Abstract Book

S257

ESTRO 37

modulated radiotherapy and Volumetric Modulated Arc Therapy treatment plans were generated for a prescription dose of 25 Gy in 10 fractions. The main objective was NC function at 3 months assessed by Free and Cued Selective Reminding Test (FCSRT). The FCSRT is a well-validated and reliable assessment of memory, including encoding, retrieval, and retention of new information over time. Results These treatment modalities spared the hippocampus, with a D100 of 8.2 ± 1.3 Gy and maximum dose of 10.3 ± 2.3 Gy. There was a decline in delayed free recall in PCI vs PCI-HA arm at 3 months (27 vs 3%; p 0.01; RR 8 [IC 95% 1.06-60.08]). There were declines in delayed free recall (48.1% vs 4.8 %; p 0.001; RR 10 [IC 95% 1.4- 71.23]) and in total free recall (33.3% vs 4.8%; p 0.01; RR 7 [IC 95% 0.9- 50.9]) between two arms at 6 months. Conclusion There was a significant decline in memory in PCI group. Further investigation to assess its impact on long-term follow-up is in progress. OC-0499 Evaluation of prostate tumor delineations on multiparametric MRI in a multicenter radiotherapy trial M. Van Schie 1 , C. Dinh 1 , P. Van Houdt 1 , J. Pos 1 , S. Heijmink 2 , L. Kerkmeijer 3 , A. Kotte 3 , R. Oyen 4 , K. Haustermans 5 , U. Van der Heide 1 1 Netherlands Cancer Institute, Radiation Oncology, Amsterdam, The Netherlands 2 Netherlands Cancer Institute, Radiology, Amsterdam, The Netherlands 3 University Medical Center Utrecht, Radiation Oncology, Utrecht, The Netherlands 4 University Hospitals Leuven, Radiology, Leuven, Belgium 5 University Hospitals Leuven, Radiation Oncology, Leuven, Belgium Purpose or Objective Focal dose escalation in prostate cancer requires reliable and reproducible delineation of the tumor on multiparametric (mp-) MRI. Guidelines to minimize variability in contouring of the visible tumor on mp-MRI are not yet available. In the absence of contouring guidelines we assessed the variability in clinical delineations in a multicenter prostate radiotherapy trial (FLAME, NTC01168479) to quantify the inter-institutional differences in interpretation of mp-MRI. Material and Methods We analyzed mp-MRI and clinical data from 230 patients in 3 institutes: UMC Utrecht (n = 144), UZ Leuven (n = 34) and NKI Amsterdam (n = 52). We performed a logistic regression to examine each institute’s weighting of T2w, ADC and K trans intensity maps in the delineation of the cancer. Differences between institutes on clinical characteristics as well as weighting of mp-MRI were examined with a Mann-Whitney U test (α = 0.05, Bonferroni corrected for multiple testing) and a Kruskal- Wallis test. Since reviewing of all delineations by a panel of experts is prohibitively time consuming, we used a published tumor probability (TP) model trained using pathological data to identify gross discrepancies with each institute’s clinical delineations [1]. We identified discrepancies based on Areas Under the ROC Curve (AUCs) between the calculated TP map and clinical delineations, and selected cases for review by a panel of experts. Results Table 1 shows characteristics of the delineations per institute. Significant differences between institutes were observed for all characteristics except tumor location (p Proffered Papers: CL 10: Prostate

= 0.44). Logistic regression coefficients for institutes I-1, I-2 and I-3 are shown in figure 1a. For each MRI modality coefficients were significantly different (p < 0.001), except between I-1 and I-2 for K trans . The strongest negative coefficient was observed in I-1 for ADC, the strongest positive coefficient in I-3 for K trans . Comparison of the TP maps with clinical delineations resulted in median (range) AUC values of 0.93 (0.50– 0.99), 0.80 (0.47–0.98) and 0.80 (0.19–0.99), for I-1, I-2 and I-3 respectively. Analysis of review cases revealed several causes of low AUC: a tumor surrounding a transurethral resection cavity, mp-MRI information that was contradictive between modalities, and tumors in the transition zone that were often compromised by benign confounders. Examples are shown in figure 1b.

Conclusion Interpretation of mp-MRI in prostate tumor delineation differed significantly between in stitutes. This was concluded from differences in delineated numbers and volumes of tumors. Also logistic regression coefficients showed different weighting of MRI modalities among the participating institutes. Disagreement between TP map and clinical delineations is a clear indication for further review. Guidelines for contouring prostate tumors on mp- MRI are needed. Correlation between observed regression coefficients and AUC values may provide input for these guidelines. [1] Dinh et al, Med Phys, 2017