Knowledge-Based Intensity Modulated Radiotherapy (IMRT) Treatment Planning for Prostate Cancer
D Dick*, S Das, J Lo, Duke Univ Medical Center, Durham, NCWE-G-BRCD-6 Wednesday 4:30:00 PM - 6:00:00 PM Room: Ballroom CD
Purpose: To verify that a knowledge-based approach to intensity modulated radiotherapy (IMRT) treatment planning can create clinically acceptable plans of higher or comparable dosimetric quality than prior clinically approved plans.
Methods: Each case in a database of 140 IMRT prostate plans treated to 54 - 74Gy (28-41 fractions) is used as a query case and is compared to the other 139 cases in the database (match cases) using a case-similarity algorithm. 2D beam's eye view (BEV) projections of the query and match cases anatomies (planning target volume (PTV), rectum, bladder, right and left femoral heads) at treatment gantry angles are captured. To quantify similarity between the query and match cases, mutual information (MI) values of the query vs. match BEVs are averaged over all treatment gantry angles. The BEV PTV projections of the best match case, identified by the highest average MI value, are deformed to the query's PTV BEVs. The deformation maps are applied to deform the match case's fluences to suit the query case and further tweaked by running 50-100 iterations (without manual intervention) of the Eclipse optimization engine with constraints directly imported from the match case.
Results: Approximately 84.3% of the bladder and rectum dose constraints were achieved for the original plans whereas 86% were achieved for the post-optimized plans. On average, the rectum and bladder volume receiving 65Gy was smaller for the original plans than the post-optimized plans by (4.52±30.84) % and (3.44±33.55) % respectively. However, the rectum and bladder volume receiving 40Gy was smaller for the post optimized plans than the original plans by (1.53±24.66) % and (3.65±24.44) % respectively.
Conclusions: The knowledge-based approach produces treatment plans of greater or equivalent dosimetric quality to prior clinically approved plans. This work has the potential to semi-automatically provide high quality plans while dramatically reducing treatment planning time.