Comparison of inverse planning simulated annealing and geometrical optimization for prostate high-dose-rate brachytherapy.

PURPOSE: An inverse planning simulated annealing (IPSA) algorithm for optimization of high-dose-rate (HDR) brachytherapy has been previously described. In this study, IPSA is compared with geometrical optimization (GO) for prostate brachytherapy.

METHODS AND MATERIALS: Using CT data collected from 10 patients, treatment plans were prepared using GO and IPSA. The clinical target volume (CTV) and critical organs (CO) including bladder, rectum, and urethra were contoured using Plato Version 14.2.1 (Nucletron Corp., Veenendaal, The Netherlands). Implant catheters were digitized using the CT planning system. All dwell positions outside of the CTV were turned off. Two optimized plans were generated for each implant using GO and IPSA. The same set of dose constraints were used for all inverse planning calculations and no manual adjustment of the dwell weight was used. Two prescription methods were used. Using the first method, coverage was prioritized: the prescription dose was normalized to the isodose volume that covers 98% of the CTV (V100 = 98% of CTV). The dose volume histograms (DVH) of CO were generated for comparison. Using the second method, sparing was prioritized: the prescription dose was normalized such that no urethra volume received 150% of the prescription dose (V150-urethra = 0 cc). The DVH of CTV and CO were generated, and the homogeneity index (HI) and conformal index (COIN) were calculated for comparison and compared using the Wilcoxon matched-pairs test.

RESULTS: Using the coverage-prioritized method, the difference in V80-bladder dose was not statistically significant (p = 0.09; median: IPSA = 0.62 cc, GO = 1.05 cc). The V80-rectum ranged from 0.20-4.8 cc, and 0.05-1.4 cc using GO and IPSA, respectively. IPSA's V80-rectum was significantly lower (p = 0.005; median: IPSA=0.38 cc, GO = 1.31 cc). V150-urethra ranged from 0.02-0.75 cc and 0.0-0.01 cc using GO and IPSA, respectively. The V150-urethra was significantly lower using IPSA (p = 0.005; median: IPSA = 0.00 cc, GO = 0.33 cc). Using the sparing prioritized method, the V100-prostate ranged from 30-97% and 95-100% using GO and IPSA, respectively. This difference was statistically significant (p = 0.008). The HI and COIN were statistically higher using IPSA (p = 0.005).

CONCLUSION: Anatomy-based inverse optimization using IPSA is superior to dwell-position-based optimization using GO as it: (1) Improves target coverage and conformality while sparing normal structures, (2) Improves dose homogeneity within the target, and (3) Minimizes volume of non-contoured normal tissue irradiated. Routine application of three-dimensional brachytherapy planning and anatomy-based inverse dwell time optimization is recommended.