Encrypted login | home

Program Information

Knowledge Based Automatic Lung IMRT Planning with Non-Coplanar Beams

no image available
L Yuan

L Yuan1*, Y Ge2 , Y Sheng3 , K Hedrick4 , F Yin5 , Q J Wu6 , (1) Duke University Medical Center, Durham, NC, (2) University of North Carolina at Charlotte, Charlotte, NC, (3) Duke University, Durham, NC, (4) Duke University, Durham, NC, (5) Duke University Medical Center, Durham, NC, (6) Duke University Medical Center, Durham, NC


TH-EF-BRD-7 (Thursday, July 16, 2015) 1:00 PM - 2:50 PM Room: Ballroom D

Purpose: We present a lung IMRT planning method which automatically determines both plan optimization objectives and beam configurations with non-coplanar beams by analyzing patient-specific anatomical information and learning from past planning experiences.

Methods: A beam navigation index map is constructed to characterize the geometry of the tumor relative to the OARs at each candidate beam direction. It is defined as the ratio of the dose deposited inside the OARs to that in the PTV along the beam path. An additional term which is a function of the beam spread is added to the navigation index to account for the effect of multiple beams on PTV dose conformity. For each case, a fixed number of beams which minimizes the total index values are selected. The relative weights of different OARs and the dose conformity term in the index are learned from the beam configurations in a database of mostly co-planar beam plans. The OAR dose sparing objectives are generated by OAR DVH prediction models which are trained by prior IMRT plans. This completely automated method was validated by re-planning 10 clinical lung plans which have used non-coplanar beams. Important dosimetric parameters of the automatically generated plans are compared with those of the clinical plans by Wilcoxon signed rank tests.

Results: All dosimetric parameters in the automatically generated plans are statistically better or comparable with those in clinical plans. The most significant improvements are: lung V10Gy with average reduction of 6% of OAR volume, Esophagus V20Gy with average reduction of 3% and spinal cord D2% with reduction of 19% of prescription dose on average.

Conclusion: Knowledge models learned from mostly co-planar beam plans can be utilized to generate IMRT plans with non-coplanar beams. This method demonstrates the potential to improve both the planning efficiency and plan quality by automatically incorporating non-coplanar beams.

Funding Support, Disclosures, and Conflict of Interest: Partially supported by NIH/NCI under grant #R21CA161389 and a master research grant from Varian Medical Systems.

Contact Email: