ESTRO 2021 Abstract Book

S417

ESTRO 2021

for other ocular morbidities could not be established.

Figure . a) 5 year LTC rate as a function of tumor apex dose and logistic TCP model fitting curve; b) cataract rate as a function of lens dose and logistic NTCP model fitting curve; (c) radiation optic neuropathy (RON) rate as a function of optic disc dose and logistic NTCP model fitting curve. Each study was weighted by the number of patients included. (1)= Ru-106, (2)= I-125, (3)= Pd-103, (4)= miscellaneous. Conclusion Despite the heterogeneity and insufficient data reporting of outcomes and dosimetric correlates in the reviewed literature, EPT outcomes were successfully modelled for 5-year LTC, cataract and optic neuropathy. Notwithstanding the uncertainty inherent to the data extraction and modelling process, the 95% 5-year local TCP was achieved at physical tumor dose of 80 Gy. More consistent recording of ocular morbidities along with accurate estimation of doses is needed to improve the robustness of toxicity risk prediction in EPT. OC-0531 a probabilistic biological atlas for prostate cancer R. Finnegan 1 , H.M. Reynolds 2 , Y. Sun 1 , M.A. Ebert 3 , L. Holloway 4 , J. Sykes 5 , J. Dowling 6 , A. Haworth 1 1 University of Sydney, Institute of Medical Physics, Sydney, Australia; 2 University of Auckland, Auckland Bioengineering Institute, Auckland, New Zealand; 3 Sir Charles Gairdner Hospital, Department of Radiation Oncology, Nedlands, Australia; 4 Ingham Institute for Applied Medical Research, Medical Physics Research Group, Liverpool, Australia; 5 Sydney West Cancer Network, Blacktown Hospital, Blacktown Cancer and Haematology Centre, Blacktown, Australia; 6 CSIRO Health and Biosecurity, The Australian e-Health and Research Centre, Herston, Australia Purpose or Objective Biologically based prostate radiotherapy treatment planning may improve the likelihood of complication-free disease-free survival. This relies on accurate localisation and characterisation of tumour lesions. Our goal is to generate a population-based biological atlas, to be used with in vivo multiparametric MRI (mpMRI) to aid biologically based treatment planning. Materials and Methods Seventy men scheduled for radical prostatectomy were recruited to this ethics approved study. Following surgery, each prostate specimen was formalin-fixed and paraffin-embedded in a custom sectioning box prior to ex vivo MRI acquisition. After imaging, histology sections were cut using sectioning box slots which corresponded with the ex vivo MRI slices. An expert pathologist annotated and graded (with Gleason scores) each tumour lesion on histology and a previously described image processing algorithm was used to derive cell density maps. A landmark-based registration process was used to map these histology-based data onto the MRI (Figure 1(a)). The population-based model required data from all patients to be co-registered. For this, a reference shape was constructed using the mean prostate, peripheral zone (PZ) and urethra volumes, to which each individual prostate was registered using anatomy-guided deformable image registration (DIR) (Figures 1(b) and 1(c)). In this reference space, the tumour location, tumour grade and cell density maps were aggregated into a statistical atlas.