The Dosimetry of Exact Couch Modelling

Background

There are many dose uncertainties which contribute to radiotherapy delivery and which might result in poorer outcome. The problem is: there is hardly any data which gives reliable figures on reduced tumor control or increased side-effects as a function of dose error. Only when dose errors add to significant levels, there is a chance to see some clinical effect.

Our approach is pragmatic: if there is a way to reduce errors without much effort, we do it. And adding the couch structures to the CT data set is a matter of a few mouse clicks.

However, unreflected use of this feature, especially for the Exact Couch with its movable rails, is discouraged. One might say: "We plan the couch, therefore everything is fine." This is not always the case.

Firstly, the rails have to be at the planned positions for every single treatment fraction. There are 4 possible rail configurations: both out, both in, left out/right in, left in/right out. We never used anything other than the "both out" configuration which is also depicted in most screenshots.

Secondly, the patient has to be at the same position relative to the rails for every single treatment. However, the lateral position of a prostate patient may have a day-to-day variation of up to 4cm, even if the patient is positioned with some well-centered device like the SinMed Combi-Fix. After such a patient is centered with the help of image guidance, at certain Gantry angles the beam might still cross a rail where it shoudn't, or it might miss the rail when it should hit it, simply because it is the patient who gets aligned with isocenter, not the couchtop.

IMRT plans that consider the rails compensate the higher absorption with an increased fluence:

(click to display BEV)

To avoid the problem that a rail might not be in the beam when ist should be, the only solution is to avoid rail-planning altogether. This is what we do for all non-head & neck patients like this fixed-angle standard prostate IMRT. We delete the rails (keeping only the couchtop),

(click to show full plan)

and take care during daily tratments, that no rail is in the beam. Then we can be sure that the setup fully corresponds to the planning setup.

There is still some improvement over the previous approach, because the modeled couchtop has the full width of 53cm, whereas the manually drawn structure could only have a width of less than 50cm. Beams that hit the couch edge are now calculated correctly.

For head & neck it is totally diferent situation: the mask system is well centered and stable. Therefore, the rails can also be defined, and they stay in the beam according to the treatment plan.

Avoidance Sectors in RapidArc

During RapidArc optimization, up to two sectors may be defined where fluence delivery is reduced to zero.

(click to enlarge)

The sectors have to be at least 15° wide.

The Exact IGRT Couchtop and RapidArc

The Exact IGRT couchtop has no rails. The parameter that needs to be specified when inserting the couch structure is the couch thickness - thin:

medium:

and thick:

corresponding to the patient's longitudinal position on the couch.

Eclipse 8.6 introduced couch modelling (see also the page on the basics).

The standard procedure in our department is to insert the synthetic couch model to a patient's 3D-model whenever a treatment plan might be created that has fields going through the couchtop (see some restrictions on the right). Couch models are mostly added for head & neck, prostate, and lung patients.

The dosimetric impact can easily be assessed by comparing plans with and without couch model. Special attention shall be paid to the two movable carbon rails, which are part of Varian's Exact Couchtop.

The Flat Panel

Our machines have the Exact Couchtop installed which consists of a flat carbon panel and two movable carbon rails. If a beam enters perpendicularly (Gantry 180°) and the default HU values of Eclipse's couch model are left unaltered, the water equivalent depth of the flat panel is about 5mm:

Water equivalent depth of Exact couchtop

This is consistent with our own measurements.

The Rails

The two carbon rails are tilted inwards by 11.2°. This has been verified on our real machines with the help of a clinometer.

One can easily find the "worst case" beam setup which gives the maximum dosimetric effect: to acheive "maximum dose error" (let's say the left rail is in the beam where it shouldn't be), the Gantry angle has to be at 168.8° and the Couch moved laterally until it is aligned with the beam.

For this setup, two plans were created which deliver a fixed number of MU with a single 6MV, 10x10 field. In one plan, the couchtop and the rail is in the beam. In the other plan, the "patient" is floating in free space*. The plans were calculated with the Anisotropic Analytical Algorithm (AAA version 8.6.14).

In the following image, the plan without the couch is substracted from the plan with the couch. The maximum dose difference for this "worst case" setup is about 0.28Gy which is equivalent to a maximum (local) dose error of 28%:

(click to enlarge)

In terms of water equivalent depth, the local difference is between 5mm (flat panel alone) and more than 10cm(!) of water in the shadow of the rail's side wall:

(click to enlarge)

Two notes:

In order to see isodoses inside the couch structures, the box which limits the isodose display has to be expanded manually, using the icon. Simply expanding the isodose area with the icon will not work. This would only expand isodose display up to the Body outline. But the Couch structures are outside the Body structure.
The source surface distance (SSD) does NOT include the Couch structures. In the image, the SSD is 95.5cm in both plans. This is the distance from focus to the Body surface.

Dose display has nothing to do with dose calculation. The absorption of the Couch structures are always taken into account, whether isodoses are displayed or not. If you are not interested in seeing isodoses inside the Couch, you do not have to do anything.

The Exact Couchtop and RapidArc

If the Exact couch with rails is installed on the treatment machine, dose planning should include the rails., which is done automatically during optimization and during dose calculation. Of course, this only works if the rails are not deleted manually!

During a full 360° RapidArc treatment, the rails are in the beam twice:

With a Flash player installed, you would see the movie RapidArc1.swf instead of this boring line.

If you still feel uncomfortable beaming through the rails, you can use Eclipse's feature of avoidance sectors and reduce fluence delivery to zero while the rails are in the beam ...

With a Flash player installed, you would see the movie RapidArc2.swf instead of this boring line.

The MLC leaves are not displayed, only the MLC aperture and the jaws, for better clarity.

If the movies appear out of sync, reload the page. If you are interested in a full-quality stand-alone movies, you may download the AVIs here and here. There is also a version of the first movie with MLC.

What is the potential dose error during RapidArc?

Let's assume the rails were at the OUT position during planning but at the IN position during treatment.

The optimized plan (45A = "A") will assume the rails OUT. A copy of this plan, (45B = "B"), was recalculated without reoptimization on a structure set with the rails IN. The plan B therefore corresponds to the wrong treatment setup. The difference of the two plans, B-A, was calculated. The dose profile through the difference plan starts behind the rectum and crosses the PTV and the bladder:

(click to enlarge)

Negative values (in the rectum) are under-, positive values (in the PTV) are overdosed. The amount of overdosage (up to 0.7Gy) has to be seen in relation to the prescribed dose of 70Gy. This gives a maximum error inside the PTV of about 1%, if the rails are in the wrong position during the treatment.

The negligible impact can also be seen in the DVH:

(click to enlarge)

Except for a slightly better sparing of the rectum, the total effect is small. One could even say that the plan with the wrong rails is the better plan (maybe we should treat all prostate patients this way?).

Conclusions

The rails should not be ignored for fixed beam techniques like IMRT, especially if beams pass through them and the number of fields is small.

During RapidArc, since there are no fixed beam angles, the percentage of MU where the rails are in the beam is generally much lower than during fixed-angle IMRT. For a typical 360° RapidArc prostate treatment with 15MV photons, the effect of wrong rail positioning is therefore very small.

This is a clear advantage of RapidArc over fixed-angle IMRT.

*In German, there is a technical device called Siemens Lufthaken, which helps stabilize free-floating structures from being attracted gravitationally.