Encrypted login | home

Program Information

Fraction-Variant Beam Orientation Optimization for IMRT Based On Group Sparsity

no image available
D O'Connor

D O'Connor*, D Nguyen , D Ruan , V Yu , K Sheng , UCLA School of Medicine, Los Angeles, CA


SU-K-FS2-11 (Sunday, July 30, 2017) 4:00 PM - 6:00 PM Room: Four Seasons 2

Purpose: Conventional beam orientation optimization (BOO) algorithms for IMRT assume that the same beam angles will be used for each treatment fraction. This seems to be an unnecessary constraint that may compromise plan quality, particularly for complex non-coplanar IMRT plans that benefit from using more beam orientations. This study investigates the potential gain from allowing beam orientations to vary between fractions. We present a BOO formulation based on group sparsity that simultaneously optimizes non-coplanar beam angles for all fractions, yielding a fraction-variant treatment plan.

Methods: Beam angles for all fractions are selected by solving a multi-fraction fluence map optimization problem involving 500-700 candidate beams per fraction, with an additional group sparsity term that encourages most candidate beams to be inactive. The optimization problem is solved using an accelerated proximal gradient method (FISTA). Our fraction-variant BOO algorithm is used to create non-coplanar, five-fraction treatment plans for a prostate case and a lung case. The treatment plans are compared with 4pi treatment plans created using an established column generation algorithm.

Results: Mean OAR dose was reduced by 12.2% of the PTV prescription dose (rectum), 4.8% (bladder), 21.9% (seminal vesicle), 22.5% (penile bulb), and 1.8% (proximal bronchus). On average, mean and max OAR dose were reduced by 5.8% and 3.4% of the prescription dose, respectively. PTV homogeneity defined as D95/D5 improved from .938 to .982 (prostate case) and from .958 to .980 (lung case). Moreover, the fraction-variant prostate and lung plans used only 9.6 and 9 beams per fraction on average, respectively, whereas the column generation plans used 20 beams. The average FISTA runtime was 77.5 minutes.

Conclusion: This work demonstrates the first BOO algorithm that simultaneously optimizes all beam angles for all fractions. The resulting plans are dosimetrically superior to the column generation plans while using half as many beams per fraction.

Funding Support, Disclosures, and Conflict of Interest: Funding sources: DOE DE-SC0017057 NIH R44CA183390 NIH R01CA188300 NIH R43CA183390 NIH U19AI067769

Contact Email: