MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Stochastic Initialization of MLC Sequences

C Locke*, K Bush, Stanford University, Stanford, CA

Presentations

(Sunday, 7/29/2018) 3:00 PM - 6:00 PM

Room: Exhibit Hall

Purpose: The problem of trying to fit a fluence map with a minimal number of apertures is NP-hard, and so typically heuristics are used to determine an approximately minimal set of apertures that produce the desired fluence map.

Methods: This work introduces a novel approach to approximating fluence maps with apertures by using leaf pair probability density functions that are used to stochastically sample apertures in a way the automatically approximates the fluence map. Using these randomly sampled apertures as a starting point for gradient descent optimization does very well at reproducing the original fluence maps. An example of the resulting fluence for 30 aperture and 10 aperture fits are shown in Figure 1.

Results: When applied to a 16 field breast treatment, the calculated 30 apertures per beam step-and-shoot plan was dosimetrically indistinguishable from the original 16 field sliding window plan shown in Figure 2, with a 65% of the MUs.

Conclusion: This work introduces a new approach to fluence map fitting using stochastically sampled MLC sequences and gradient descent. The technique is able to determine MLC sequences that are dosimetrically equivalent to the original plans using less MUs. Future work will look at applying this technique to dynamic gantry plans to improve VMAT's plan quality through better choices of initial coarsely sampled apertures.

Funding Support, Disclosures, and Conflict of Interest: This work was supported in part by Varian Medical Systems.

Keywords

Optimization

Taxonomy

TH- External beam- photons: IMRT dose optimization algorithms

Contact Email