Automated treatment planning engine for prostate seed implant brachytherapy

To develop a computer-intelligent planning engine for automated treatment planning and optimization of ultrasound- and template-guided prostate seed implants.

The genetic algorithm was modified to reflect the 2D nature of the implantation template. A multi-objective decision scheme was used to rank competing solutions, taking into account dose uniformity and conformity to the planning target volume (PTV), dose-sparing of the urethra and the rectum, and the sensitivity of the resulting dosimetry to seed misplacement. Optimized treatment plans were evaluated using selected dosimetric quantifiers, dose-volume histogram (DVH), and sensitivity analysis based on simulated seed placement errors. These dosimetric planning components were integrated into the Prostate Implant Planning Engine for Radiotherapy (PIPER).

PIPER has been used to produce a variety of plans for prostate seed implants. In general, maximization of the minimum peripheral dose (mPD) for given implanted total source strength tended to produce peripherally weighted seed patterns. Minimization of the urethral dose further reduced the loading in the central region of the PTV. Isodose conformity to the PTV was achieved when the set of objectives did not reflect seed positioning uncertainties; the corresponding optimal plan generally required fewer seeds and higher source strength per seed compared to the manual planning experience. When seed placement uncertainties were introduced into the set of treatment planning objectives, the optimal plan tended to reach a compromise between the preplanned outcome and the likelihood of retaining the preferred outcome after implantation. The reduction in the volatility of such seed configurations optimized under uncertainty was verified by sensitivity studies.

An automated treatment planning engine incorporating real-time sensitivity analysis was found to be a useful tool in dosimetric planning for prostate brachytherapy.