2119 - ShotCaller: Custom SFRT Vertice Generation Made Easy

02:30pm - 04:00pm PT

Hall F

Screen: 5

POSTER

Presenter(s)

Arjun Karnwal, Student - CHLA, Los Angeles, CA

A. Karnwal^1,2, A. J. Olch³, H. Zhang⁴, and K. Wong³; ¹University of Southern California - Alfred E. Mann Department of Biomedical Engineering, Los Angeles, CA, ²Childrens Hospital of Los Angeles, Los Angeles, CA, ³Department of Radiation Oncology and Pediatrics, Children's Hospital Los Angeles, Keck School of Medicine of the University of Southern California, Los Angeles, CA, ⁴Department of Radiation Oncology, University of Southern California, Los Angeles, CA

Purpose/Objective(s):

Spatially fractionated radiation therapy (SFRT), also known as GRID / LATTICE radiotherapy (GRID/LATTICE RT), has been increasingly used to treat bulky and radioresistant tumors. The majority of planning time for LATTICE RT goes towards designing the optimal arrangement of spherical targets that fit the clinic's goals. Maximizing the number of spherical targets within an irregular-shaped tumor while avoiding edges and organs at risk (OARs) is time consuming and error prone. The purpose of this study is to develop and test various automated algorithms that can assist radiation oncologists and physicists with GRID / LATTICE RT treatment planning by finding the maximum arrangement of targets for a given tumor volume.

Materials/Methods: Using the Eclipse Scripting Application Programming Interface (ESAPI), a script and graphical user interface (GUI) were developed to automate target placement. The GUI has user editable parameters for key planning considerations. An alternate version was developed using MATLAB for clinics without the specific treatment planning system used.ShotCaller can produce either cylindrical or spherical targets for either GRID or LATTICE RT. The most recent algorithm uses a King Of the Hill Monte Carlo approach, with multiple user-selectable ‘biases’ to optimize sphere arrangement for various treatment goals (maximizing number of vertices, increasing distance to OARs, clustering vertices closer to tumor center, or minimizing/maximizing the number of planes for vertices). The script was retrospectively tested on tumor volumes for patients previously treated with SFRT (n=11). The associated target arrangement, time to generate targets, and target violation count were compared to identical metrics for a human operator's target arrangement. Additionally, ShotCaller was tested against freely available software solutions for SFRT sphere optimization. Among tested tumor volumes were tumors with cavities, and tumors with non-overlapping volume regions.

Results: The script was significantly faster than human staff for all cases. For a very bulky ~20 cm diameter tumor (4200cc), the script took eight minutes while it took two hours to manually place over 50 vertices, a 15x speed differential. In all cases tested, the script had zero geometric placement violations, compared to an average of 2-4 violations for the human operators. Additionally, ShotCaller was able to consistently find as many or more targets in the tumor volume compared to the manually designed target arrangements.

Conclusion: The script was proven to be a reliable, quick, and convenient method of SFRT vertice placement for a wide variety of tumor geometries. The user specific parameters and biases will allow for wide clinical application and site specific adaptation. The script, either ESAPI or MATLAB-based, will allow all clinics to streamline and standardize SFRT treatment planning, while saving valuable staff time and potentially decreasing errors in treatment planning.