clinical experience of monte carlo dose calculation in

22
Current and future of independent dose calculation using Monte Carlo in Nagoya Proton Therapy Center Toshiyuki Toshito, Chihiro Omachi Nagoya Proton Therapy Center, Nagoya City University Tsukasa Aso National Institute of Technology, Toyama College Shogo Okada, Koichi Murakami, Takashi Sasaki High Energy Accelerator Research Organization International Conference on Medical Physics and School Jul 31, 2019@ICISE, Quy Nhon, Vietnam

Upload: others

Post on 01-May-2022

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Clinical experience of Monte Carlo dose calculation in

Current and future of independent dose calculation using Monte Carlo in Nagoya

Proton Therapy Center

Toshiyuki Toshito, Chihiro OmachiNagoya Proton Therapy Center, Nagoya City University

Tsukasa AsoNational Institute of Technology, Toyama College

Shogo Okada, Koichi Murakami, Takashi SasakiHigh Energy Accelerator Research Organization

International Conference on Medical Physics and SchoolJul 31, 2019@ICISE, Quy Nhon, Vietnam

Page 2: Clinical experience of Monte Carlo dose calculation in

• Introduction

• Development of MC system

• Independent dose calculation

• Speeding up by GPU

Outline

Page 3: Clinical experience of Monte Carlo dose calculation in

• Introduction

• Development of MC system

• Independent dose calculation

• Speeding up by GPU

Page 4: Clinical experience of Monte Carlo dose calculation in

PBA vs. MC

Paganetti PMB 53 4825 2008MGH, US Yamashita PMB 57 7673 2012

HIBMC, Japan

Schuemann IJROBP 92 1157 2015MGH, US

Range uncertaintyDose to OAR

Max. error in dose to target: 5%TCP: 11%

• “The most accurate, and hence desirable, dose-estimation algorithm are

Monte Carlo models.”

– ICRU 78, 2007

TPS (TPS): Pencil beam algorithm (PBA)Large uncertainty in inhomogeneous structure

Page 5: Clinical experience of Monte Carlo dose calculation in

Spot scanningGantry 1for head and neck, etc.

Gantry 2 Double scatt.for lung and liver

Fixed BeamDouble scatt.for prostate

250 MeVsynchrotron

Injector7 MeV Linac RFQ+DTL

NPTC -Nagoya Proton Therapy Center

Broad-beam

Page 6: Clinical experience of Monte Carlo dose calculation in

Patients

2013-May 31, 2019

Total 2778

Prostate45%

Liver20%

Lung13%

Pancreas2%

H&N4%

B&S1%

Pediatric2%

Others13%

Page 7: Clinical experience of Monte Carlo dose calculation in

• Introduction

• Development of MC system

• Independent dose calculation

• Speeding up by GPU

Page 8: Clinical experience of Monte Carlo dose calculation in

Data flow

MIM Maestrocontouring

dose evaluation

MOSAIQ (Elekta)OIS

CT

VQA (Hitachi)TPS (PBA)

Irradiation machine

Monte Carlo

CT image, coordinate system,Range shifter, energy absorber, aperture

Beam parametersposition(xi,yi), Energyi, MUi

half a day for one patientall patients for scanning 1.2x108 protonssome patients for double scatt.

Linux PC clusterIntel Xeon 364 cores

Page 9: Clinical experience of Monte Carlo dose calculation in

Treatment Nozzles

Broad-beam Double scatt.

Spot scanning

range modulation wheel

range compensator

MLC

2nd scatt.

scanning magnets

~3m

aperture

Page 10: Clinical experience of Monte Carlo dose calculation in

Software development

Geant4

1M lines

International collaboration 100 developers, 20 years

PTSIM

100k lines

10 developers mainly in Japan, since 2003

Dedicated for Nagoya Proton Therapy Center

10k lines

2 medical physicists, since 2012

Page 11: Clinical experience of Monte Carlo dose calculation in

71.6MeV

139.3MeV

221.4MeV

(mm)

I=77eV

PDD in water OCR for single spot in water139.3MeV Depth: 100mm

0

0.2

0.4

0.6

0.8

1

1.2

-100 -50 0 50 100

Measurement

MC

0

0.2

0.4

0.6

0.8

1

1.2

-100 -50 0 50 100

Measurement

MC

(mm) (mm)

0

5

10

15

20

25

0 50 100 150

Measurement in

MC in

Measurement cross

MC cross

0

5

10

15

20

25

30

35

50 100 150 200 250

Measurement in

MC in

Measurement cross

MC cross

Depth (mm) Energy (MeV)

FWH

M(m

m)

FWH

M(m

m)

139.3MeV

D200mmD100mm

D50mm

D25mm

In Cross

Verification

Page 12: Clinical experience of Monte Carlo dose calculation in

Verification and absolute dose normalization

R30 SOBP10 FS10

0

20

40

60

80

100

120

-300 -250 -200 -150 -100 -50 0 50 100 150

0

20

40

60

80

100

120

-200 -150 -100 -50 0 50 100 150

R20 SOBP10 FS10(mm)

