Encrypted login | home

Program Information

OPTIMIZER, a New Optimization Algorithm for a Co-60 Arc Therapy Machine


O Chibani

O Chibani, C Ma, Fox Chase Cancer Center, Philadelphia, PA, (2) Fox Chase Cancer Center, Philadelphia, PA

Presentations

SU-I-GPD-T-406 (Sunday, July 30, 2017) 3:00 PM - 6:00 PM Room: Exhibit Hall


Purpose: To present a new optimization algorithm (OPTIMIZER) for arc therapy using a new Co-60 machine. Results for different anatomic sites (prostate and Lung) are presented and compared to a commercial TPS.

Methods: The Co-60 machine has a 70-cm SAD (Source to Axis Distance) and 0.5-cm-witdh MLC system. The maximum MLC speed is assumed to be 2.5 cm/s at isocenter. Dose kernels for varying SSDs (from 40 to 100 cm) are generated for 0.5x0.5-cm2 beamlets using the GEPTS Monte Carlo system. Heterogeneity correction is used. OPTIMIZER uses independent (i) random gradient descent method and (ii) local minima escape strategy. The optimization process takes into account the machine physical constraint (available activity, maximum gantry rotation speed, and maximum MLC speed) to make sure the plan is deliverable with the desired number of arcs.

Results: OPTIMIZER is first used to reproduce a VMAT prostate plan using a 6 MV linac beam. Dose kernels similar to those generated for the Co-60 machine were generated for the 6 MV beam. The same set of DVH objectives for PTV80, PTV56-80, Rectum-PTVs, and Bladder-PTVs were used in the two systems. OPTIMIZER produces a plan of similar quality using the same number of arcs (2). See Table I. The OPTIMIZER plan requires about 10 mn execution time. Figure I shows OPTIMIZER prostate plans obtained with a 6 MV 100-cm SAD linac and the Co-60 70-cm SAD machine. Finally, the new optimizer was used for a lung case treated with a single partial Co-60 arc (Fig.2). A highly conformal plan is obtained (CI = 1.01).

Conclusion: A new optimization algorithm is presented. Although designed for a Co-60 arc therapy machine, the new algorithm can be used for linacs as well. Comparison with Eclipse VMAT shows similar plan quality in relatively short computing times (~ 10 mn).


Contact Email: