Encrypted login | home

Program Information

Improving Non-MCO VMAT Planning in RayStation for Prostate Treatment with An Automated Knowledge-Based Model Created Using a Database of Clinical MCO-IMRT Plans

no image available
M Young

M. Young*, L Vanbenthuysen , B Crawford , Y Wang , Massachusetts General Hospital and Harvard Medical School, Boston, MA


SA-B-BRD-7 (Saturday, March 5, 2016) 10:30 AM - 12:30 PM Room: Grand Ballroom D

To overcome the computational challenges of multi-criteria optimization (MCO) for volumetric-modulated arc therapy (VMAT) in RayStation using a knowledge-based model constructed with sixty MCO-based intensity-modulated radiation therapy (IMRT) prostate plans.

The same prescription was used for the IMRT and VMAT plans – 50.4 Gy to PTV5040 (expanded from prostate + proximal seminal vesicles) followed by a boost of 28.8 Gy to PTV7920 (expanded from prostate). The dose-volume histogram (DVH) for each target and organ-at-risk (OAR) was averaged over the MCO-IMRT plans. The average DVH for each structure was used to calculate model constraints (upper range) and objectives (lower range) at select dose-volume points. The model was implemented to guide non-MCO VMAT optimization for ten randomly-selected clinical VMAT prostate plans. Each plan previously required an expert planner to initially generate a patient-specific MCO-IMRT plan to inform non-MCO VMAT plan optimization. Model-generated plan quality was compared to that of the planner-generated plan using homogeneity index (HI) and conformity number (CI) for PTV5040, PTV7920, CTV5040, and CTV7920, and multiple DVH indices for bladder, rectum, anterior rectal half, posterior rectal half, femurs, and penile bulb. Statistical significance was assessed using the Wilcoxon rank sum test.

All model-generated plans achieved prescribed target coverage and satisfied OAR sparing per QUANTEC protocol. The model-generated plans showed statistically significant improvement on target coverage, with HI decreased by 27.9%, 15.0%, and 21.2%, for PTV5040, PTV7920, and CTV7920, respectively. The model-generated plans also showed statistically significant sparing improvements on the bladder, rectum, and femurs. The most improved OAR sparing was observed for Dmean and D2 of the bladder, which were reduced by 2.0±2.6 and 5.7±0.7 Gy, respectively.

The automated, knowledge-based model used an MCO-IMRT plan database to substantially improve VMAT prostate planning plan quality over the manual VMAT planning technique in RayStation.

Contact Email: