Let's Play Snooker!

Literature

Ann Van Esch et al., Implementing RapidArc into clinical routine: A comprehensive program from machine QA to TPS validation and patient QA. Med. Phys. 38 (9), September 2011.

Plan Generation

Compared to IMRT plans, artifical RapidArc plans are a little tricky to construct. A simple text editor is not sufficient any more, because the MLC movement is coded in the RT Plan file. A DICOM editor is needed.

We start by building a phantom in Eclipse,

and define two target structures (high resolution segments) on the surface. This will be used to fit the MLC later on. The red and blue tubes have 6 mm diameter, and are aligned with the Beekley spots on the real phantom.

Before diving into DICOM, we start with a conformal MLC arc field. (The full use of the phantom can be acheived by setting up four arc fields in the four quarters of Gantry angle.)

We add an MLC, set the technique to Arc Dynamic, and change the number of control points to the desired value.

We also change the energy from the default 6X to 10FFF, because this time we want to test the highest dose rate.

The next window shows the MLC properties of the conformal MLC arc field:

It will look quite different after the field will be converted to a VMAT field with the help of the DICOM editor.

This gives a good base for later DICOM editing.

Final adjustments of jaw size, dose prescription etc. should be made before export, because it is usually easier to find the fields which need to be changed in Eclipse. However, changes can also be made with the editor or after re-import.

The red and blue target structures may be used to fit the MLC. The 180 -> 90 arc field focusses on the red target:

However, we will later edit all leaf positions in the file.

Note that dose calculations of artifical machine QA plans require some adjustment of local DCF settings, if the control point angle interval is too large:

If you override the local DCF settings to enable calculation, don't foreget to undo this before calculating the next clinical plan.

This is how the calculated plan looks right now, using a coarse angular resolution in conformal arc calculations of 5°:

We now export the plan and open the RP-file with the editor. We use DicomEdit 1.7:

Most of the DICOM tags speak for themselves. We therefore only add some notes.

Regarding MLC tag editing, we prefer to click the "Edit in grid ..." button,

and change the default display

to 60 columns without Row order.

The advantage is that now the Import/Export buttons can be used to transfer leaf positions between Excel and the current DICOM tag via *.csv files. This makes leaf editing easy.

Meterset weights have to be edited too.

As a final step, some IDs need to be changed, otherwise Eclipse will refuse to import the plan.

When an edited plan is reimported into ARIA, a warning is issued, because plan integrity check fails. This can be ignored.

This is the delivery of a 10FFF plan (the short bursts of radiation at 2400MU/min really sound impressive):

RapidArc machine QA should be as simple and efficient as possible. The Snooker Cue test was "designed to allow a quick routine assessment of the correct interplay between gantry angle, MLC position, and dose delivery in one single treatment plan by means of the EPID." (Ann Van Esch, 2011).

We set up a dedicated RapidArc plan, built the necessary phantom, and performed the test on our TrueBeam:

If everything goes as planned, small dots are visible in each of the vertical stripes. The planned stripe width is 6 mm.

This is a screenshot halfway into delivery of the first out of of four arc fields:

(you are right: the MV imager asks for a calibration of the dosimetry modes ...)

The Phantom (Revision 1)

The small dots on the integrated MV images are radioopaque markers. We tested the suitability of different marker types (lead, steel) and diameters. One possibility is to use Beekley spots made of steel (Y-Spots/2mm, V-Spots/2.5mm, Z-Spots/4.3mm).

The phantom we used for MV imaging contains five Beekley Y-Spots (2mm diameter), which were placed on a scaled printout. The outer square of the printout is 10 x 10 cm.

The sole purpose of the block of radiation transparent styrofoam is to raise the BBs above table surface, so that the couchtop cannot disturb the MV images at oblique beam incidence.

The next photo gives an idea of the setup. When the Gantry is at 0°, SSD is 90 cm. It is best to align the BBs with the horizontal side lasers first, and then raise the couch by 10 cm. In our case, the block is around 8 cm thick, so isocenter will be below couch surface.

The test is very demanding. Small deviations in RapidArc delivery can happen, even on a TrueBeam ;-)

The above animation blends two successive deliveries of the first RapidArc field. While the first three stripes are at constant positions relative to the markers, the rightmost stripe wiggles around a little. Here is a zoom:

Portal dosimetry workspace can be used to quantify the shift. We are interested in a horizontal (x-) profiles across the images, which have four peaks:

This is a zoom into the rightmost of the four peaks:

The central profile depression is from the marker. Although it is not possible to measure distances directly in this graph, by looking at the horizontal axis one can see that the profiles are shifted by approximately 0.3 mm.

Discussion

The radioopaque markers serve as landmarks in space¹, just like the tumor, which has to be hit with high precision. In this sense, the Snooker Cue test is very realistic. But while clinical plans rarely push the envelope of the machine, the Snooker arcs require high values of gantry acceleration and deceleration, high and low values of gantry speed (0.5 deg/sec during delivery, 6.0 deg/sec while beam is on hold) and high and low values of MU/deg. Only the correct interplay of all components will provide successful delivery.

The result is immediately visible on the images, due to the markers.

Revision 1 of our phantom serves as a proof of principle. A Revision 2 model would have to be made with higher precision. Small deviations in marker placement could be misinterpreted as delivery errors. Laser misalignment or simple setup errors are other possible sources of error.

Notes

¹This makes a big difference to "EPID only" delivieries without phantom, such as Varian RapidArc commissioning tests, where there is no relation to fixed points in space.