Using EPIQA for the verification of FFF-RapidArc Plans on the TrueBeam

Gaussian Smoothing?

The controversial issue of Gaussian convolution (GC) was already discussed elsewhere, where the MLC leaf width had been 1 cm. GC of the measured images helped to improve the agreement between calculated and measured distributions. On the 80 leaf MLC, we often used Sigmas of 2 to 2.5 mm in order to get good results.

Our TrueBeam STx has a central leaf width of 2.5 mm, which makes GC less important. However, the measured images show high spatial frequencies which simply cannot be found in the calculated images, even if the sharpest profiles are fed into AAA 10.0.28.

One reason is different depth of calculation and measurement. Eclipse dose is usually calulated in dmax (14 to 27 mm, depending on energy), but measurement depth in the detector is a fixed 8 mm.

Another justification for some smoothing is the discrepancy between the relatively large Eclipse calculation grid size (0.125 cm for EPIQA evaluations) and the small width of the interleaf-leakage tongues and grooves that can be assessed with film or the EPID. Eclipse simply cannot model them accurately enough (yet).

This is the "licence to smooth".

How much smoothing is OK?

Here is an example of a 10X arc field, evaluated using the 2%/2mm Gamma criterion.

First, unsmoothed:

The inplane profiles show that aS1000 has higher frequencies than Eclipse:

Now the same for Gaussian convolution with a radius of 1 mm,

1.5 mm,

and 2 mm:

Up to 1.5 mm, Y profiles get closer and closer regarding spatial frequency (visually more "similar" regarding the peaks and valleys). But at 2 mm, there are peaks and valleys in the Eclipse calculation that cannot be found in the measured image!

This is the point where the measured image was smoothed too much.

We found that for our system (HD120, AAA 10.0.28, EPIQA 3.0) the optimal smoothing radius is 0 to 1 mm for 6FFF, 1.5 mm for 6X, 2 mm for 10X, and 2.2 to 2.5 mm for 15X.

Vienna, October 24, 2012 - Today KFJ successfully started Portal Dosimetry of FFF-Beams on the TrueBeam, using the highest available dose rates and EPIQA version 3.0.5.

Click all images to enlarge:

(EPIQA evaluation of 10FFF-RapidArc test plan with a Gamma criterion of 3%/3mm and active Jaw Tracking. Dose Rate was 2400 MU/min. See also the evaluation of the first clinical FFF-plan treated at KFJ, which can be found here.)

While Varian Portal Dosimetry is straightforward for flattened beams, the unflattened energies of the TrueBeam (6FFF and 10FFF) are currently not supported by Varian.

Epidos (again) is one step ahead and offers a solution to this problem with Version 3.0 of EPIQA, which supports all 5 Photon energies of the TrueBeam (6X, 6FFF, 10X, 10FFF, 15X).

Configuring the FFF Beams in EPIQA

The configuration steps for flattend and unflattend beams are very similar. The only difference is that for the highest dose rates (2400 MU/min for 10FFF), the source detector distance for the configuration should be moved from 100 cm to 150 cm. This avoids saturating the Portal Imager during image aquisition.

As for flattened beams, the first configuration step is to provide ionization chamber measurements ("output factors"). We preferred to calculate these factors in Eclipse instead of measuring them. The following screenshot is for 10FFF, 30x30 cm:

The phantom is at SSD = 147.8 cm, which is 150 cm - dmax for 10FFF. The dose prescription is 100 Gy, this gives an output factor of 207.37 MU/Gy for this field size.

The second step is the aquisition of integrated images for the same series of field sizes ranging from 3x3 cm to 30x30 cm, with the detector positioned at 150 cm. This completes the configuration of the primary beam.

Here are our results for 6FFF:

and 10FFF:

Four more images are acquired for the configuration of the MLC-transmitted beam. The measured MLC transmission also has to be entered in EPIQA configuration software. We used the same values as for the "Dosimetric Data" of Eclipse Beam Configuration: 1.0% for 6FFF and 1.2% for 10FFF.

As a last step, since the TrueBeam does not allow to acquire "raw" images, the dose profile for the same flattened energy (6X in the case of 6FFF, 10X for 10FFF) has to be provided, and also 2D-dose distributions for a 40x30 cm flattened beam and a 30x30 cm unflattend beam ("Gain correction"). These are calculated in Eclipse at the same distances as above: 150 cm - dmax.

Conclusion: FFF-RapidArc-QA solved!