ESTRO 2020 Abstract book

S407 ESTRO 2020

care diagnostics. With the rise of big data analysis, radiomic features (RF) from multimodal images are derived to characterize tumor biology non-invasively. To our knowledge this is the first assessment of RF derived from PSMA-PET images to detect invisible intraprostatic microlesions. Material and Methods 20 patients were prospectively enrolled in this study. Eligible patients underwent 68 Ga-PSMA-PET before prostatectomy. Whole-mount histopathology, evaluated by a pathologist, was co-registered with 68 Ga-PSMA-PET images. Two experts delineated intraprostatic gross tumor volume (GTV), which was subtracted from prostatic gland volume, to obtain a 3D volume of non-PCa tissue in PET (nGTV). RF were extracted from nGTV with the open source python package Pyradiomics. Alongside the original image a local binary pattern (LBP) filter was applied. Additionally, a phantom study was conducted to discard features with significant inter-scanner variability (3 different scanners). 100 of 237 features proved robust and were subsequently statistically analyzed. The analysis was conducted on half prostate and entire prostate level, respectively. Further feature reduction and selection was performed by p-adjustment with Benjamini and Hochbergs False Discovery Rate and a clustered correlation heatmap based on the Spearman’s correlation coefficients. Thus, representative RFs were chosen from each cluster and areas under receiver-operating curves (AUC) were calculated. Results On a half gland basis 31 % of RF significantly (p<0.05) discriminated between nGTV with or without microlesions. All were altered by the LBP filter. LBP-Firstorder- Interquartile-Range had an AUC of 0.78 whereas LBP- Firstorder-Uniformity correlated negatively with microlesions and an AUC of 0.76. Cross-validation of the significant RF was performed on whole prostates. The accuracy of the predictive model improved though for LBP- Firstorder-Range (AUC 0.87) and LBP-Firstorder-Uniformity RF derived from nGTV in PSMA-PET significantly detected the presence of microlesions. Uncovering multifocality in prostate cancer could help guide focal therapies and personalize treatments. Consequently, the aforementioned radiomic biomarkers need to be validated externally before translating into clinical practice, this validation is still ongoing. PD-0670 Population outcomes of EBRT with LDR brachytherapy boost for intermediate, high-risk prostate cancer J. Oh 1 , I. Spadinger 2 , V. Lapointe 1 , J. Morris 1 1 BC Cancer, Radiation Oncology, Vancouver, Canada ; 2 BC Cancer, Medical Physics, Vancouver, Canada Purpose or Objective To evaluate and compare the prognostic and treatment factors and oncologic outcomes of the high-intermediate (HIR) and high-risk (HR) prostate cancer patients treated with LDR-PB and EBRT with the results of Androgen Suppression Combined with Elective Nodal and Dose Escalated Radiation Therapy (ASCENDE-RT). Material and Methods HIR or HR prostate cancer patients were eligible for combined LDR-PB, EBRT, +/- androgen depravation therapy (ADT) treatment from 2010 – 2016 at British Columbia Cancer. All patients received prostate +/- pelvic irradiation to 46Gy followed by LDF-PB boost. ADT was (AUC 0.81). Conclusion

given at the discretion of the treating oncologist. Disease and treatment characteristics and oncologic outcomes were prospectively collected and reviewed. Results 568 patients received the combined modality treatment. Median age was 69 years. 69% (N=391) had NCCN HR disease, 291 (51%) had Gleason grade group (GGG) 4 or 5, and the median iPSA was 12ng/mL. 97 (17%) had clinical stage T3a or higher, and the median positive number of cores (PPC) was 57%. 243 (43%) had 2 or more high risk features (HRF). The median D90 for LDR-PB was 124.5Gy. 93% of HR and 55% of IR received ADT. 5-year relapse-free survival (RFS), metastasis-free survival (MFS), and overall survival (OS) was 85±4%, 90±3%, and 89±3%. On univariate analysis, GGG, HRF, PPC were associated with RFS. GGG and HRF were also associated with MFS. Patients who relapsed in 5 years had worse OS.

Conclusion The combined modality treatment factors and outcomes are comparable to the results of ASCENDE-RT and remain an effective treatment option for IR and HR prostate cancer. Higher GGG, HRF, PPC are potentially associated with worse outcomes. People who had early relapse had worse OS as demonstrated by ASCENDE-RT.