Geant4 applications to proton radiation therapy. Harald Paganetti

Transcription

1 Geant4 applications to proton radiation therapy Harald Paganetti

2 Geant4 application to proton radiation therapy Proton therapy techniques in a nutshell Monte Carlo applications: Treatment head developments Quality assurance Phase space calculations Patient simulations Clinical use Time dependent patient geometries Neutron dose in proton beam therapy PET imaging for quality assurance

3 Geant4 application to proton radiation therapy Proton therapy techniques in a nutshell

4 Dose shaping for passive scattered protons Hanne Kooy, MGH High-Density Structure Target Volume Beam Critical Structure Double scattering system Aperture Body Surface

5 SOBP: Spread-Out Bragg Peak

6 SOBP Modulation High-Density Structure Target Volume Beam Critical Structure Hanne Kooy, MGH Aperture Body Surface Prescription: Range Modulation Compensator Aperture

7 Range Modulator Wheel

8 Hanne Kooy, MGH Aperture and Range Compensator + = To be designed by the planning system!

9 Typical treatment head Range Modulator Wheels IC1 Aperture, Compensator First Scatterers Second Scatterers IC2 and IC3

10 Alex Trofimov, MGH Proton beam scanning Bragg peaks of pencil beams are distributed throughout the planning volume Pencil beam weights are optimized for several beam directions simultaneously (inverse planning)

11 Proton beam scanning Intensity Modulated Beam Fast Slow Vacuum Chamber Y Z Pair of Quads Scanning Magnets Beam monitor 0.8 m 0.6 m X Alex Trofimov, MGH

12 Monte Carlo for proton therapy simulations Monte Carlo code Physics Setup is not a Monte Carlo simulation executable is an Object Oriented Simulation Toolkit in C++ advantage over other codes: flexibility

13 Monte Carlo for proton therapy simulations

14 Geant4 application to proton radiation therapy Treatment head developments

15 Ben Clasie, MGH Proton therapy nozzle Vacuum chambers for scanned beam delivery Quadrupole magnets Dipole magnets Proton beam Proton beam Quadrupole vacuum chamber Quad chamber Dipole vacuum chamber Schematic of the chamber positions Dipole chamber

16 Ben Clasie, MGH Possible design improvement Contamination comes from protons with reduced energy Smoothing edges and high-density absorbing material reduces contamination

17 Ben Clasie, MGH Possible design improvement Better agreement in the slope 1.5% change in the output factor

18 Ben Clasie, MGH New chamber solution Top View 3 mm thin fiberglass walls Tapered inner surface Side view High-density absorber with a parabolic taper

19 Ben Clasie, MGH New chamber solution Monte Carlo simulations Field flatness is greatly improved, but there is still a reduction in the field diameter Slope has good agreement 0.3% change in output factor

20 Geant4 application to proton radiation therapy Quality assurance

21 Examples: Quality Assurance / Tolerance Studies Beam scraping (device alignment sensitivity) total dose relative dose total dose minus scraping 0.2 dose from scraping the FS frames depth [cm]

22 Examples: Quality Assurance / Tolerance Studies Alignment of second scatterer Dose [%] Dose [%] second scatterer aligned Depth [mm] Lateral distance to isocenter [mm] Dose [%] Dose [%] second scatterer tilted by 5 o Depth [mm] Lateral distance to isocenter [mm]

23 Examples: Quality Assurance / Tolerance Studies Wheel rotation potentiometer setting Dose [%] Depth [mm]

24 Geant4 application to proton radiation therapy Phase space calculations

25 Phase Space plane Phase Space Format Example: (part, x, y, p x, p y, p z, flags )

26 Parameters to characterize the beam at nozzle entrance 1. Beam size and spread (IC measurement) 2. Beam angular spread (manufacturer info) 3. Beam energy (control system) 4. Beam energy spread (manufacturer info, measured) Are these parameters correlated?

