A Shape Relationship Descriptor for Radiation Therapy Planning Michael Kazhdan 1 , Patricio Simari 1 , Todd McNutt 2 , Binbin Wu 2 , Robert Jacques 3 , Ming Chuang 1 , and Russell Taylor 1 1 Department of Computer Science, Johns Hopkins University, USA 2 Department of Radiation Oncology and Molecular Radiation Science, Johns Hopkins University, USA 3 Department of Biomedical Engineering, Johns Hopkins University, USA ⋆ Abstract. In this paper we address the challenge of matching patient geometry to facilitate the design of patient treatment plans in radio- therapy. To this end we propose a novel shape descriptor, the Overlap Volume Histogram, which provides a rotation and translation invariant representation of a patient’s organs at risk relative to the tumor volume. Using our descriptor, it is possible to accurately identify database pa- tients with similar constellations of organ and tumor geometries, enabling the transfer of treatment plans between patients with similar geometries. We demonstrate the utility of our method for such tasks by outperform- ing state of the art shape descriptors in the retrieval of patients with similar treatment plans. We also preliminarily show its potential as a quality control tool by demonstrating how it is used to identify an organ at risk whose dose can be significantly reduced. 1 Introduction In the treatment of patients with malignant tumors, the goal of intensity-modulated radiation therapy (IMRT) is to deliver a high dose of radiation to the tumor vol- ume while sparing nearby organs at risk (OAR). In practice, a pre-treatment computed tomography (CT) scan of the patient is segmented to identify the tumor volume and OAR. The segmented scan is then used by a dosimetrist to guide the settings for each multi-leaf collimator (MLC) on a radiotherapy machine targeting the tumor. This step is referred to as IMRT planning. The design of a high-quality IMRT plan is one of the most time-consuming and least automated steps of the treatment cycle. The dosimetrist must optimize the MLC settings to achieve a dose distribution that most closely meets a set of physician-driven constraints. For example, in a commonly used treatment proto- col for head-and-neck cancer [1] at least 95% of the tumor volume should receive a dose of at least 70 Gy, while no more than 50% of each parotid should receive ⋆ This research was supported in part by the generosity of Paul Maritz, Philips Radi- ation Oncology Systems (Madison, WI), and by Johns Hopkins University internal funds.