Treatment Planning - Inhomogeneities

When looking at beam profiles, isodose charts, or percentage depth dose charts, the dose
displayed is modeled in a phantom comprised of a homogenous, water or water-equivalent
material. Linear accelerators are calibrated to deliver an equivalent amount of dose/MU at a
certain depth within a phantom under reference conditions. However, the patients we treat are
not a single, uniform density as the phantom is. Different tissues within the patient have different
densities. For the megavoltage energy range utilized in linear accelerators, Compton scattering is
the primary interaction process that occurs, and the attenuation of the beam is dependent upon
the electron density (g/cm3) of the material it transverses.1 The density of the medium not only
has an effect on attenuation but on absorption and scattering. These effects are particularly
significant in, and at boundaries of, big changes in density – such as the interface between lung
and surrounding soft tissue, or at the interface of soft tissue and a metal implant.

For air, or air-equivalent cavities, such as the lung but also sinuses and air pockets in the
digestive system, the beam penumbra widens due to increased range of laterally scattered
electrons – the range of these electrons in air is about three times as far as it is in soft tissue. As a
result of this, there is decreased scatter back towards the central axis as there are fewer
interactions occurring. This reduced attenuation in lung leads to an increase in range of electrons,
of about three times. This leads to an increase in dose at depth, both to lung tissue and to
structures beyond the lung. This also results in a loss of electronic equilibrium at the soft tissue
interface after passing through air, with reduced dose on the soft tissue side and increased dose
on the lung side, creating a secondary build-up effect. Since this soft tissue interface is likely the
tumor that we are trying to treat, the resultant secondary build-up effect leads to under-dosing at
the periphery of the target.

For materials that are denser than soft tissue, such as bone, but particularly metal implants made
of gold, titanium, or some other such material, there are more interactions occurring within the
denser material. Backscatter into the preceding soft tissue is more pronounced, leading to areas
of higher dose. Meanwhile, there is dose perturbation in the following soft tissue after the
implant. At lower energies, there initially is an area of lower dose than might be expected within
a homogenous phantom, but then an increase in dose from the forward scatter of electrons from
the high density material. At higher energies, where pair production is also a frequent interaction,
there is higher dose deposition immediately after the bone, decreasing with the range of the
electrons set in motion.1 When reconstructing transmission data to produce CT images of a
patient, this effect is rendered by streaking and artifacts within the image. This appears as areas
of low density – i.e. air – around the implant, when the density is in fact of soft tissue.

Historically, the differing densities within the patient were simply not accounted for. Treatment
plans were created on the basis that the patient was a uniform density – that of water. Eventually,
methods were determined for accounting for the changes in density and representing them within
a dose calculation. Inhomogeneity corrections such as the tissue-air ratio method, isodose shift
method, etc., were applied sporadically within departments. As technology became more
developed and advanced, making dose modelling more accurate, the American Association of
Physicists in Medicine (AAPM) commissioned Task Group 65 (TG65)2 to review the topic of
inhomogeneity correction and assess its clinical need. The task group looked at previous studies
– one that determined that 80% of its patients studied had underestimated probabilities of
radiation damage within the lung, in the range of 5%, when inhomogeneity corrections were not
applied2 – and recommended that inhomogeneity corrections be applied in the interest of
accurately determining dose delivered to targets and organs-at-risk.

There are various algorithms within the treatment planning software that attempt to model the
dose whilst accounting for these changes in density. The most basic algorithms simply don't take
the densities into account at all - they assume the entire medium is a single, bulk density. The
most sophisticated algorithms determine the probability and simulates paths of electrons and
photons and their scattered counterparts. Monte Carlo, one of these algorithms, is considered the
most accurate algorithm to date.3

In order to better illustrate these concepts, a patient with T3N0M0 squamous cell carcinoma of
the right upper lobe was planned with and without inhomogeneity corrections. This patient was
simulated on a vacuum-bag for immobilization, with arms raised above the head, to reduce the
amount of normal tissue in the path of the beam, depending upon the eventual beam
arrangement. For simplicity, and to best illustrate the effects of densities, the patient was planned
with parallel opposed (POP) beams – one entering anteriorly, and the other entering posteriorly.
The isocenter was placed at the center of the tumor volume, rather than mid-plane as is often the
case with POP arrangements, because of the difficulties in normalizing dose in lung tissue. There
was a 2cm MLC margin applied around the tumor volume on each beam. The beams were then
weighted to ensure dose homogeneity across the beam path – as is often standard. The
prescription was set to a single fraction – 2Gy. The lowest energy was used, 6MV. The effects of
inhomogeneity correction is more pronounced at higher energies – but the calculation algorithms
we use clinically do not accurately display the effects of beam transmission through the lung. It
is also recommended to use the lowest energy possible within the clinical setting.2

