An effective calculation method for an overlap volume histogram descriptor and its application in IMRT plan retrieval

To effectively calculate an overlap volume histogram (OVH) descriptor and improve intensity modulated radiation treatment (IMRT) planning by basing it on previous plans with similar features, a method based on morphology for OVH calculation was proposed and a novel similarity measurement was employed for retrieval of a suitable IMRT plan. First, the minimum and maximum distances between the tumor and organs at risk (OARs) were calculated as the start and end points for contraction or expansion, and a suitable step size for contraction or expansion was determined according to these distances. Then, a dilation or erosion morphology operator was employed to compute the OVH descriptor. Finally, the performance of IMRT plan retrieval was evaluated, where the area between OVH descriptors was taken as the similarity measurement, and a 3D reconstruction for each case was also performed for visual comparison. Twenty-eight nasopharyngeal carcinoma (NPC) cases were evaluated. The results show that OVH descriptors can be calculated effectively with the proposed method, and match well to the 3D geometrical features of the tumor and OARs. Further, the IMRT plan retrieval results match well based on a visual inspection of their 3D geometrical features, and an increase of the area between OVH descriptors leads to a decrease of visual similarity. Therefore, the proposed method can be used effectively for the calculation of an OVH descriptor as well as the retrieval of similar IMRT cases.