and the robustness of photon dosimetry to small geometric and density variations. However, the presence of steep dose gradients makes rectal dose more sensitive to geometric changes that were clearly reflected in PCC variations. Therefore, PCC can be used effectively to guide online image matching as well as for adaptive planning.