Dose calculation algorithms Flashcards
what is TG43?
dose calculation formalism introduced by the AAPM in which the effects of several physical factors (e.g., ISL, scatter, attenuation) on the dose rate distribution are considered separately. Tabulated values of functions representing these different contributions are interpolated to determine the dose distribution for a particular source
limitations of TG43
o Assumes everything is water (even outside the patient) – does not account for heterogeneities, presence of applicator (especially relevant for gynecological treatments).
Inhomogeneities affect fluence reaching point of interest by modifying scatter and attenuation, and also affect energy deposition at point of interest.
o For permanent implants, orientations of seeds may change over time. Dose calculation assumes either a point source, or a line source approximation.
o Interseed shielding is ignored (pre-calculated dose distributions for single sources are superimposed).
o Seed orientation is unknown.
effects of TG43 limitations vs MC calc
o Interseed attenuation can be 10% effect along needle direction for LDR prostate implants.
Also, MC D90 is 6% lower than TG-43
Calcifications in the prostate can lower D90 by up to 25%
o Intraoperative I-125 lung brachytherapy:TG-43 D90 is 36% lower than MC
What is Advanced collapsed cone engine?
Elekta’s collapsed cone convolution
describe collapsed cone convolution
uses superposition-convolutions techniques with angular discretization of pre-calculated kernels along a radiation transport grid (these are referred to as “collapsed cones”)
Energy released within some solid angle (i.e., within a cone) is transported and deposited along the cone central axis. Electron transport is not modelled.
advantages of collapsed cone over TG43
models dose deposition in non-water media
patient’s tissue and finite size are accounted for
models applicator
tissue assignment in Oncentra
assign a particular mass density for an ROI, or mass density can be image based
the 3 contributions to the final dose per Elekta’s advanced collapsed cone engine
Primary photons: these are transported using a ray tracing algorithm which assumes dose = collision kerma (charged particle equilibrium is assumed). Non-water mass energy absorption coefficient (muen) data are used. Primary photon dose is an analytical calculation
Once-scattered photons: energy deposition from these is calculated using collapsed cone superposition/convolution method using pre-calculated first scatter kernels for water which are scaled according to the density of the medium (consider effective pathlength through cone lines).
Higher order scattered photons: these are handled the same as once-scattered photons; however, multiple-scatter kernels are used.
what does the kernel describe?
describes the spread of energy deposition surrounding an interaction point (which is taken to occur at the origin of the coordinate system of the kernel)
what is multi resolution voxelaization in collapsed cone convolution?
- variable computation grid size
- used to speed up computation
- higher spatial resolution near the sources
o The collapsed cone convolution method takes advantage of the fact that at large distances the kernel values fall off rapidly (such that the decrease in resolution as you move away from the centre of the kernel does not have a considerable effect on the final dose distribution).
how is degree of angular discretization determined in advanced collapsed cone engine?
- depends on number of dwell positions
- more dwell positions = fewer transport directions
what is ray effect?
arises due to the collapsed cone approximation i.e., due to angular discretization; washes out when there are more dwell positions. If the cone opening is smaller than the voxel size, then the artefact is not present (i.e., can use a higher resolution to get rid of this artefact).
how does collapsed cone algorithm reduce computation time vs MC?
reduces NxNxN calculations (3D voxel grid) to MxN calculations (M rays and N voxels along each ray direction)
What is Acuros BV Algorithm
Varian’s brachy dose calc algorithm
describe acuros BV algorithm
- primary photon dose and first scatter photon fluence is calculated using ray tracing (same as Elekta ‘s advanced collapsed cone)
- grid-based Boltzmann solver calculates the scattered dose
- iteratively solves the LBTE numerically
how does Acuros BV discretize?
discretizes three space variables (giving the position), two angular variables (giving the direction) and one energy variable. Acuros solves this discretized equation to determine the angular and energy dependent photon fluence at every spatial degree of freedom in the computational domain.
how are brachy sources modelled in acuros BV?
modelled as point sources since they are geometrically small compared to the spatial domain in which dose is calculated
does acuros BV vary the computational grid size?
yes, to balance speed with accuracy
once fluence is solved, how does acuros BV solvve for dose?
- electron transport is ignored due to short range of electrons vs voxel size
- dose is assumed to be equal to the kerma [also assuming kerma = collision kerma i.e., radiative losses are ignored].
fundamental data used by Acuros BV
macroscopic atomic cross sections
determined based on the material present in the voxel, which is determined based on pixel intensity
types of systematic errors for Acuros BV
Angular discretization typically results in ray effects.
Energy discretization errors: solution biases present over a large region.
Spatial discretization errors: local solution over/under shooting; generally most pronounced in high dose gradients regions (e.g., in penumbra regions behind shields)
explain ray tracing for advanced collapsed cone engine and acuros BV
-model energy transfer to the medium and attenuation of incident fluence according to the attenuation coefficient
-To get the terma, multiply the fluence by the mass attenuation coefficient of the medium at the location of interest (or integrate over the spectrum if polyenergetic).
- In ACE, the energy lost by the primary photon (the terma) is spatially distributed according to the scatter kernel.
o Photon transport using ray tracing is in a straight line unlike Monte Carlo photon transport…
briefly explain MC
-photon interactions (and resulting changes in direction, creation of secondary particles) are explicitly modelled.
o Sample incident photon energy, direction, position and transport to the first boundary.
o Sample distance to the first interaction according to the attenuation coefficient, and transport photon to this interaction location.
o Choose type of interaction and sample energy, direction of new particles and original particle if it still exists.
o Transport photon until it leaves the geometry or its energy is below the transport threshold.
o Also transport secondary electrons, keeping track of delta rays and bremsstrahlung as well; score resulting energy deposition.
o Repeat for all histories