Dosimetric Performance of Octavius

The Reference Plane

The reference plane is the geometrical plane that represents the measurement matrix (the matrix of dose values measured with the ARRAY seven29). When the beam enters Octavius from all directions, the reference plane is the plane running through the chamber's centers.

Geometrically, the reference plane of the ARRAY is located 7.5 mm behind its surface. This is the dose plane that is usually exported from the treatment planning system for comparison.

When the ARRAY is positioned inside Octavius, the reference plane should be 8.5 mm lower than the horizontal laser marks (grooves on the phantom's side wall). Then the center of the central chamber is exactly 16 cm away from the 8 surfaces, in the geometrical center of the phantom (the phantom diameter is 32 cm).

Our ARRAY has a caliper-measured thickness of 2.17±0.01 cm. The slot of Octavius (where the ARRAY is inserted) has an inner height of 2.27±0.01 cm.

This means that if everything is perfect, there is a 1 mm air gap above the ARRAY:

Originally, our ARRAY was equipped with four rubber pads (disks with 3 cm diameter), which served as "feet" and prevented slipping of the ARRAY on flat surfaces.

To get the geometry right, we had to remove them:

(click to enlarge)

The pads were about 1 mm thick, located in 0.25 mm deep millings in the corners of the ARRAY's back side:

(click to enlarge)

(click to enlarge)

They were pushing the seven29 up inside Octavius and forced a 1 mm air gap BEHIND the ARRAY. The reference plane was therefore too high by 1 mm during the first scan:

(click to enlarge)

Compare this to the corrected scan after removal of the pads:

(click to enlarge)

It is a very small modification, though. But now geometry is fine: we can be sure that the reference plane is where the laser marks are.

NOTE: Later models of the ARRAY (ours is S/N 227) use thinner (0.15 mm) pads, so there is no need of modifying.

In the 0° setup, the vertical play of the ARRAY inside Octavius has no negative effect, thanks to gravity.

For other orientations, one can use some cardboard strips or similar to fill up the air gap.

Download Area

For those who want to perform their own evaluations with the 10x10 fields, the calculated dose matrices and the ARRAY measurements can be downloaded as ZIP-archives. The individual files can be opened with VeriSoft. The name of the file is the Gantry angle.

6MV-calculations (23MB)
6MV-measurements (200kB)

15MV-calculations (23MB)
15MV-measurements (200kB)

The 6MV, 90° field was measured with two different dose rates (300 MU/min and 600 MU/min). All other measurements are done at 600 MU/min.

The Sign of Deviations

In VeriSoft 4.0, the "Compare" window (the lower right window) can be used to display local percentage differences ("Loc%Diff") in the status bar. One has to watch out when interpreting the results.

To our understanding, a percentage dose deviation should have a positive sign if the measured dose is higher than the calculated dose. The calculated matrix serves as reference to us: if we measure less than calculated, the deviation should be negative.

Of course, it's all convention. We think that if we check an IMRT plan, we assume that the calculation is correct. When we perform IMRT verification, we strive to verify that the calculated plan is correctly delivered on the treatment machine.

With VeriSoft 4.0, PTW reversed the sign. Look at this measurement: the measured value is lower than the calculated value, but the deviation is stated positive (0.298%).

At a different location, the measurement is higher than calculation, but deviation is negative (-3.326%).

By the way: it makes no difference whether the calculation or the measurement is loaded in the upper left corner. Here, the matrices are swapped, but deviation is still -3.326%.

One has to get used to this, not only because the previous version (VeriSoft 3.2) used the opposite sign convention! If the same two matrices are loaded in VeriSoft 3.2, the deviation at LR=0 mm, TG=70 mm is stated as +3.6% (screenshot 1, screenshot 2). This is what we would expect. Measurement is higher, deviation is positive.

The VeriSoft 4.0 Manual defines the ARRAY-measurement as the reference matrix, the TPS-calculation as the comparison matrix.

Underdosage is defined as "dose in reference matrix" (= measurement) "smaller than dose in comparison matrix" (= calculation). We totally agree with that.

But it also means that starting with VeriSoft 4.0, underdosage is represented by positive deviations!

The new PTW viewpoint could mean: "A measurement is a measurement, measured dose is always correct. Therefore measurement should become the reference". A positive deviation (=underdosage) of 2% tells me: "For this IMRT plan, the planning system calculated 2% more dose than I measured".

But to swap viewpoints consistently, one should also redefine the terms underdosage and overdosage: if the measurement serves as reference, underdosage should mean: the calculated value is less than the measured value.

But this step was not taken: underdosage still means "less measured dose".

We find this (halfhearted) swapping of viewpoints with the introduction of VeriSoft 4.0 rather confusing. What is your opinion?

Literature

Ann Van Esch, Christian Clermont, Magali Devillers, Mauro Iori, Dominique P. Huyskens: On-line quality assurance of rotational radiotherapy treatment delivery by means of a 2D ion chamber array and the Octavius phantom, Med. Phys. Volume 34, Issue 10, pp. 3825-3837 (October 2007)

Martin Böttcher, Einsatzmöglichkeiten des 2D-Arrays seven29 mit Octavius-Phantom und des 2D-Arrays MatriXX Evolution mit MULTICube-Phantom für die Planverifizierung in der Tomotherapie, Diplomarbeit im Studiengang Medizintechnik, Fachhochschule Gießen-Friedberg, Juni 2009 (in german).

Most measurement systems on the market which are based on 2D-arrays and which are used for the verification of rotational therapies like RapidArc or Tomotherapy exhibit a directional depencence of detector sensitivity. Some systems perform software-compensation after capturing the Gantry angle in real-time with the help of a clinometer. In contrast to such systems, Octavius does not need Gantry readings to know where the beam comes from! The compensating air cavity in the lower half of the LINAC phantom automatically takes care of the compensation.