27 Commissioning of the Monte Carlo Dose [%] Dose [%] Depth [mm] Angular Spread [rad] Angular Spread [rad] N Beam Size Sigma [cm] Beam Size Sigma [cm] Depth [cm] N 0.02 Angular Spread [rad] 0.01 Angular Spread [rad] Beam Size Sigma [cm] Beam Size Sigma [cm]

28 Snout retraction area Range Modulator Wheels Magnet 2 Jaws (X and Y) (& Range Verifier) IC1 Snout First Scatterers Magnet 1 Second Scatterers IC2 and IC3

29 Monte Carlo model of the nozzle (~1000 objects) - Not patient specific -

30 High Z Low Z Dose [%] Depth [mm]

31 Modeling time dependent geometrical setups Key to 4D Monte Carlo: Geometry changes during the simulation via C++ class architecture based on GEANT4 Geometry update command in DetectorMessenger DetectorConstruction: rot_rmw = new G4RotationMatrix(); rot_rmw->rotatez(wheel_angle*degree); RMW_Phys -> SetRotation(rot_RMW); G4RunManager* therunmanager = G4RunManager::GetRunManager(); therunmanager->defineworldvolume(worldphys); therunmanager->geometryhasbeenmodified(); therunmanager->resetnavigator();

32 Range Modulator Wheel Issues 1. Beam Gating Dose [%] Beam range: cm Modulation: 6.78 cm Depth [mm] 2. Beam Current Modulation 120 Dose [%] Beam range: cm Modulation: 6.78 cm Depth [mm]

33 Dose [%] Dose [%] Dose [%] Dose [%] depth [cm]

34 Dynamic Systems in Radiation Therapy - Beam Delivery - IMRT Tomotherapy Proton Therapy IMPT Types of variations: IMRT: moving leafs Tomotherapy: rotating beam Protons: rotating wheel IMPT: changing magnetic field

35 Scanning Magnet simulation

36 IMRT simulation

37 Aperture and Compensator Monte Carlo simulation based on milling machine files

38 Simulation of ionization chambers Volume for absolute dosimetry Output Factor D cal i ic cgy MU

39 Absolute dosimetry (output factor prediction) by simulating the ionization chamber output charge Output Factor D cal i ic cgy MU i ic e ε ic de = p dξdf W dx air air Volume for absolute dosimetry cm < SOBP r < 15 cm cm < SOBP r < 24 cm Output Factor [cgy / MU] Output Factor [cgy / MU] (SOBP r - SOBP m ) / SOBP m (SOBP r - SOBP m ) / SOBP m

40 Geant4 application to proton radiation therapy Patient simulations

41 Patient information Example: CT scan: 134 CT slices, voxels/slice mm mm 1.25/2.5 mm Treatment planning grid: 2.0 mm 2.0 mm 2.5 mm

42 Issues related to dose calculations on CT data: Memory Consumption - Solution within the Geant4 framework Parameterized volume: allows the least computer memory usage for CT voxels (two bytes per voxel) Modification of the Geant4 source code Slow tracking through Much faster way to transport particles in CT voxels Abandoned parameterized the general, but volumes low-efficient algorithm in GEANT4.

43 CT conversion Density and Material composition Density and Material composition

44 Issues related to dose calculations on CT data: CT conversion Solution within the Geant4 framework Solution: HU space is divided into 27 groups with members of each group sharing the same element composition but differ in mass density Modification of the Geant4 source code Dynamic assignment of mass density Only one material is defined for each CT HU group. The composition of the material is preserved, but the mass density varies ~ 3000 materials (cross section sets)

45 HU conversion Group HU range Density Density Material [g/cm3] correction composition weights [%] (center of HU bin) H C N O Na Mg P S Cl Ar K Ca Ti 1 [ ; -951 ] [ -950 ; -121 ] [ -120 ; -83 ] [ -82 ; -53 ] [ -52 ; -23] [ -22 ; 7 ] [ 8 ; 18 ] [19 ; 79 ] [ 80 ; 119 ] [ 120 ; 199 ] [ 200 ; 299 ] [ 300 ; 399 ] [ 400 ; 499 ] [ 500 ; 599 ] [ 600 ; 699 ] [ 700 ; 799 ] [ 800 ; 899 ] [ 900 ; 999 ] [ 1000 ;1099 ] [ 1100 ; 1199 ] [ 1200 ; 1299 ] [ 1300 ; 1399 ] [ 1400 ; 1499 ] [ 1500 ; 1599 ] [ 1600 ; 1999 ] [ 2000 ; 3060 ] [ 3061 ; ]

46 Gantry Angle Position XYZ Rotation Pitch Roll Air Gap Gantry Couch ISO AP1A 0º 0º 1 AS1A 65º 270º 1 RS1A 305º 50º 1 RA1A 295º 0º 2 RS2A 300º 60º 2 AS2B 90º 270º 3

47 Geant4 application to proton radiation therapy Clinical use

48 Patient Database Treatment Planning FOCUS/XiO Treatment Control System Treatment Head Geometry Patient Geometry Beam Setup Patient Setup Phase Space Calculation Dose Calculation Monte Carlo Dose Cube Monte Carlo Dose Cube FOCUS/XiO format FOCUS and XiO: Computerized Medical Systems Inc.

49 GUI program

50 Logistics MGH: Linux cluster with 70 processors Runtime per patient (using 20 processors): 5 hrs (~2.5% stat. accuracy in the GTV) ~4 hrs for treatment head ~1 hr for patient based on simulation of the entire plan No sophisticated variance reduction techniques used Full physics

51 Example 1 Case 1: Para-spinal tumor 176 x 147 x 126 slices voxels: x x mm 3 3 fields: ~15.0 Gy each

52 Case 1: Para-spinal tumor slices mm 3 3 fields: ~15.0 Gy each PB Dose-to-water PB MC Dose-to-tissue MC MC Dose-to-water 1 Gy 3 Gy 5 Gy 7 Gy 9 Gy 11 Gy 13 Gy 15 Gy 17 Gy Range

53 10 Gy 15 Gy 20 Gy 25 Gy 30 Gy 35 Gy 40 Gy 45 Gy 46 Gy

54 Total DVH Volume [%] Brainstem GTV Dose [Gy] XiO MC (DTW) MC (DTT)

55 MC:Dose-to-tissue 10 Gy 20 Gy 30 Gy 35 Gy 40 Gy 42 Gy 44 Gy 46 Gy 48 Gy PB 1 Gy 2 Gy 3 Gy 4 Gy MC:Dose-to-water Penumbra Range Dose homogeneity

56 Planned SOBP versus delivered SOBP Dose [relative] Depth [cm] Dose [relative] Depth [cm]

57 Example 2 Case 3: Maxillary sinus 121x121x101 slices voxels: x x mm 3 19 fields: 15 protons + 4 photons PTV A Critical Structure B C

58 Proton dose in the presence of range uncertainty

59 Proton dose in the presence of range uncertainty

60 PB CTV MC 100% 5 Gy 10 Gy 15 Gy 20 Gy 25 Gy 28 Gy 30 Gy 32 Gy 34 Gy 1 Gy 2 Gy 3 Gy 50% 4 Gy 5 Gy 6 Gy Range Patch Line 0% Fields

61 PB CTV MC 100% 5 Gy 10 Gy 15 Gy 20 Gy 25 Gy 28 Gy 30 Gy 32 Gy 34 Gy 1 Gy 2 Gy 3 Gy 50% 4 Gy 5 Gy 6 Gy Range Patch Line 0% Fields

62 PB CTV MC 100% 10 Gy 20 Gy 30 Gy 40 Gy 50 Gy 56 Gy 60 Gy 64 Gy 68 Gy 2 Gy 4 Gy 6 Gy 50% 8 Gy 10 Gy 12 Gy Range 0% Fields 5,6,8,9 0 50

63 Case 2: Nasopharynx & both necks 169 x 155 x 125 slices voxels: x x mm 3 8 fields: 7 protons + 1 IMRT Monte Carlo XiO

64 Example 3 Case 2: Nasopharynx & both necks 169 x 155 x 125 slices voxels: x x mm 3 8 fields: 7 protons + 1 photons

65 PB MC MC Range 10 Gy 20 Gy 30 Gy 40 Gy 50 Gy 60 Gy 65 Gy 70 Gy 75 Gy Dose-to-tissue Dose-to-water 1 Gy 2 Gy 3 Gy 4 Gy 5 Gy 6 Gy

