MC model of the Elekta EPID

Comments Off on MC model of the Elekta EPID

This page describes some of the background and motivation behind some aspects of the Kairn BEAMnrc/DOSXYZnrc model of the Elekta iView EPID. This model is widely used for research at QUT (and is the basis for some subsequent models). You can tell that you are using the Kairn DOSXYZnrc  model (or a modified version of the Kairn model) if the printed circuit board layer of the EPID is modelled as copper with a density of 3.95 g/cm^3. You can tell that you are using the Kairn BEAMnrc model if the SLABS layers above and below the gadox are called “Agadox” and “Bgadox” respectively. If you are using any of these models (or any models based on them) or if you find any of the information below useful, please cite: Kairn et al, Phys Med Biol, 53(14) 3903-3919 (2008).

The a-Si Electronic Portal Imaging Device

  • The a-Si EPID is comprised of several layers of material combined to give the maximum imaging efficiency in response to a high energy photon beam.
  • Beneath a protective covering layer, a metallic buildup plate a few millimetres thick acts to create a shower of electrons in response to an incident photon beam.
  • These electrons are projected down into a layer of scintillating material (in our case, terbium-doped gadolinium oxysulphide or ‘gadox’) where they interact to produce optical photons.
  • These optical photons are then detected  by an array of photodiodes residing on the silicon microarray, which produce an electrical signal which is read out via the transistor array.
  • The images thus produced are automatically gain and dark-field corrected and stored as computer data files.

The Elekta iView

The Elekta iView a-Si EPID is modelled according to manufacturers specifications. Key features of this particular panel are:

  • The lateral dimensions of the active area of the panel are 41cm x 41cm.
  • This area is covered by 1024×1024 pixels.
  • The electronic buildup layer above the scintillator is a thin metal sheet.
  • The Gd2O2S active layer is situated in direct contact with the a-Si layer.
  • A printed circuit board supports the electronics.
  • A second metallic filter plate is located below all of these components, to eradicate low-energy backscattered radiation coming from downstream support structures, or the floor.
  • The entire panel is enclosed in a cover made of high-density polystyrene and acrylonitrile butadiene styrene.
  • The top surface of the EPID housing is approximately 157.1cm from the photon source.
  • The Gd2O2S active layer of the EPID is approximately 160cm from the photon source.
  • The total thickness of the panel from top cover to bottom cover is nearly 10cm.

Coordinate System

  • The z axis is parallel to the beam, with positive being directed away from the source.
  • The x and y axes are normal to the beam and to each other, with x being parallel to the direction of motion of the MLC leaves (at the RBWH) when the collimator angle is zero.
  • The surface of the EPID is parallel to the x-y plane.
  • The gantry angle is equal to zero degrees when the beam is pointed straight at the floor.

Modelling the Elekta iView EPID using BEAMnrc

  • The model EPID panel is also included in a series of BEAMnrc simulations of the entire linac-phantom-detector system.
  • This allows us to record and examine phase space files (describing the positions, directions, energies, types, etc of particles present) in planes at a range of locations above, below and within, the EPID.
  • Consequently, many separate ‘SLABS’ component modules (rather than one CM with many layers) are used to simulate:
    • The region between linac and top of phantom;
    • The phantom;
    • The region between phantom and (optional) additional buildup material on top of EPID;
    • (Optional – it could be made of air) buildup material;
    • Set of EPID layers above active layer;
    • Active layer of EPID; and
    • Set of EPID layers below active layer, as well as air and floor below detector (if floor is parallel to EPID plane).
    • The EPID model in BEAM can be easily manipulated by:
      • Including or omitting an external metallic buildup (or filter) plate on the top surface of the EPID;
      • Including, excluding or varying a range of reclilinear phantoms between linac and detector; and
      • Modifying the composition of the detector itself.
      • To include components which cannot be described using this simple ‘SLABS’ setup, a new linac model must be built.
      • Eg. Addition of a ‘MIRROR’ CM allows the floor to be included in simulations where the gantry angle is not 0o, 90o or 270o.

