Triple VMAT Pretreatment Verification

Configuration

To configure DIAMOND, certain basic data have to be measured in the water phantom and in a mini phantom.

To have a starting point, DIAMOND comes with an extensive library of preconfigured machines, which were collected from other users (see the left column in the following screenshots).

To configure an own machine(s), one has to provide several input data.

The user can choose to provide either percentage depth dose or TPR data. One is converted into the other by DIAMOND. We imported depth dose curves:

DIAMOND also needs off-axis factors,

scatter factors (measured in the ESTRO mini phantom),

MLC transmission,

and some more.

Before purchasing DIAMOND, we had concerns that HD120 MLC modelling might not be possible. But the MLC configuration is very flexible: it was no problem to define the HD120 MLC.

Regarding dosimetric MLC properties, some mechanisms are provided which help to optimize the agreement between measured and calculated leaf edge profiles. The next two screenshots show for 6 MV and a 4x4 cm field the measured (blue) and DIAMOND calculated (red) profiles across the rounded MLC edge. In iteration 2,

calculation inside the field and close to the edge is still a little too high compared to the PinPoint measured curve. Outside the field, calculation is a little too low.

In iteration 3, some parameters in the MLC model have been changed:

Now the agreement is much better, both inside the field and outside.

Regarding wedges, no extra beam data has to be measured, if only EDW is used (this is our situation), and if DIAMOND is configured to use the "Direct Calculation method". If the user chooses to use measured data for the calculations, this data has to be provided.

If Machine ID matches the ID in ARIA, Diamond will automatically choose the correct Machine data for the calculation.

We have two identical TrueBeams. Therefore, only one machine was configured in DIAMOND in the first place, and validated. This machine was then duplicated and the ID changed.

DIAMOND can be used on a network, so that multiple users can work at the same time. We have all patient and machine data on a central network drive.

Validation

PTW provides a test patient (validation patient) who can be used to validate DIAMOND configuration.

The validation patient contains a series of test plans. The plans contain a series of static fields, which are calculated and compared with measurements.

Our "Cal" plan for instance contains for each of our five energies five fields with sizes from 5x5 cm to 40x40 cm. DIAMOND calculates the number of Monitor Units needed to deliver 100 cGy to the ACP.

The results can be conveniently printed in tabular form, and compared to measurements.

Similar validation plans exist for scatter factors (ScSp), off axis factors, wedges, etc.

All validation plans can be modified by the user, according to his needs.

With the introduction of PTW DIAMOND software (version 5.3.2), KFJ broadened its VMAT QA spectrum: Now most VMAT plans¹ are verified in three ways before the first clinical delivery: with EPIQA portal dosimetry, Varian Portal Dosimetry, and finally with a DIAMOND calculation².

In the first 14 months after going clinical with TrueBeam, over 520 arcs were verified using one or both Portal Dosimetry methods (see the results for the first 313 arcs here). Since March 2014, DIAMOND is routinely used. "Older" arcs are retrospectively analyzed, to evaluate the performance of the software.

(Left: 10FFF, single arc VMAT Eclipse plan to be verified. Right: DIAMOND calculation result³.)

Introducing DIAMOND

The software can be used for all kinds of plans, static, IMRT (sliding window), and VMAT. Here, we only focus on VMAT calculations.

VMAT Calculation Workflow

To check a VMAT plan, we do the following: first, we load the plan in Eclipse. In the example, it is a rectum VMAT plan with a single 10MV arc. We right-click on the plan, select Export > Wizard ... and then check "Include structure set" and uncheck "Include setup fields" because setup fields have no associated dose. The structure set is checked because DIAMOND needs it for the calculation of depth and densities.

In DIAMOND Patient View, we click on the Import button, select the patient, and select both the plan file and the structure set file. The plan file starts with RP, the structure set with RS. The we switch to Plan View to perform the calculation. From the ROI window that pops up, we only select BODY for calc. "For display" is selected automatically. When we switch to Plan View, the uncalculated plan is shown.

DIAMOND calculates dose in a single point. This is called the ACP, the Active Control Point. If the plan in Eclipse would contain a Primary Reference Point, the coordinates of this point would be contained in the exported plan, and DIAMOND would use this point as ACP. But as we have not assigned a location to the reference point in Eclipse (dose is normalized to Target Mean), DIAMOND will calculate dose in isocenter. In our situation, we therefore have to measure the isocenter dose in Eclipse using the Point Dose Tool⁴.

