Objectives:

Current patient specific QA procedures for relatively small SBRT targets are inadequate. They are usually performed using either a cylindrical or planar phantom embedded with electronic detectors and utilize commercial analysis algorithms. Although efficient, these commercial devices do not provide data in the axial plane commonly most useful for clinical plan evaluation, have relatively coarse spatial resolution for small SBRT targets, provide no positional data relating results to patient anatomy and cannot be easily altered for customized calculations besides the standard Gamma Index to include for example information about a specific ROI. Commercial film techniques have excellent spatial resolution but are often limited to a single plane within a 3D dose distribution. Film calibration with multiple color channels is not standardized.
This work describes an innovative cylindrical phantom for simultaneous acquisition of dose at the phantom surface encompassing the entire treatment geometry and cumulative axial dose in a preselected clinically relevant plane. The use of film facilitates very fine resolution dose measurement. A set of algorithms has been developed for accurate dose calibration with verification, scanner correction, and versatile analysis tools such as Gamma Index in specific geometrical regions of interest.

Methods:

A cylindrical 0.93 gm/cm3 plastic phantom consists of a peripheral section with a 4 cm thick outer ring of 32 cm outer diameter containing a film insert slot at a depth of 1.5 cm. The central solid 10 cm diameter section extends 2 cm past the center of the outer cylinder and contains an axial film slot in the central plane. The phantom was scanned in 1 mm slices. Treatment plan axial doses computed with 1mm resolution were exported as 33x33 cm / 500x500 pixels for the periphery and 10x10 cm / 200x200 pixels for the center.
Gafchromic film in 10x25 cm sections was measured using the red channel only in the center of an Epson 10000XL scanner according to manufacturer recommendations at 50 dpi.
Two treatment plans were created to determine the scanner and phantom uniformity. An 8x20 cm 6MV beam was exposed at 1.5 cm depth and the calculated plan was compared to the scanner measurement.
A phantom plan consisting of multiple static fields with a common isocenter at the phantom origin point was created to evaluate the uniform phantom geometry, isocenter location, and phantom density.
Converting measured film density to delivered dose is accomplished by creating a density to dose calibration curve. A few select dose points are commonly used to establish the density to dose relationship over a wide range. A unique feature of our process is to generate a calibration plan in the phantom with the actual clinical setup, including isocenter and beam energy. A statistically large sample assures accuracy and robustness.
A fourth-degree polynomial was fit to the density-dose curve and verified by comparing the computed measured dose from the film to the treatment plan dose extracted from cylindrical verification plan.
For demonstration purposes a representative VMAT SBRT lung patient plan with two 180 arcs was created with the PTV near the brachial plexus.
The corresponding verification plan was created on the phantom with doses extracted in the periphery encompassing the entire treatment field and the clinically significant axial plane where the brachial plexus is closest to exceeding tolerance. As an added unique feature we are able to represent various plan structures such as target and brachial plexus in on the axial verification plan so that the analysis can focus on specific geometrical regions.
MATLAB with the imaging toolbox was used for computation. All scans and exported dose planes were converted to 0.5x0.5 mm pixels and registered to corresponding data sets. Global gamma index was calculated using the classical formula with 3% and 2mm criteria with 10% threshold. Given the versatility of a customizable algorithm, global and local indices can be calculated for clinically relevant areas such as the quadrant containing the brachial plexus in the axial plane. For 180-degree treatment arcs the analysis can be limited to the entrance portion of the peripheral cylinder since the relatively uniform exit dose distribution is clinically irrelevant and with global settings artificially skews gamma to a high passing rate.
Besides Gamma, the individual measured versus plan dose profiles in the periphery contain detailed information that may be useful to identify specific faults in individual control points, faults at certain gantry angles, or excess failures in high gradient sections.
As an internal QA process the registered plan dose can be plotted as a function of the corresponding film density superimposed on the known calibration curve to highlight any discrepancy. The registered plan dose can also be plotted as a function of the measured film dose. The best least square linear fit should have a slope of 1.0, intercept at 0.0, a minimal standard deviation. Deviation from these parameters is indicative of problems with the calibration curve, registration, scanner, scaling, etc.

Results:

Scanner uniformity in the central region encompassing the stated film sections was acceptable without linearity correction as noted in AAPM TG235. The slope of the ratio of the measured film density to the computed plan density for a uniform radiation field was -0.0004. Other guidelines were implemented as necessary.
AAPM TG218 recommends that spatial resolution of the evaluated dose should be no greater than 1/3 the DTA criterion, i.e. o.67 mm for 2mm DTA. Matlab image size tool was used to convert all planar data to 0.5x0.5 mm pixel size. The 1 mm slice thickness however was not interpolated.
Phantom uniformity initially was unacceptable due to small variations in the film position within the peripheral slot. Acceptable phantom uniformity was achieved by overriding the measured phantom CT density with a uniform density.
Scatter plots to assure process integrity corroborated the calibration curve. For the center plane the planned versus measured slope and intercept were 0.997 and 1.12.
The gamma index calculation was verified by comparison with a commercial product for various fields and parameters.
Gamma passing rate in the periphery for the sample case was 98.4 %. Many failures occurred along two straight lines possibly indicating film scratches or scanner artifacts warranting further investigation. The remaining failures were clustered in two low dose regions.
For the central region the gamma passing rate was 100 %. To verify this exceptional result the plan and measured doses were misregistered and as expected the passing rate decreased proportionally. Vertical and horizontal profiles comparing plan and measured doses corresponded almost exactly even in the high gradient region across the brachial plexus.

Conclusion(s):

A significant number of IROC credentialing submissions fail even though they have passed in house QA. The proposed phantom and supporting software provide simultaneous peripheral and axial data, the ability to focus on specific geometrical regions, superb resolution with statistical validity, and built-in versatility and verification. These innovative factors lead to enhanced performance that warrant the additional effort often associated with film dosimetry and custom phantoms.