Development of a Stand Alone Comprehensive Collision Prediction System for Modern Radiation Therapy
MetadataShow full item record
Purpose: Studies exploring the use of non-coplanar radiotherapy involving unconventional combinations of treatment couch and gantry orientations have demonstrated significant dosimetric benefits. However, clinical implementation of optimum non-coplanar radiotherapy has not become a routine practice mainly due to the absence of a comprehensive collision avoidance tool. Without such utility, during treatment delivery, the moving mechanical components of the Linear Accelerator (Linac) may collide with the patient, patient support system and treatment accessories. In this work, we developed and evaluated a stand-alone collision prediction system (CPS), which can fit in a typical clinical workflow. Methods: The CPS is based on a geometric model of the Linac and patient morphology acquired by a 3-D vision camera (Microsoft, Kinect). Physical dimensions of the mechanical components of the Linac (Varian TrueBeam) were quantified to construct a prototype geometric model with the isocentre as the origin of the co-ordinate system. Other components of the Linac model include the treatment couch, gantry, kV source, and both MV and kV imaging panels. A combination of contours from the planning CT and the 3-D depth data acquired with the Kinect camera were used to develop a patient specific model. The Linac couch coordinate was estimated based on a correspondence map between the CT couch top and the choice of the treatment isocenter coordinates. An algorithm that relies on a dot product between vectors denoting points in Linac components and patient morphology is used to predict collisions. Virtual collision test cases were calculated with our CPS and experimentally verified in the treatment room using an ArcCheck and Rando phantoms to simulate a patient.Results: For a set of 111 collision test cases, the sensitivity and specificity of the CPS model were calculated to be 0.95 and 1.00, respectively. Validation experiments demonstrate that the CPS consistently predicts collisions prior to actual collision locations. The average difference between the predicted and measured collision states was 2.3 cm for lateral couch movements. The predicted couch rotational position for a collision between the gantry and a patient analog differed from actual values on average by 3.8 degrees. An end to end testing simulating CPS usage during a typical clinical workflow had a prediction error on average by 3.4 degrees with a standard deviation of 1.3 degrees.Conclusion: This study outlines a framework for the development and clinical implementation of a solution to collision management challenges during clinical treatment planning. The CPS can be a valuable tool for treatment planners who seek to implement non-coplanar radiotherapy into routine clinical practice.