The above image displays the plan without any correction. The patient is a homogenous, uniform
density – that of water. The tumor volume is outlined and shaded in pale green. The position of
the lungs are outlined in blue and yellow. The dose has been normalized to the isocenter – the
100% isodose line in red is visibly running through this point. The Dmax depth for 6MV is
approximately 1.5cm – and at both beam entry points within the patient, there are areas of higher
dose – over 107%. This is the region of electronic equilibrium. The lower isodose lines, <90%,
follow close to the beam edge throughout the treated volume. As the patient is of a uniform,
medium density, there is adequate scatter forward and laterally to maintain electronic
equilibrium. The only dose variation that we see – highlighted in the axial and sagittal views – is
with the higher isodose lines, 95% and 100%. They are bowing in centrally, and seem tilted
towards the patient’s right side. They also bow in inferiorly. This is not due to any
inhomogeneity, this is because the patient separation increases left-right and superiorly-
inferiorly. The more material a beam traverses, the more attenuated it becomes. This is similar to
the effect seen when a beam traverses high density materials. The effective path through higher
density materials is greater than the same length of lower density materials because of the
increased electron density (g/cm3). In this plan, the tumor coverage appears to be relatively good,
with 100% of the volume receiving 95% of the dose, as visualized in the dose-volume histogram
(DVH) below. The mean dose throughout the tumor volume is 100.3%.

Meanwhile, the same patient and plan is calculated with inhomogeneity correction. The
weighting of the beams has actually been changed to achieve the same dose homogeneity as
before – but differences can be seen between this plan and the plan without any heterogeneity
correction. The isodose lines are bowing in centrally through the treated volume, producing an
hourglass shape, which is a classic dose distribution for this beam and density scenario. This can
be seen clearly in the plan overview below.
This hourglass shape is produced from the reduction of lateral scatter back to the central axis as
the beam traverses low density material – i.e. the lung. The plan is still normalized at the
isocenter, which lies at the center of the tumor volume, and the 100% isodose line still runs
through this point. However, compared to the plan without heterogeneity correction, the 100%
isodose line does not extend into the lung tissue at all.

This distribution is a more accurate representation of what is really happening in low density
tissue – the range of the electrons are increased and there are less interactions and dose
deposition occurring. The secondary build-up effect mentioned earlier can be seen as the 95%
isodose line only just covers the periphery of the tumor, compared to the earlier plan, where it
extends throughout the entire treated volume. Even the effect of higher density materials can be
visualized on the plan that accounts for density changes. At the medial beam edge, the 100% and
30% isodose lines bow away slightly more from the spinal cord than is evident in the bulk-
homogeneity plan. This is because of the increased attenuation of the beam as it traverses bone –
the vertebra posteriorly, and the sternum anteriorly.
These differences can be further visualized on the DVH, above. The mean dose for the tumor in
the plan with inhomogeneity correction is 98.5% - a few percent less than the plan without any
correction. The minimum dose the tumor receives in the respective plans is 89%, compared to
94.4%. These can be better seen on the comparative DVH, below. The difference in tumor
coverage is evident. The plan without any correction (triangles) has visibly better tumor coverage
overall. However, due to the effect of density inhomogeneity, as discussed earlier, this is not an
accurate representation of the dose deposition that is actually occurring. Relying on information
from a plan that hasn’t had any inhomogeneity correction applied, leads to overestimation of
tumor coverage, and therefore tumor control.
The above comparative DVH also demonstrates that the plan without correction overestimates
the amount of dose deposition within the ipsilateral lung – the blue line. The effect of higher
density material in bone is shown in the dose reported to the spinal cord – the red line. This
reflects the bowing in of the isodose lines, as seen on the medial edge of the beam in the plan
that had a correction applied.

There are also observed differences in the number of monitor units (MUs) needed to deliver the
prescribed dose for each plan, as seen below. The total number of MUs needed for the plan with
no correction applied is 246. Meanwhile, the plan with inhomogeneity correction applied
requires 227MU – less than the plan with no correction. This is a result of the perceived
equivalent path of beams within the plan. Since the plan with no correction applied is calculated
on a water-equivalent homogenous patient, the total amount of electron density across the entire
beam path is much greater than when lung tissue is taken into account, being that the relative
electron density of lung is approximately 0.2-0.3 that of water. Therefore, more monitor units
need to be applied to the plan with no correction applied in order to deliver the same dose to the
same prescription point, as it assumes that there is greater attenuation of the beam over its path,
due to the increased density.

Inhomogeneity correction OFF Inhomogeneity correction ON

In conclusion, when comparing two equivalent plans side-by-side – one with inhomogeneity
correction applied, and one without – the dose perturbation and distribution, caused by the
presence of density changes within an actual patient, can be clearly seen. Without applying an
inhomogeneity correction of some sort, the dose to the tumor can be overestimated in lung plans,
and over- or underestimated elsewhere. This can lead to incorrect reported values for organs-at-
risk, potentially exceeding tolerance doses without being aware; and can result in insufficient
tumor control. Therefore, it is important that we use all the tools available to us in order to make
dose reporting as accurate as possible, and as close as we can model the actual processes
occurring.

References

1. Khan F, Gibbons J. Khan's The Physics Of Radiation Therapy. 5th ed. Philadelphia:
Wolters Kluwer; 2014.

2. Papanikolaou, Battista, Boyer, Kappas, Klein, Mackie, Sharpe, Van Dyk. Tissue
inhomogeneity corrections for megavoltage photon beams. AAPM Report No. 85, Task
 Group No 65 of the Radiation Therapy Committee of the American Association of
Physicists in Medicine. 2004.

3. Cherry P, Duxberry A. Practical Radiotherapy Physics and Equipment. 2nd ed. London:
Wiley Blackwell: 2009.