Claire Dempsey
Department of Radiation Oncology, University of Washington, Seattle WA, USA
Assistant Professor

Landon Wootton
Medical Physics Resident
Department of Radiation Oncology, University of Washington, Seattle WA, USA

Juergen Meyer
Associate Professor
Department of Radiation Oncology, University of Washington, Seattle WA, USA

Lori Young
Assistant Professor
Department of Radiation Oncology, University of Washington, Seattle WA, USA

Claire Dempsey
Assistant Professor
Department of Radiation Oncology, University of Washington, Seattle WA, USA

Background and Purpose: Accurate source positioning is critical in HDR brachytherapy treatments and film based tests are typically performed to verify source position. Varian provides a device to perform this test consisting of a treatment channel in plastic with metal bars embedded in the plastic corresponding to expected dwell positions. Scatter from these bars creates dark strips on film to facilitate determination of whether the source position is within tolerance. However, in practice it can be difficult to judge the exact dwell positions on the film due to blurring, introducing uncertainty and inter-observer variation. Therefore, an image analysis program utilizing cluster analysis was developed to reproducibly and objectively determine source positions on film irradiated using a Varian supplied source position testing device.

Methods: A position verification test was performed using RT-QA radiochromic film, using the Varian source position test device. 10 dwell positions 10 mm apart were delivered with a VariSource iX HDR afterloader using an Ir-192 source. The film was scanned on a typical office flatbed scanner (HP M4555P) as a 150 dpi tiff file. The program used image thresholding to identify irradiation-induced darkening and the DBScan clustering algorithm was used to identify the 10 largest clusters of dark pixels. The actual dwell position was determined by calculating the centroid location of each cluster. Then, an intensity profile was taken on a line fit to the centroids, and the expected dwell positions were determined from this profile using a peak-finding algorithm which identified the dark strips created by the metal bars in the vendor device. The average distance between dwell positions was then calculated, as well as the maximum, minimum, and average deviation between the expected and actual positions.

Results: The algorithm correctly identified the actual and expected dwell positions on the irradiated film. The average distance between actual dwell positions was 9.9 mm. All dwell positions were within 1 mm of the expected dwell position, with an average difference of 0.5 mm and a maximum difference of 0.9 mm. Visual inspection of the film validated that these results were reasonable.

Conclusions: A programmatic method of analyzing position verification films has been developed and demonstrated to work as expected. This method provides quantitative dwell position information, eliminates inter-observer variation, and does not require a specialized scanner. Quantitative information provides robust comparison between tests performed at different times and allows trends over time to be identified.


← Back to all abstracts