1. Harvard Medical School Massachusetts General Hospital In-vivo Monitoring of Proton Therapy With PET: Biological Considerations Kira Grogg 1, Nathaniel Alpert 1, Xuping Zhu 1, Chul Hee Min 2, Mauro Testa 2, Brian Winey 2, Marc Normandin 1, Helen A. Shih 2, Harald Paganetti 2, Thomas Bortfeld 2, Georges El Fakrhi 1 IEEE-MIC 15 Nov, 2014 Seattle, WA 1 Center for Advanced Medical Imaging Sciences, Department of Nuclear Medicine and Molecular Imaging, MGH 2 Department of Radiation Oncology, Massachusetts General Hospital

2. Finite Proton Range: a blessing and a curse target Low entrance dose Sensitive to range uncertainty Localized dose 2 target Low entrance dose Sensitive to range uncertainty Localized dose Ideal treatment Potential actual treatment Dose to critical structure Tumor Critical structure Beam

3. PET Signal from nuclear interactions Positron emitters are produced from interactions of protons and nuclei Proton Neutron Positron Electron Photon 3 AtomreactionIsotopet 1/2 (min) 16 O p, pn 15 O 2.037 12 C p, pn 11 C 20.39 14 N p, 2p2n 11 C 20.39 16 O p, 3p3n 11 C 20.39 16 O p, 2p2n 13 N 9.965 14 N p, pn 13 N 9.965 31 P p, pn 30 P 2.498 40 Ca p, 2pn 38 K 7.636

4. 4 Proton beam from cyclotron sent to patient’s tumor Creates positron-emitting radioactive isotopes that can be imaged PET cameras detect the photons emitted Create images of radioactivity generated

5. Compare measured PET to simulated PET Monitoring Proton Therapy with PET Planned dose Predicted PET distributionMeasured static PET image 5 with washout applied Any discrepancies? Make improvements! Simulate Compare Correct washout using dynamic PET images instead

6. Biological effects on PET: rabbit head model Dead conditionLive condition Proton beam is into the screen Same color scale 6 Radioactivity is not static – biological clearance affects the distribution!

7. Biological correction to MC: Patient example Individual variation and unpredictable tumor perfusion –Not included in current washout correction Monte Carlo (MC) simulation with generic tissue-type based washout applied Static in-room PET measurement 7

8. Dynamic PET imaging Typical dynamic PET imaging –Inject single radiotracer, at one time point –Track over time the tracer distribution as it moves preferentially into specific tissues Dynamic PET imaging for proton therapy –Protons create multiple radioisotopes at one location Continuous production over 30-90 seconds Isotopes may also incorporate into multiple types of molecules –Track the loss/dispersion of radiotracers 8

9. Kinetic Modeling for biological washout Kinetic modeling to obtain original production rate maps of 15 O –Patient specific maps, independent of biological factors –Can be compared directly to MC simulations Biological clearance rate maps calculated concurrently –Potential biomarker for tumor treatment response Irradiation 9 C : PET activity concentration R 0 : Production rate during irradiation k r : radioactive decay constant k b : biological clearance rate t : time

10. Kinetic Modeling Assumptions 10 Necessitates scanning ASAP – 15 O: T 1/2 = 2 min At later times, 11 C and 13 N dominate –In-beam is ideal Expensive integration Limited coverage –Exploring in-room PET/CT scanning delay between treatment and imaging of ~3 minutes One k b 15 O H H H H 15 O dominates at early times

11. PET /CT Proton beam Mobile NeuroPET/CT  In-Room scanning In-room scanning for higher activity levels Integrated PET/CT scanner for improved attenuation correction and anatomical registration 11

12. Validation of kinetic model in rabbit thigh Kinetic model fits of time activity data for an ROI in the thigh, live and dead conditions. ROI in the thigh Beam direction Dead thigh Live thigh Live and dead rabbit thigh irradiated and scanned under identical conditions Ratio of production rates in live and dead conditions: 1.02 ± 0.03 12

13. Rabbit Thigh Production Rate Maps 13 R 1 : Rate of 15 O production –(a) Live R 1 map –(b) Dead R 1 map –(c) Profile through R 1 maps LiveDead

14. Rabbit #2: Beam through the thighs Dead conditionLive condition 14 R dead /R live = 1.02 ± 0.01 Thigh Muscle ROI

15. Rabbit #2: Beam through the thighs Dead conditionLive condition 15 Hip Bone ROI R dead /R live = 1.03 ± 0.03

16. First-in-man in-room NeuroPET/CT patient 16 Dose plan

17. PET/CT for 1 st patient MLEM, 100 iterations, no decay correction 12 minutes, 60 second frames AllEyeMaskBone 17

18. Patient 1: eye and mask ROIs Mask: no washout –k 1 = 0.0057 s -1 –R 1 = 73 Bq/ml/sec Eye: some washout –k 1 = 0.00612 s -1 –R 1 = 168 Bq/ml/sec 18 Results are consistent with MC and previous experiments 15 O decay constant: 0.00567 s -1

19. Patient 1: bone and tumor ROIs Bone: slow washout, less 15 O –k 1 = 0.00578 s -1 –R 1 =148 Bq/ml/sec Tumor: faster washout –k 1 = 0.00738 s -1 –R 1 = 216 Bq/ml/sec 19 Results are consistent with MC and previous experiments 15 O decay constant: 0.00567 s -1

20. Note the difference in loss of activity between the skull and the brain tissue Patient 2: PET Images and dynamic results ROIs in the brain and bone are plotted to the right Brain R 1 = 222 ± 19 Bq/ml/s k 1 = 0.0116 ± 0.0004 /s Bone R 1 = 199 ± 32 Bq/ml/s k 1 = 0.00652 ± 0.0009 /s 20 15 O decay constant: 0.00567 s -1

21. Conclusions/Future work Production rate maps are a better tool for comparisons of measured to simulated PET –3-D Maps of the produced radioactivity are created without confounding biological effects –Subsequent range determinations will be more accurate 21 More sophisticated reconstruction methods are being explored to improve upon the noisy images –e.g. Direct estimation of kinetic parameters Evaluate biological clearance rate (k 1 ) as a biomarker for treatment response

22. Acknowledgements Thanks to all the physicists, therapists, and engineers at the MGH Burr Proton Center Thanks to all of the members of the El Fakhri lab 22 http://nmmiweb.mgh.harvard.edu/CAMIS Funding –National Institutes of Health under Grant No. R21CA153455, R21EB12823 and T32RB013180