ESTRO 2023 - Abstract Book

S1882

Digital Posters

ESTRO 2023

Our novel results show promise but are limited by the small dataset used, particularly in a deep learning setting. However, given the small dataset and the relative success of the methods proposed, we believe that CNNs should be investigated further to predict regions at high risk of harboring LRR.

PO-2099 Bladder dose surface maps identify subregions associated to late toxicities after prostate cancer RT

E. Gioscio 1 , A. Cicchetti 1 , J. Iacovacci 1 , S. Spampinato 2 , J.M. Waskiewicz 3 , B. Avuzzi 4 , E. Garibaldi 5 , E. Villa 6 , A. Magli 7 , D. Cante 8 , G. Girelli 9 , M. Gatti 10 , L. Ferella 11 , B. Noris Chiorda 4 , P. Ferrari 12 , C. Piva 13 , M. Pavarini 14 , R. Valdagni 1,4,15 , V. Vavassori 6 , F. Munoz 16 , C. Fiorino 14 , C. Cozzarini 14 , T. Rancati 1 1 Fondazione IRCCS Istituto Nazionale dei Tumori, Prostate Program, Milan, Italy; 2 Århus University Hospital, Medical Physics, Århus, Denmark; 3 Azienda Sanitaria dell'Alto Adige, Radiotherapy, Bolzano, Italy; 4 Fondazione IRCCS Istituto Nazionale dei Tumori, Radiation Oncology 1, Milan, Italy; 5 AO SS Antonio e Biagio e Cesare Arrigo, Radiotherapy, Alessandria, Italy; 6 Cliniche Gavazzeni-Humanitas, Radiotherapy, Bergamo, Italy; 7 Azienda Ospedaliero Universitaria S. Maria della Misericordia, Radiotherapy, Udine, Italy; 8 ASL TO4, Ospedale di Ivrea, Radiotherapy, Ivrea, Italy; 9 Ospedale degli Infermi, Radiotherapy, Biella, Italy; 10 Istituto di Candiolo, Fondazione del Piemonte per l'Oncologia IRCCS, Radiotherapy, Candiolo, Italy; 11 Ospedale Regionale Parini-AUSL Valle d'Aosta, Radiotherapy, Aosta, Italy; 12 Comprensorio Sanitario di Bolzano, Medical Physics, Bolzano, Italy; 13 ASL TO4, Ospedale di Ivrea, Radiation Oncology, Ivrea, Italy; 14 San Raffaele Scientific Institute, Università Vita Salute San Raffaele, Medical Physics, Milan, Italy; 15 Università degli Studi di Milano, Oncolgy and Hemato-Oncology, Milan, Italy; 16 Ospedale Regionale Parini-AUSL Valle d'Aosta, Radiation Oncology, Aosta, Italy Purpose or Objective To identify sub-regions of the bladder associated with different patient-reported late persistent urinary symptoms after RT for prostate cancer using Dose-Surface Maps (DSMs). Materials and Methods Patients (pts) from a prospective, multi-institutional trial, including pts treated with radical RT with conventional/moderate hypofractionation were considered. One item from CIQ-SF (incontinence) and 3 items from IPSS (frequency, urgency, weak stream) were considered to assess persistent symptoms scored by adapting a previously introduced methodology (LAPERS). We defined persistent symptoms as mild or moderate/severe if the median score over late follow-ups (FUPs) was ≥ 1 or ≥ 2, respectively; in addition, a median score worse than the baseline (BL) value was considered to split pts that worsened after RT. The minimum FUP was 5 years, and only pts with at least 3 late FUPs were included. The analysis was conducted considering as end-points all persistent symptoms and only the ones that worsened at FUP compared to BL. The bladder DSMs were generated by an automatic dedicated tool written in Python, starting from the CT scans/dose cubes of treatment planning. The maps were opened in their anterior surface, aligned and normalised both in left-right (LR) and cranial-caudal (CC) directions to compare pts and apply voxel-wise analysis (VWA) spatially. Specifically, for L-R direction, we normalized to the maximum bladder extent within the population. In the CC direction, we: (a) aligned all maps at the most caudal slice, (b) kept the most caudal 1.5 cm fixed (as a surrogate of the bladder neck, considering this part of the bladder fixed and not influenced by bladder filling), (c) normalised the remaining bladder surface to the CC extension of the median bladder height (6 cm in this population). In this way, “small” bladders were stretched to the median and “large” bladders were shrunken to the same median. Doses were converted to EQD2 (a/b=1Gy, previously found for late urinary symptoms) at the pixel level. A pixel-wise t-test was used to highlight anatomical areas with significant differences between the average dose maps in the groups with/without toxicity. Results 384/539 pts had complete FU data and DSM. Table 1 reports toxicity rates. Fig.1 shows DSMs and summary results of the VWA. For severe urinary frequency, the larger dose differences (4-8Gy) were placed posteriorly, including the upper part of the bladder neck. The bottom part of the neck and the cranial bladder were highlighted for moderate/severe urinary urgency (differences 4-7Gy). The severe weak stream was associated with the anterior bladder. We found no specific significant regions for the other symptoms.