AN OPEN SOURCE PROCESSING TOOL FOR KNOWLEDGE-BASED METRICS COMPUTATION

Introduction The Overlap volume histogram (OVH) is a geometric metric commonly used for knowledge based automatic planning approach. Despite its effectiveness, there is a lack of an open source software to compute it. In this work we aim to develop an open source tool to generate the OVH, and test it on a large cohort of patients. Materials & Methods A MATLAB code was developed to compute the OVH curves of organs at risk (OARs) starting from PTV expansion of 2 mm steps. The code was tested on a public imaging database, containing 85 prostate patients. OVH curves of the bladder and rectum were generated, with the prostate as potential target. Boxplots of the distances covering 20% and 50% of the volume; R20 and R50 were generated. Results Figure 1a,1. b. demonstrate the OVH curves for bladder and rectum. Figure 1.c and 1.d. evince a remarkably higher median values for the R20 and the R50 of the rectum, as compared to the bladder (2.5 vs 0.5 mm) and (5.9 vs 4.5 mm) respectively. This notable difference reflects a significant degree of overlap between this OAR and the prostate. Additionally, the R20 and R50 interquartile range for both rectum and bladder illustrate an important overlap variability between patients Summary A MATLAB code was developed and tested in this work, to automatically generate OVH curves for a large amount of data, prompting for the evaluation of geometrical variability at a multicentric level.