Encrypted login | home

Program Information

Effect of Grid Size On Optimization Using the MultiPlan Monte Carlo Algorithm

no image available
R Krauss

R Krauss*, J Tubbs , St. Francis Hospital, Memphis, TN


PO-BPC-Exhibit Hall-14 (Saturday, March 5, 2016)  Room: Exhibit Hall

Purpose:To determine the effect of optimization grid size on the full resolution calculation of lung plans using MultiPlan’s fast Monte Carlo algorithm by comparing visual prescription lines and new conformity index.

Methods:Three lung lesions were contoured with a 5 mm PTV expansion in the MultiPlan treatment planning system. A set of beams were generated for each PTV using the Ray-Tracing algorithm and the plan saved. Using the saved beam set, a single optimization was performed for each PTV using MultiPlan’s Monte Carlo algorithm with identical constraints on low (64x64), medium (128x128), and high (512x512 “native”) resolution. No further optimization or time or beam reduction was done to ensure each resolution was optimized under the same conditions. Each plan was prescribed to 54 Gy. The conformity was then visualized in three planes and calculated using Paddick’s new conformity index. All optimized plans were then calculated on high resolution and re-prescribed as above. The conformity was again visualized in the three planes and conformity calculated.

Results:Comparison of the conformity of each low, medium, and high resolution optimization to the conformity of the corresponding high resolution calculation showed significant improvement in agreement between the optimization and final calculation as the resolution increased. Each optimization produced plans with high conformity (nCI=1.1-1.2) but at lower resolutions, the high resolution calculation deviated significantly from the original optimization (low: nCI=1.4-1.6, medium: nCI=1.2-1.4). High resolution optimizations were accurately reflected in the final high resolution calculation, as expected (nCI=1.1-1.2).

Conclusion:This work indicates that optimizations using lower resolutions of MultiPlan’s fast Monte Carlo algorithm do not accurately reflect the dose distribution calculated at the highest resolution. While high resolution optimizations can take two to three times longer than at lower resolutions, the conformity of the final plan may offset the additional optimization time.

Contact Email: