Proton Dose Calculations Semi-empirical models for scanning beams and IMPT Uwe Oelfke DKFZ Heidelberg Department of Medical Physics (E040) Im Neuenheimer Feld 280 69120 Heidelberg, Germany u.oelfke@dkfz.de Outline Phase Space, Physics and Dose Dose Kernels: Water Tissue Inhomogeneities Consequences Seite 1 1 Phase space and dose of a beam spot Dose = Phase Space ** Physics Pencil Beam (H20) Depth dose curve A simple model Stopping power, CSDA Range Straggling, energy-spektrum Inelastic nuclear interactions Basic physics Dose factorizes in depth dose and lateral spread Laterally integrated depth dose curve Energy loss due to collisons with e- Straggling Energy spread of dose delivery system Primary and nuclear interactions Seite 2 2 Fluence and depth dose curve Protons D(z,r) = ??? H2O E(0) dN p dN p Φ= Ψ= Ee Energy -Fluence: Fluence: dA ⊥ dA ⊥ Dose and vectorial energy fluence D(x) = dE ab 1 (x) = ∇ Ψ (x) ρ dm Vectorial energy fluence Depth dose curve: a simple model Protons D(z,r) = ??? E(0) D(z) = - H2O 1 ∂ ρ ∂z (Φ(z) E(z)) = - 1 ρ E(z) Inelastic nuclear inter. ∂Φ( z) 1 ∂E( z) Φ( z ) ∂z ρ ∂z Mass Stopping Power (collision) Seite 3 3 Coulomb interaction with electrons Stopping power - Range Stopping Power Microscopic def. : S(E) = E tr ( E) L L = mean free pathlength E R CSDA = ∫ dE' CSDA Range: 0 1 S(E' ) Bethe Bloch Formula 2 Z eff e4 2m c 2 β 2 dE = −4πN Z T ln e − β 2 dx el . me c 2 β 2 I (1 − β 2 ) [ ( Zeff = Z p 1 − exp − 125 ⋅ β ⋅ Z −p2 / 3 )] Barkas-Formel Mass stopping power Seite 4 4 CSDA Range Coulomb interaction with electrons Straggling and Bragg Peak Statistical fluctuations: broadening of energy spectrum Straggling width: Microscopic def. : ∝ R Bragg Peak: Σ(E) = E tr E tr ( E) L L = mean free pathlength ∗∗ ) Depth dose: Straggling – Bragg Peak Seite 5 5 Proton Bragg Peaks – Straggling Energy spectrum of accelerator Inelastic Nuclear Interactions Strong interaction with nuclei in the body creation of slower secondary protons, recoil heavier fragments creation of new background ‚Bragg peaks‘ Lateral scattering at lower energies - halo Seite 6 6 Inelastic nuclear interactions Lateral Scattering Coulomb interaction with nuclei shielded by electron cloud Moliere Theory: Gaussian width + Broader Background Drifting + Scattering = Transport Eq. 2 Gaussians for Modelling – Primary + Secondary (Field Size Effect) Distribution of scattering angles Seite 7 7 Lateral spread of beam (Highland) FWHM = σ * 2.35 160 MeV 200 MeV Multiple Coulomb scattering The Source(s): phase space of the accelator Energy spectra Lateral distributions Angular distributions Source size, position Collimator scattering Extrafocal radiation, ... Calculation ? f i (E, r, z,θ x , θ y ) i = particle type Calibration via a set of measurments or Monte Carlo Techniques ! Seite 8 8 Characterization of Phase Space ‚Physics‘ dose in water is defined by either semiempirical source models or MC calculations Use a set of measurements to calibrate the parameters of a phase space model Aim: To determine a few parameters of your ‚machine‘ such that the dose for all beam spots in water is correct Assumptions: Parameters: Seite 9 9 Beyond Gaussians…. Seite 10 10 Seite 11 11 • Pencil beam in water: • Standard 1D scaling: pathlength scaling Independent of inhomogeneity position and scattering properties • New 2D scaling: pathlength AND radial scaling Depends on inhomogeneity position and scattering properties Pencil beam scaling methods m r S 1 P (r , z , E 0 ) = P w , zeq , E 0 2 ρ w (Fr ) Fr • 1D scaling: Fr=1 • 2D scaling: Fr = σ (z , E 0 ) σ w (zeq , E 0 ) Use of Highland’s formula for Multiple Coulomb Scattering Seite 12 12 Implementation of the 2D scaling We need: Depth dose curves in water CT calibration curve relating Hounsfield numbers to relative stopping powers Simple analytical formula for calculating the standard deviation of lateral spread (function of depth and energy) CT calibration curve relating Hounsfield numbers to material specific lateral scaling factors CT Calibration curves Lateral Scaling Factors Relative Stopping Powers 1,5 1,0 0,5 0,0 -1000 Lateral Scaling Factors 1000 2,0 100 10 1 -200 -100 0 100 200 Hounsfield Units 1 0,1 -500 0 500 1000 1500 2000 -1000 -500 Hounsfield Units 0 500 1000 1500 2000 Hounsfield Units Example: Homogeneous Media Hanitra Szymanowski and Uwe Oelfke: Two-dimensional pencil beam scaling: an improved proton dose algorithm for heterogeneous media 2002 PMB 47. Seite 13 13 Example: Homogeneous Media Water Bone 0,4 MC 1D 2D 0,3 0,2 Sigma (cm) Sigma (cm) 0,4 0,1 0,0 0 5 10 MC 1D 2D 0,3 0,2 0,1 0,0 0 15 Depth (cm) 5 10 Depth (cm) Homogeneous medium: bone Lateral dose distribution Bone at z=4cm Bone at z=8cm 20 MC 2D 1D 15 10 10 5 5 0 -0,6 -0,4 -0,2 0,0 0,2 Tiefe (cm) 0,4 0 -0,6 -0,4 -0,2 0,0 0,2 Tiefe (cm) 0,6 0 ,3 MC 1D 2D 0 ,1 0 2 4 6 Sigma (cm) S ig m a (c m ) 0,4 bone 0 ,2 0,2 2 4 0 ,1 0 ,0 0 2 4 6 8 10 12 14 16 18 D e p th (c m ) S ig m a (c m ) S ig m a (c m ) 0 ,2 6 8 10 12 14 16 18 A ir s la b MC 1D 2D 0 ,3 bone Depth (cm) A ir s la b a ir 0,6 0,1 0,0 0 8 10 12 14 16 18 MC 1D 2D 0,3 D e p th (c m ) 0 ,4 0,4 Bone slab B o n e s la b 0 ,4 0 ,0 MC 2D 1D 15 Dosis (%) Dosis (%) 20 0 ,4 MC 1D 2D 0 ,3 a ir 0 ,2 0 ,1 0 ,0 0 2 4 6 8 10 12 14 16 18 D e p th (c m ) Seite 14 14 Comparison to GEANT4 Lateral inhomogeneities Influence of dose algorithm on IMPT Clinical example: • Lung tumor • 5 equally spaced proton beams, 3mm wide proton spots • 3D spot scanning technique • Distal-Edge-Tracking technique Treatment plans: • Optimisation with 1D-scaling - Recalculation with 1D-scaling • Optimisation with 1D-scaling - Recalculation with 2D-scaling • Optimisation with 2D-scaling - Recalculation with 2D-scaling Seite 15 15 Lung: Systematic error Optimisation: 1D scaling Lung DET Volume (%) Recalc:1D scaling Recalc:2D scaling 110 100 90 80 70 60 50 40 30 20 10 0 PTV opt:1D calc:1D Lung opt:1D calc:1D PTV opt:1D calc:2D Lung opt:1D calc:2D 0 10 20 30 40 50 60 70 80 Dose (Gy) % D m in P TV Lun g D m ax D m e an D m in D m a x D m ea n ∆ s (D E T) 7.5 3.9 5.4 0.0 3.9 0.2 3.8 9.0 4.6 0.0 9.0 0.2 ∆ s (3D ) Lung: Convergence error Recalculation: 2D scaling Optim:2D Volume (%) Optim:1D Lung DET 110 100 90 80 70 60 50 40 30 20 10 0 PTV opt:1D calc:2D Lung opt:1D calc:2D PTV opt:2D calc:2D Lung opt:2D calc:2D 0 10 20 30 40 50 60 70 80 Dose (Gy) % D m in P TV Lung D m ax D m ean D m in D m ax D m ean ∆ c (D ET) -7.1 -1.9 -5.4 0.0 -1.9 -0.5 -4.9 -8.6 -5.8 0.0 ∆ c (3D) -8.6 -0.6 Conclusions Calibration of dose for beam spots in water has to be done with outmost accuracy Integration of phase space and physics is done with phenomenological models There are clear limitations of this approach for tissue inhomogenieties Seite 16 16