66 Volume [%] Left Neck GTV Left Neck CTV Dose [Gy] XiO MC (DTW) MC (DTT)

67 100% 50% MC PB 0% Gy DVH for left submandibular gland

68 Geant4 application to proton radiation therapy Time dependent patient geometries

69 Dynamic Systems in Radiation Therapy - Patient - Tumor motion is of particular interest for intrathoracic tumors Inter-patient variations due to tumor position in the lung, patient breathing pattern, lung capacity, etc. Organs in the thorax and abdomen move approximately periodically along the craniocaudal and anterior-posterior axes with a period of about 4 seconds. Tumor motion in the lung can be up to 4 cm and is on average in the order of 3 7 mm.

70 Lateral Motion of Lung Tumor Significant Lateral Motion

71 Dosimetric effects due to motion in proton therapy S. Mori, MGH

72 - Breathing Patient - Eike Rietzel R L posterior view posterior cut

73 10 x CT 10 x VDM 10 x Dose Dose at a reference phase

74

75 y Deformable Registration Methods y Transform Fixed Image x Moving Image x

76 4D Dose Deposition Beamlet 2 Beamlet 1 Beamlet 2 Beamlet 1 A B C D D1 E F G H I J D2 K L M N O P D1 D2 T = t 1 T = t 2

77 4D Dose Deposition 4D Dose Deposition Dose deposition defined via voxel identifiers, not position in space! Beamlet 2 Beamlet 1 Beamlet 2 Beamlet 1 A B C D D1 E F G H I J D2 K L M N O P A B C D A E B F G C HD E I D1 F J G K HL M I N J K O D2 LP T = t 1 T = t 2

78 Dose calculation during non-rigid motion T = t 1 T = t 2

79 Four-dimensional Monte Carlo simulation based on 4D CT Inhale Intermediate Exhale

80 Four-dimensional Monte Carlo simulation based on 4D CT Solid lines: Dashed lines: Patient in inhale Considering the entire breathing phase

81 Interplay effects of motion (4D Monte Carlo) Time dependent scanning magnet settings Time dependent patient geometry E. Rietzel

82 Interplay effects between organ motion and MLC movement JH Kung and P Zygmanski

83 Single-dynamic (patient movement ; static beam delivery) Beamlet 2 Beamlet 1 Beamlet 2 Beamlet 1 D1 D2 D1 D2 T = t 1 T = t 2 Interplay

84 Double-dynamic (patient movement; dynamic beam delivery) Interplay Beamlet 1 Beamlet 2 Beamlet 2 Beamlet 1 D1 D2 D3 T = t 1 T = t 2 Effect can be reduced by repaiting

85 4D Monte Carlo of IMPT Beamlets and patient are moved continuously (rigid) 4D Monte Carlo Assumptions: Irradiation time per slice is 0.4 seconds (on average) Changing the cyclotron beam energy with a degrader takes a few seconds Breathing cycle is 4 seconds Choose a specific scanning pattern Choose a specific number of protons per second Update the beam delivery setup and the patient setup every 0.1 virtual seconds

86 4D Monte Carlo of IMPT 4D Monte Carlo DVH for left anterior field Static Patient moves ±0.5 cm Patient moves ±1.5 cm Patient moves ±0.5 cm; repainting Patient moves ±1.5 cm; repainting

87 Geant4 application to proton radiation therapy SUMMARY / TAKE-HOME MESSAGES Proton therapy techniques in a nutshell Treatment head developments Quality assurance Phase space calculations Patient simulations Clinical use Time dependent patient geometries Neutron dose in proton beam therapy PET imaging for quality assurance

88 Acknowledgements NIH support: P01 CA21239 RO1 CA RO1 CA Ben Clasie Martijn Engelsman Hongyu Jiang Hanne Kooy Greg Sharp Ziji Wu