Development of 3D Slicer based film dosimetry analysis

Abstract

Radiochromic film dosimetry has been widely adopted in the clinic as it is a convenient option for dose measurement and verification. Film dosimetry analysis is typically performed using expensive commercial software, or custom made scripts in Matlab. However, common clinical film analysis software is not transparent regarding what corrections/optimizations are running behind the scenes. In this work, an extension to the open-source medical imaging platform 3D Slicer was developed and implemented in our centre for film dosimetry analysis. This extension streamlines importing treatment planning system dose and film imaging data, film calibration, registration, and comparison of 2D dose distributions, enabling greater accessibility to film analysis and higher reliability.

Full text

Development of 3D Slicer Based Film Dosimetry Analysis
KM Alexander1, A Robinson2, C Pinter2, G Fichtinger2 and LJ Schreiner1,3
1 Department of Physics, Queen’s University, Kingston, Ontario, Canada K7L 3N6
2 School of Computing, Queen’s University, Kingston, Ontario, Canada, K7L 3N6
3 Department of Medical Physics, Cancer Centre of Southeastern Ontario at Kingston
General Hospital, Kingston, Ontario, Canada, K7L 5P9
E-mail: kevin.alexander@krcc.on.ca
Abstract. Radiochromic film dosimetry has been widely adopted in the clinic as it is a
convenient option for dose measurement and verification. Film dosimetry analysis is typically
performed using expensive commercial software, or custom made scripts in Matlab. However,
common clinical film analysis software is not transparent regarding what
corrections/optimizations are running behind the scenes. In this work, an extension to the open-
source medical imaging platform 3D Slicer was developed and implemented in our centre for
film dosimetry analysis. This extension streamlines importing treatment planning system dose
and film imaging data, film calibration, registration, and comparison of 2D dose distributions,
enabling greater accessibility to film analysis and higher reliability.
1. Introduction
Dose delivery validation in modern clinical radiation therapy (RT) is difficult, as treatment techniques
have become increasingly complex. This has led to the need for high-resolution detectors to accurately
measure radiation dose distributions. One convenient clinical dosimetry option is radiochromic film as
it is near tissue equivalent, energy and dose rate independent, and is an integrating dosimeter. Film
dosimeter use is widely published and has been used for quality assurance, small field dose
measurements, and as part of the commissioning of new machines and treatment techniques. At the
last IC3DDose conference, we presented new, open-source 3D gel dosimetry analysis software, which
streamlined gel dosimeter analysis from hours to 5-10 minutes [1]. Since then, we have recognized
that an open-source film dosimetry analysis software would also be a useful clinical tool for film
dosimetry. This work presents the initial development and implementation of such software.
2. Materials and Methods
In our clinic, film dosimetry analysis was previously performed using commercial software which was
expensive and which performed corrections and optimizations (data smoothing, non-uniformity
corrections), sometimes without the user’s knowledge. To overcome these challenges and software
license costs, a streamlined user interface for a 2D film dosimetry workflow has been implemented in
3D Slicer, as an extension called a slicelet. 3D Slicer is a medical software platform with advanced
graphics rendering, segmentation, and registration capabilities. 3D Slicer is a good fit for film
dosimetry analysis, as it is a free, open-source and customizable computational tool used for image
analysis and visualization. The film dosimetry analysis slicelet draws on tools developed for radio-

Figure 1. The workflow of the Film Dosimetry Analysis slicelet showing the planning data (upper
left), film imaging data (top), film calibration (right), and dose comparison (bottom left).
therapy called SlicerRT [2] (www.slicerrt.org), which contains a multitude of RT-specific features,
such as loading of DICOM-RT data, computation and display of dose-volume histograms,
manipulation of structures, dose comparisons, and dose distribution visualization.
In this work, we present the initial development, implementation, and testing of the Film Dosimetry
Analysis slicelet, which follows the general workflow shown in figure 1. The testing was performed
using Gafchromic EBT3 film (International Specialty Products (ISP) Corporation, Wayne, NJ)
coupled with our custom built, in-house CCD lightbox film dosimetry imaging system, using single
red channel dosimetry methods. Using the CCD lightbox film imaging system with a diffuse red light
field, flood field images are acquired without any film placed on the light field surface to provide a
reference against which to normalize the image of the irradiated film. The irradiated film images are
acquired with the film of interest placed flat on the light field surface. Using these images, the optical
density for each pixel of each film is calculated. A sample film dosimetry case taken from our clinic’s
commissioning work on a new fractionated stereotactic radiation therapy (FSRT) brain metastases
treatment is presented to demonstrate the functionality of the slicelet.
2.1. Calibration
In order to calibrate the film for dosimetry, a relationship between the optical density and the absorbed
dose during irradiation is determined. This is performed by irradiating several films to known doses
under well-defined conditions (e.g. 10 x 10 cm2 fields) in order to establish the film response model.
These calibration films are imaged and a flood field image is acquired. The images of these films and
the flood field are then imported into the slicelet and assigned their appropriate dose value, or assigned
as the flood field image. For example, in this case calibration doses of 0, 50, 100, 200, 300, 400, and
500 cGy were loaded along with a flood field image (figure 2a). Next, a region of interest is chosen,
which defines a uniform region of the calibration films (figure 2b) and an average optical density
value is calculated within that region for each calibration film. These corresponding optical density
and dose values are then plotted and fit using the following function [4] (figure 2c):
Dose = a + b·OD +c·ODn,
where fit parameters a, b, c are constants determined by a least squares fitting function, and n is
limited to the range of [1.00, 4.00]. When this calibration routine is complete, the fit parameters
automatically populate the fields of the slicelet. Fit parameters can be saved to a text file for future
use.

