Incorporating Prior Knowledge into IMRT Beam Orientation Optimization

Transcription

1 Incorporating Prior Knowledge into IMRT Beam Orientation Optimization Lei Xing, Ph.D. and Andrei Pugachev, M.S. Department of Radiation Oncology Stanford University School of Medicine Stanford, CA

2 Fixed gantry IMRT 140 o 180 o 220 o 100 o 260 o Center 60 o 300 o 20 o 9-field head and neck Treatment 340 o

3 Things need to be optimized Number and incident directions of beams Beam modality Beamlet weights

4 Beam orientation optimization Clinically, beam orientations are selected empirically. Excessive computational time is required for full beam orientation optimization. F n N = = n= 1 r σ [D c (n) D 0 (n)] 2 D c (n) --- function of beamlet weights, beam orientation.

5

6 Beam orientation optimization More intelligent and less computationally intensive methods of beam orientation selection are needed.

7 Beam s-eye-view Dosemetrics (BEVD) for IMRT Beam Selection BEV volumetrics in conventional radiotherapy. Only prior geometric information of the system is used. BEV Volumetrics Gantry Angle

8 BEVD We extend the original BEV to IMRT with consideration of 1. prior geometric and dosimetric information of the system, and 2. beam intensity modulation.

9 Beam s-eye-view Dosemetrics (BEVD) for IMRT Beam Selection BEVD for IMRT. Maximum beam profile A Pugachev & L Xing, IJROPB 51, (2001). oth prior geometric & dosimetric knowledge of the system are used.

10 BEVD For each organ at risk (OAR), assign a tolerance dose. For every possible beam orientation, calculate single beam dose distribution that can be achieved without exceeding the tolerances of OARs and normal tissue.

11 BEVD Calculation Divide a beam into bixel map. m 1 1 c Introduce an empirical score function to evaluate the goodness of each beamlet/beam. Obtain BEVD score for each incident beam direction. Select peaks of the BEVD score function, taking into account the principle of maximum beam separation. Perform beam intensity profile optimization.

12 BEVD ranking for an IMRT treatment of paraspinal tumor (Coplanar beam) Selected beams Five beams with high BEVD score. Selected gantry angles: 225, 175, 140, 85, 260

13 BEVD ranking for an IMRT treatment of paraspinal tumor (Non-coplanar beam) Selected beams Selected beams (gantry, table): (180,20) (180,-20) (230,20) (230,-20) (130,0)

14 0 PBEV Selected Beam Orientations Equiangular Beam Orientations Liver % 50% 65% 80% 95% GTV Kidney 90 Beams (gantry,table): (180,20), (180,-20) (230,20), (230,-20) (10,0) 180 Gantry angles:40,110,180,255, 325

15

16 Beam orientation selection Selected beams 225 (15MV), 180 (6MV), 140 (6MV), 85 (15MV), 260 (15MV)

17 Results Nasopharyngeal cancer BEVD method applied to the selection of coplanar beams

18 Dose distribution of two-stage optimization 0 35% 50% 65% 80% 95% 270 GTV 90 Cord 180

19 (a) (b) Fig. 2

20

21 Results III Model cases 2D. Simplified dose model. beam orientations are selected using BEVD method.

22 0 0 OAR 2 OAR 2 PTV OAR 3 90 OAR 3 PTV 90 OAR 1 OAR 1 Dose distribution of five IMRT beams selected with BEVD 0.1 Tolerance:

23 Volume, % Equiangular BEVD selected OAR 3 PTV Dose

24 0 0 PTV PTV OAR 90 OAR 90 Dose distribution of five IMRT beams selected with BEVD Equally spaced five beams

25 S i Gantry angle

26 Volume, % Equiangular BEVD-selected PTV OAR Dose

27 BEVD-guided beam optimization for IMRT In IMRT, the quality of a beam direction can be measured by a BEVD score. Select beams with the highest BEVD score, taking into account the angular separation of the beams. BEVD is a computer assisting tool, not an automated tool. It can, however, be used as a priori knowledge of the system to improve the beam orientation calculation.

28 (a) (b)

29 BEVD-Guided Beam Orientation Optimization A. Simulated annealing (SA) optimization B. BEVD-guided simulated annealing optimization Initial set of beam orientations Initial set of beam orientations Vary a gantry angle randomly Reduce temperature Vary a gantry angle randomly Reduce temperature Beam intensity profile optimization Calculate probability of acceptance based on BEVD score Accept or reject the trial beam configuration based on SA probability No Temperature low enough? Accept or reject trial beam configuration Accept Reject Yes Beam intensity profile optimization Optimized plan Accept or reject the trial beam configuration based on SA probability No Temperature low enough? Yes Optimized plan

30 Liver GTV Spinal cord 180 Kidney

31 BEVD-Guided Beam Orientation Optimization T=0 With annealing Number of iterations is reduced by a factor of 10. BEVD-guided optimization improves convergence behavior of the system. Speed depends weakly on the cooling schedule. Results are consistent with BEVD ranking. Formalism is applicable for functional imaging-guided IMRT.

32 Conclusions A practical BEVD tool has been developed for beam orientation selection in IMRT. The technique allows one to select beam orientations without paying the excessive computing overhead of computer beam orientation optimization. The approach has considerable potential for simplifying the IMRT planning process to maximize the potential of IMRT.

33 Conclusions Incorporation of prior knowledge significantly improves the calculation: speed and convergence. The methodology also applies to the optimization of other system parameters, like beamlet weights.

34 ACKNOWLEDGEMENTS S. Crooks, C. Cotrutz, S. Hunjan, J. Lian, D.Y. Yang, A Boyer, Q.T. Le, S.S. Donaldson, S. Hancock, C. King, A. Koong This work was partially supported by grants from the American Cancer Society, Whitaker Foundation and Department of Defense.