This dose (49 Gy), divided by the number of fractions (25), is the dose per fraction (1.96 Gy). In DIAMOND we click on Edit Plan Data, and replace the nominal fraction dose (200 cGy) by the isocenter dose (196 cGy). Now we can click on "Calculate Plan" (or "Calculate Field", which is the same for a single arc plan), and wait for the calculation result.

In this example, the calculated isocenter dose is a little less (194.36 cGy) than the Eclipse reference dose (196 cGy). DIAMOND gives the deviation %Diff Calc/Plan with positive deviation (+0.84%) which is a little odd, since Calc (194.36 cGy) is less than Plan (196 cGy). At least the signs can be configured to some extend. For now, we are happy with the result, and always look at the dose values themselves:

This was the typical workflow for most VMAT plans. Head and Neck plans sometimes require one more step on the Eclipse side, at least for us. Since we include all immobilization devices in the Body outline, this means that at isocenter the Body outline is not always a single, closed loop, but sometimes contains two closed loops: the patient plus the carbon base plate of the head support. DIAMOND is not happy with that. It needs a single, closed loop at the Z coordinate of the ACP.

The case is easy to handle. In Eclipse, we first duplicate the Structure Set, then create a Dummy structure that forms a small 3D bridge between the two Body parts that shall be connected (VOI, CT-ranger from -1000 HU to 0 HU, Boolean Algebra: Body = Body OR Dummy). Then we assign this Structure Set to a plan copy of the VMAT plan to be verified. This plan copy is then exported to DIAMOND. Important: it is not necessary to calculate 3D dose of this plan in Eclipse! The reference dose at isocenter is taken from the original plan, which stays untouched.

DIAMOND RS Contours Manager shows the modified Body outline. Note the small vertical bridge in the lower part of the image, that connects the patient ant the support.

With such a workflow, also the results for head and neck VMAT plans are usually very good:

Challenges

Verification results are not always satisfactory, and sometimes no plausible reason for the deviation can be found. It is easier to guess the reason if large inhomogeneities are present (lung!): DIAMOND calculates in water, and in the current release (5.3.2) the software knows nothing about the density distribution inside the Body outline.

This is the uncorrected calculation result for a plan with two 10MV arcs (going from 181° to 30° and back):

Large portions of the right lung are penetrated by the beam. Deviations are in the range of 6 to 8%. Since calculated dose is lower than plan dose, it can be assumed that the deviation mainly comes from the fact that DIAMOND does not know about the low density lung.

The above screenshot also shows the Eclipse WED tool ("Water Equivalent Depth for Field"). For both arcs, the measured WED of isocenter is 9.84 cm in Eclipse. This is much less than the geometrical depth.

The information can be used to apply a first order correction in DIAMOND. For this, we right-click in Plan View on the IsoCenter point, select "Edit All Points", and get this window. We check that "Enter SSD & Depth Manually" is selected, enter the WED from Eclipse, highlight all segments (could also be a subset), and hit Apply. Then we recalculate the arc, and repeat the procedure for the second arc. The results now look much better:

Conclusions

DIAMOND proved to be a valuable QA tool for VMAT plans, and also for static and IMRT plans (not covered here). As a point dose calculation tool, it has its limitations. However, it is imaginable that under certain conditions, like limited access to the treatment machine, it can replace pretreatment QA measurements. For us, it is currently an additional tool to verify the plans, giving triple confidence that plan delivery is correct⁵.

Notes

¹Since October 2012, VMAT arcs with FFF energies (6FFF, 10FFF) and full dose rate can be verified with EPIQA, but still not with Varian Portal Dosimetry. Although TrueBeam 2.0 now offers separate image calibration of FFF modes (TrueBeam 1.6 simply used the 10MV calibration to acquire integrated images with 10FFF), the necessary Eclipse Portal Dose Image Prediction (PDIP) is still not possible in Eclipse 11.
² Our fourth method, measurement with the 2D-Array OCTAVIUS 729, is used on a case-by-case basis. For very large fields, where the IDU20 sometimes has problems, satisfactory results can often only acheived with the 2D-Array.
³The use of DIAMOND for flattening-filter free energies is not yet approved by PTW.
⁴We assume AAA. If plans with more than one arc are calculated with AcurosXB, Plan Dose Calculation should be OFF to use this workflow. Otherwise, the arc contributions to isocenter are not known. Of course, using a Primary Reference Point with a location always works.
⁵This is not to be confused with plan quality. Even a suboptimal or poor plan can be verified with excellent results.