Encrypted login | home

Program Information

Impact of the History Size M Used in Limited-Memory Broyden-Fletcher-Goldfarb-Shanno Optimization Algorithm On Intensity-Modulated Proton Therapy Treatment Planning

no image available
I Khan

I Khan1*, W Liu2 , (1) Arizona State University, Tempe, AZ (2) Mayo Clinic Arizona, Phoenix, AZ


SU-I-GPD-T-159 (Sunday, July 30, 2017) 3:00 PM - 6:00 PM Room: Exhibit Hall

Purpose: In the process of constructing intensity-modulated proton therapy (IMPT) treatment plans, the Limited-memory Broyden-Fletcher-Goldfarb-Shannon (L-BFGS) optimization algorithm has been increasingly gaining attention for the treatment of cancer. Originating from the family of quasi-Newton methods, it takes into the account the limited amount of memory in a computer. The L-BFGS retains the past m updates of the solution variable as well as the gradient of the objective function. Thus, the question arises as to whether varying the history size m makes an impact on the quality of the IMPT treatment plan.

Methods: IMPT plans were generated for a prostate, head/neck, and lung patient. Six plans were generated for each case using L-BFGS by varying the history size m for each distribution. The following history sizes were used for each case: 1, 2, 3, 5, 100, and 1000. Dose-value histograms (DVH) were generated for each plan in order to examine the quality of the plan. A number of metrics were examined for each plan, including D_95% as well as D_5% - D_95% for the target tumor, and the D_1% for each organ at risk (OAR). Vā‚‚ā‚€ was examined for lung cases. These benchmarks were utilized to compare the tumor dose coverage, the tumor dose homogeneity, as well as the OAR protection in between the different plans that were generated.

Results: The L-BFGS optimization can yield clinically acceptable plans within 60 minutes for all cases on a laptop. When the history size m is changed, the D_1% for all OARs and the tumor dose homogeneity (D_5% - D_95%) appeared to stagnate for each plan in each patient.

Conclusion: It appears that modifying the history size m used in L-BFGS has no impact on the IMPT treatment plan quality.

Funding Support, Disclosures, and Conflict of Interest: Supported by the National Cancer Institute (NCI) Career Developmental Award K25-CA168984, by the Fraternal Order of Eagles Cancer Research Fund Career Development Award, by The Lawrence W. and Marilyn W. Matteson Fund for Cancer Research, by Mayo Arizona State University Seed Grant, and by The Kemper Marley Foundation.

Contact Email: