Clinical Implementation of an In-House Developed Plan Quality Control Tool to Guarantee High Quality Automatically Generated Radiation Therapy Plans for Prostate Cancer Patients

Abstract

In our institute all prostate radiotherapy plans are automatically generated with a single treatment technique template in an auto-planning module on treatment planning software. Since the planning process is fully automatic it is more difficult for the planner to decide if the plan is optimal or needs improvement. Currently, a scorecard is used to evaluate if a plan meets the clinical dose criteria. However, this scorecard only represents dose criteria for the population, and does not guarantee an optimal plan for each individual patient. We developed an in-house plan quality control (QC) tool to check if the automated prostate treatment plans are optimal or need further improvement. This tool predicts a personalized dose-volume histogram (DVH) for each organ-at-risk (OAR) based on the anatomy of each individual patient. All prostate plans have one single VMAT arc (95 to 265°), by which 70 Gy is given in 28 fractions of 2.5 Gy. Historical data consisting of 129 automated prostate plans which all fulfill the clinical dose criteria are used in this study. The model is trained on 100 plans and validated on a separate test set of 29 plans based on a random split. The model is made by first calculating principal components (PCs) of the DVH and overlapping volume histogram (OVH), which is a measure to capture individual patient anatomical information. Second, the calculated PCs are used to fit a support vector regression model that predicts DVH PCs from OVH PCs. The plan QC tool is implemented in our clinical workflow. The plans are exported from the TPS in DICOM format and sent to a PACS that triggers an application. The application automatically calculates the predicted and planned DVH curves and generates a personalized scorecard in a PDF report. The report is added to the departmental oncology information system (OIS) and reviewed by the planner. Threshold levels of 3Gy for dose and 3% for volume metrics are set to detect if a plan needs to be replanned. Plans are replanned by changing the treatment technique template for each of the OAR dose parameters that can be improved and start the Auto-Planning again. All plans in the test dataset were checked by the plan QC tool. For the 29 plans, 8 plans had OAR dose parameters that deviated more than the 3Gy and 3% thresholds from the predicted values. These 8 plans were further optimized after which the DVH metrics were within the threshold levels with respect to the predicted DVH metrics. The mean improvement is 4,2Gy for mean dose and 4,0% for V30Gy and V60Gy of the rectal and anal wall. An independent plan QC tool is necessary to make sure that automated prostate treatment plans are optimal for each individual patient. This tool generates a personalized scorecard for each plan and can be safely implemented in the clinic to guarantee high quality automated prostate plans, and thus optimize and individualize patient treatment.