"Less parts" usually means "higher reliability".

Bending air*

The semicircular air cavity is usually never seen on a CT scan, because the lower part of the phantom which contains the cavity is used during ARRAY measurements only. But for the curious, here is a scan which shows the air cavity (and the chamber inserts instead of the seven29):

(click to enlarge)

The cavity starts at Gantry angle 103.2° and reaches its full width of 2 cm at 106°. With the ARRAY in place, compensation of the central chamber's signal is effective between 106° and 254°.

The ARC test

The magnitude of compensation which the air cavity provides can be calculated without performing any measurements. We constructed two semi-ARC treatment plans in Eclipse and compared them. One ARC plan ("UpperARC") runs clockwise from 270° to 90°, the other plan "treats" the lower part of the phantom, where the beam crosses the air cavity (Technically, the lower ARC had to be split into two subARCs). Dose to isocenter was calculated with 200 MU fixed, using AAA 8.6.14. Field size was 10x10 cm. The couch was not modeled, since this would spoil the comparison.

The following image shows the UpperARC plan for 6MV. Calculated dose to isocenter is 1.258 Gy for 200MU:

(click to enlarge)

A screenshot of the LowerARC (6MV) is here.

Plan (200MU)	Isocenter Dose, 6MV	Isocenter Dose, 15MV
UpperARC	1.258 Gy	1.574 Gy
LowerARC	1.338 Gy	1.642 Gy
Dose increase	+6.36%	+4.32%

Dose is higher by 6.36% and 4.32% (for 6MV and 15MV, respectively), when the beam crosses the air cavity (LowerARC plans). This is to give an idea of the amount of compensation during rotational treatments.

Static fields

Simple static AP/PA plan comparisons give slightly higher factors:

Plan (200MU)	Isocenter Dose, 6MV	Isocenter Dose, 15MV
AP (Gantry 0°)	1.274 Gy	1.587 Gy
PA (Gantry 180°)	1.374 Gy	1.671 Gy
Dose increase	+7.85%	+5.29%

Note that these were evaluations in the treatment planning system alone. The next step is to do some measurements with the seven29/Octavius combination.

How good does it work?

For each energy, 37 isocentric treatment plans were created in Eclipse. Each plan contained a single static 10x10-field (200 MU) with the isocenter in the center of the ARRAY's central chamber. Gantry angle increased from 0° to 180° in 5° steps. The Exact couch was modelled with the rails on the outer position. AAA dose calculation grid size was 1.5 mm. The dose planes through isocenter were exported as 512x512 pixel matrix (size of matrix: 27x27 cm).

The following animation shows the crossplane profile through the ARRAY center for 6MV (here is a zipped MPEG-4 QuickTime movie). Lines are the AAA calculation, dots are the ARRAY measurements. Linac dose rate was 600 MU/min. The first frame is at 0°, the last at 180°. Gantry increment between the animation frames is 5°:

The measured isocenter dose and the Eclipse dose are compared in the following table for 6MV:

(click to enlarge)

Each point is a single 200 MU measurement (no averaging over multiple measurements) with 600 MU/min dose rate. The zero level of dose deviation has no significance. During the first measurement at Gantry 0° it was tried to adjust k_User such that dose deviation is zero. To give the "Average" value in the last row some significance, one should add 0.31% to the values in the last column, thereby "normalizing" the dose deviation values to Gantry 0°.

A note on the sign in the deviation plots: positive deviations mean that measured dose is higher than calculated dose (see discussion in the side panel). Between 80° and 100°, agreement is suboptimal, as Ann would call it. The deviation is >2%. The effect of the carbon rail is visible between 130° and 140°.

As expected, the central chamber is fully compensated by the air cavity starting at 105°. At this angle, the right part of the profile still suffers from the lack of compensation:

(click to enlarge)

At 130° and 140° (but not at 135°) there are kinks in the profile:

(click to enlarge)

The kinks originate from the massive upper and lower walls of the transversed carbon rail. At 130°, the central axis is in the shadow of the upper carbon wall:

(click to enlarge)

At 135°, the beam is centered in the rail, at 140° the lower wall of the rail shadows the central axis. In all cases, the performance of Octavius and the couch modelling of Eclipse is optimal.

The results for 15MV are naturally better than for 6MV, due to higher penetration:

(click to enlarge)

Clinical Applications

We currently use the seven29/Octavius combination for the verification of fixed-field head and neck IMRT plans.

If there are abutting fields, we are usually interested in the abutment region (where the IMRT plan and the forward-planned Supra field meet):

(click to enlarge)

Here is another screenshot of the same plan. A relatively tight Gamma criterion (3mm/2%) was chosen, although the beams enter from multiple directions. Also the lateral dose profiles through the IMRT region usually agree very well.

The decreasing sensitivity of the ARRAY between ~80° and 90° (or 270° and ~280°) should not be overvalued! Here is another clinical 6MV IMRT plan where one of the beams enters at a "dangerous" 274°:

(click to enlarge)

But it is only 1 out of 5 beams - the verification result is OK.

Measurement resolution can be improved by merging two measurements which are shifted longitudinally by 5 mm. The longitudinal density of measurement points is therefore doubled:

(click to enlarge)

Conclusions

There is good agreement between AAA calculated dose planes and Octavius measurements, except for a small range of Gantry angles (~80°/85° to ~100°). Taking into account that the single field checks are much more sensitive than rotational checks like ConformalMLC or RapidArc (where the seven29/Octavius system easily passes the 2%/2mm criterion), we consider Octavius and the seven29 qualified for the verification of static and rotational treatments.

*It is generally accepted theory that air is flat ;-)

back to ARRAY main page