ESTRO 37 Abstract book

ESTRO 37

S501

PO-0928 Normal tissue dose estimation using large databases for automatic plan selection of similar patients W. Van Elmpt 1 , T. Lustberg 1 , J. Van Soest 1 , M. Gooding 2 , A. Dekker 1 1 MAASTRO Clinic, Department of Radiation Oncology, Maastricht, The Netherlands 2 Mirada Medical Ltd, Science and Medical Technology, Oxford, United Kingdom Purpose or Objective Pre-defined planning constraints cannot always be met during the planning procedure, then decisions need to be made for prioritizing of the objectives for targets and/or organs-at-risk (OAR). Whether planning constraints will be met is difficult to predict up front. Therefore the goal of this project was to automatically select similar patients from a large database of previously treated patients. Such a method gives the radiation oncologists an indication of target coverage and OAR dose levels achieved in the past based on the geometric characteristics of the current patient. Material and Methods A database of non-small cell lung cancer patients was created: stage I treated with SBRT (N=142) and stage II-III treated with VMAT based concurrent chemo-radiotherapy (N=191). These patients were automatically processed in a dedicated medical image analysis framework: Distances between the center of the PTV and a reference point (center of mass of both lungs), dose-volume histograms and volume of the PTV and OARs were calculated. Next, a near neighbour search strategy (kNN) was optimized to select 5 similar patients from the database based on volume and location of the PTV (= the spatial coordinates relative to the reference point) only. Validation was performed by comparing the DVH metrics using a leave- one-out strategy. Results For the vast majority of the patients, the similar patients were also from the same cohort (SBRT 90.1%; VMAT 92.1%). For every patient, the difference of the median DVH metric of the 5 similar patients was compared to the difference for a random 5 patient selection or the average entire cohort. Relative differences per OAR are show in Figure 1. Absolute differences for (similar patient; random selection; average cohort) were: mean heart dose (1.4 Gy; 3.2 Gy; 2.9 Gy); spinal cord maximum dose (4.8 Gy; 15.1 Gy; 15.3 Gy); esophagus V35 (2.4%; 8.7%; 3.0%); maximum dose mediastinal structures (3.6 Gy; 23.8 Gy; 9.7 Gy) and mean lung dose (1.7 Gy; 5.8 Gy; 5.5 Gy), respectively.

Figure 1: Difference in selected DVH metric for the similar patients, random selection of 5 patients and the average of the entire cohort (N=333). Using a similar patient search strategy improved accuracy compared to random selection or the average of the patient cohort. Conclusion Simple geometric features combined with a kNN search was able to return dosimetrically similar patient(s) with their DVH metrics typically within a few Gy or percent points. The radiation oncologist may use this information up front to get an estimate of tumour and OAR dose levels (i.e. DVH) based on the current patient anatomy without any treatment planning performed. Furthermore, since these similar patients have clinical feasible plans, these previously delivered plans may quick start the manual treatment planning process and reduce treatment optimization time, e.g. by copying the weights and objective parameters in the treatment planning system. PO-0929 Exploring dose-effect relationships for late fecal incontinence after modern radiotherapy W. Heemsbergen 1 , L. Incrocci 1 , F. Pos 2 , R. Wortel 1 , J. Lebesque 2 , M. Witte 2 1 Erasmus MC Cancer Institute, Radiation Oncology, Rotterdam, The Netherlands 2 Netherlands Cancer Institute, Radiation Oncology, Amsterdam, The Netherlands Purpose or Objective Prostate cancer patients treated with external beam radiotherapy are at risk to develop late fecal incontinence, which is a complex phenomenon affecting Quality of Life. It may involve changes in bowel frequency, failure to resist urge, affected sphincter control, uncontrolled flatus, leakage of feces or mucus, and/or changes in consistency. Recent investigations on dose-effect relationships focused mainly on anal canal Poster: Physics track: (Radio)biological modelling