0

20

40

60

80

100

120

-100 -50 0 50

R8 SOBP10 FS10(mm)

(mm)

(mm)0

20

40

60

80

100

120

-150 -100 -50 0 50 100

R12 SOBP10 FS10

10x10 field PDD Normalization

Measurement

MC

Page 13: Clinical experience of Monte Carlo dose calculation in

• Introduction

• Development of MC system

• Independent dose calculation

• Speeding up by GPU

Page 14: Clinical experience of Monte Carlo dose calculation in

Work flow for patient-specific QA

TPS OIS irradiation machine Measurement

Log-file analysisMonte Carlo

Log-file

irradiation fileDICOM-RT

Comp.

Comp.

We have carried out MC dose calculation as independent dose calculationin patient-specific QA for more than 400 patients since 2013.

Page 15: Clinical experience of Monte Carlo dose calculation in

70.2 GyE to the CTV in 26 fraction (2.7 GyE/fr)

MC

TPS (PBA)

3 fields

Clinical Case 1

MCTPS(PBA)

CTV

GyE

Page 16: Clinical experience of Monte Carlo dose calculation in

Clinical case 2Paranasal sinus 70.2GyE/26frac SFUD

3 fields 30°,80°,120°

MCTPS(PBA)

CTVPTV

GyE

54.7GyE

49.9GyE

MC

TPS (PBA)

80°

modified plan

• Proton scanning for H&N: Overdose to optic nerve was revealed by MC in two cases out of 78 head and neck patients, and treatment plan was modified.

T. Toshito et al., The 30th Annual meeting of the Japanese Society for Radiation Oncology 2017

Page 17: Clinical experience of Monte Carlo dose calculation in

Clinical case 370.2 GyE/26frac IMPT

3 fields 195°,80°,110°

80°

re-plan from 7th fraction

⇒104%

MC

TPS(PBA)• IMPT for H&N: Intolerable dose to OAR was revealed by MC and

treatment plan was modified in two cases out of 43 cases. T. Toshito et al., The 116th Scientific meeting of the Japan Society of Medical Physics 2018

Dmax to skinMC 114%PB 106%

Page 18: Clinical experience of Monte Carlo dose calculation in

Clinical case 4Artificial bone made by titanium

MC

TPS (PBA)

one field6 fields

Page 19: Clinical experience of Monte Carlo dose calculation in

• Introduction

• Development of MC system

• Independent dose calculation

• Speeding up by GPU

Page 20: Clinical experience of Monte Carlo dose calculation in

MPEXS: Massive Parallel Electro X-ray SimulatorGeant4 EM on CUDA for medical applicationSpeedup factor : ∼440References

N. Henderon, et al. Joint International Conference on Supercomputing in Nuclear Applications and Monte Carlo 2013 (SNA + MIC 2013) DOI: https://dx.doi.org/10.1051/snamc/201404204 K. Murakami, et al. Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC), 2013 DOI: 10.1109/NSSMIC.2013.6829452

MPEXS-DNAGeant4-DNA on CUDA DNA-scale physics and chemical processesSpeedup factor : ∼1000References

S. Okada, et al. M&C + SNA + MC 2015 Proceedings CD-ROM: http://www.ans.org/store/item-700397/S. Okada, et al. Innovation in Medicine and Healthcare 2015 DOI: https://doi.org/10.1007/978-3-319-23024-5_29

MPEXS [Massive Parallel Electro X-ray Simulator]

:Geant4-based GPU Simulators

Developed in Stanford univ./SLAC/KEK https://wiki.kek.jp/display/mpexs/MPEXS+Project

NVIDIA CUDA computing platform

GeForce GTX 1080Ti 1480 MHz 3584 CUDA coresTesla K40 745 MHz 2880 CUDA cores

Parallel tracking

Page 21: Clinical experience of Monte Carlo dose calculation in

Migration from PTSIM to MPEXS-h

• CPU-based Particle Therapy MC Simulation framework using Geant4

• Versatile• Dose calculation• Machine design• Commissioning• Neutron dose

• Since 2003

MPEXS-h

• GPU-based MC for hadron on MPEXS platform

• Standard EM and Binary Cascade model in Geant4 are transplanted to CUDA

• Dedicated to dose calculation• Since 2017

Page 22: Clinical experience of Monte Carlo dose calculation in

Conclusion

• In Nagoya Proton Therapy Center, in-house Monte Carlo dose calculation system was developed.

• It has been used for independent dose calculation for patient specific QA.

• Treatment planning was modified for a few percent of clinical cases according to Monte Carlo dose calculation.

• We demonstrated that Monte Carlo dose calculation was clinically useful.

• Migration to GPU-based system is conducted to speed-up calculation.