Modelling the Elekta iView using DOSXYZnrc

  • DOSXYZnrc is used to model the phantom and EPID, with the incident particle source being defined as the phase space output of a BEAMnrc simulation of the linac treatment head.
  • This allows calculation of dose in the phantom and in the EPID itself.
  • Usually, we’re interested in the signal at the Gd2O2S active layer of the detector.
  • Because of the large number of media comprising the phantom-detector system, you may need to modify $MXMED in dosxyznrc_user_macros.mortran and then re-compile your phantom (‘make’) dosxyznrc.
  • The real EPID panel has an active area with dimensions 41x41cm2 in the x-y plane, but the panel is surrounded by additional material (electronics and housing) not described in the manufacturer’s specifications.
  • The original DOSXYZnrc model EPID had dimensions 42x42cm2 in the x-y plane, to allow for some additional scattering into the active area from electronics and housing materials not included in the model, although this has been removed in more-recent versions, to simplify comparison with experimental images.
  • This 42cm width was divided into 168 equally-sized voxels, in each direction, although this, too, has been altered to simplify experimental comparison.
  • The number 168 was chosen so that each pixel was 0.25cm across. Reducing this to 0.1cm increased the noise in the image (requiring an increase in the total number of histories simulated), while also extending the simulation time (even for the same number of histories).
  • The z voxel dimensions are defined in terms of groups (unlike Mohammad’s model)
  • The minimum z boundary is defined as zero.
  • The maximum z boundary occurs at the base of the EPID housing, approximately 115cm from the Elekta linac’s exit plane.
  • Between the linac and the top surface of the EPID, four groups (set to air) are included to allow for the inclusion of a very adaptable phantom volume, surrounded by air.
  • In the EPID itself, each group is defined as containing a single voxel in the z direction, with the ‘width’ of that voxel being equal to the thickness of the appropriate layer of the EPID.
  • This allows changes to the thickness of each layer (of EPID or phantom or air) to be made straightforwardly, and without the necessity of redefining the positions of all the other layers.
  • This also allows easy increases to the resolution of each layer to be made, by simply increasing the number of voxels in the requisite group.

Phantom and EPID layers in the DOSXYZnrc model

  • The DOSXYZnrc model of the EPID includes all layers down to the base of the ABS housing of the panel.
  • Above the EPID, voxels 1, 2, 3 and 4 are used to model the setup of whatever (rectilinear) phantom is being examined.
  • The active layer of the detector is found at z voxel number 12, which is also z group number 12.  (Note that in IDL this is z voxel number 11.)
  • The x-y dimensions of the phantom need not equate to those of the detector.
  • The position, thickness, material and density of the phantom can very easily be modified .
  • Some of the materials appearing in the EPID can be modelled straightforwardly:
    • Copper: CU700ICRU at r=8.96g/cm3
    • Aluminium: AL700ICRU at r=2.7g/cm3
    • Polystyrene: PS700ICRU at r=1.03/cm3
  • The materials and properties used in the EPID model which require justification are:
    • Air: density of dry air at sea level r=0.00129g/cm3
    • Lanex Fast intensifying screen: el Mohri et al examined Lanex Fast B, Lanex Regular and Lanex Fine scintillation screens. “For all cases, the phosphor layer was modeled as Gd2O2S with a density of 3.67g/cm3. This is approximately 50% of the bulk density of Gd2O2S. This reduction in density accounts for the polymer binder and small air pockets contained within a realistic phosphor layer.”
    • Carbon fibre: 170C700ICRU (graphite) at the lower density of carbon fibre, r=1.6g/cm3
    • Printed circuit board (PCB): The most electron-dense contribution to this material comes from its approximately 35%(vol) copper component. So, as above, let the material be CU700ICRU, but at the lower density of 3.95g/cm3.

One way to justify these choices is to cite el Mohri et al, Medical Physics 28(12) 2538 (2001), and this website. Another way to justify these choices is to acknowledge that you are using the Kairn model: Kairn et al, Phys Med Biol, 53(14) 3903-3919 (2008).

Comments are closed.