(a)
(b)
(c)
Figure 2. a) A screenshot of the graphical user interface of the calibration tab in the 3D Slicer Film
Dosimetry Analysis slicelet. b) A region of interest drawn about the central, uniform region of a 600
cGy irradiated film (10 x 10 cm2 field size). c) Calibration curve generated using the slicelet and
fitted using the n-power polynomial function.
2.2. Load experimental data and apply calibration
The experimental film for this test case is a VMAT FSRT brain metastases treatment with 4 Gy
prescribed to each of the four metastases (varying in size from 8-20 mm in diameter), and a 2 Gy
whole brain dose. In this step of the slicelet, the DICOM dose and structures, and the experimental
film image is also imported. The imported experimental film image, flood field image, and the
DICOM dose volume are then assigned their respective roles, and the film resolution is input, in units
of mm/pixel (figure 3a). The calibration function is then applied to the experimental film to convert to
dose (figure 3b).
2.3. Film registration
In order to register the film measurement to the calculated dose plane, the coronal plane at isocenter is
extracted from the DICOM dose volume. The experimental film is also resized from pixels to distance
in millimetres. The experimental film is then registered using a rigid registration built-in to the
BRAINSfit registration module in 3D Slicer (figure 4a).
(a)
(b)
Figure 3. a) Screenshot showing the ‘load experimental data’ options and an image of the
experimental film. b) Screenshot of the ‘apply calibration’ step, where fit parameters appear after
performing the calibration step (step 1). These values can also be manually input.

(a)
(b)
(c)
Figure 4. a) Film (green) after completing a rigid registration to the planned dose (grey) and
structure set. b) Dose profiles about the registered experimental film (orange) and Eclipse (blue)
dose distributions, as indicated by the line drawn in figure 4a. c) Screenshot of the user interface for
the gamma comparison tool, and plane of gamma value calculated within the brain structure.
3. Results and Conclusions
After calibrating and registering the film measurement, the final step in the film dosimetry analysis
slicelet is to compare that measurement to the planned dose distribution by using a 2D gamma dose
comparison (figure 4c). The gamma dose comparison tool allows the user to set the distance to
agreement and dose difference, as well as the option to only calculate the gamma values above a
minimum dose threshold, or within a particular structure. In this example, excellent agreement
between the measured and planned dose distributions was shown, resulting in a 3%/3mm gamma pass
rate within the brain structure of 99.3%. Dose profiles also show excellent agreement (figure 4b).
Overall, the film dosimetry analysis slicelet streamlines complex film analysis; another example of
its use can be found in a clinical example reported in this conference’s proceedings [3]. This open-
source analysis tool helps make film dosimetry more accessible and transparent. It also enables
convenient film dosimetry, free from the constraint of commercial analysis or read out systems. Future
work will look at implementing multichannel dosimetry methods in the slicelet. This poster, along
with a workshop during the meeting, will show how 3D Slicer can facilitate efficient 2D and 3D
dosimetry.
Acknowledgements
Research funding for this work has been provided by the Canadian Institutes of Health Research
(CIHR) and the Ontario Consortium for Adaptive Interventions in Radiation Oncology (OCAIRO).
References
[1] Alexander KM et al. 2015 J. Phys.: Conf. Ser. 573 012042.
[2] Pinter C et al. 2012 Med. Phys. 39 6332-7
[3] Alexander KM et al. 2016 J. Phys.: Conf. Ser. (accepted to IC3DDose 2016)
[4] Devic S et al. 2004 Med. Phys. 31 2392-2401