Click here to


Are you sure ?

Yes, do it No, cancel

CBCT-Based Radiomics of Prostate Cancer for Predicting Patient Outcome to Radiotherapy

R Delgadillo*, B Spieler, J Ford, D Kwon, F Yang, M Studenski, K Padgett, M Abramowitz, R Stoyanova, A Pollack, N Dogan, University of Miami, Miami, FL


(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: Daily Cone beam CT (CBCT) images may be useful in detecting early morphological changes during prostate cancer radiotherapy. The aim of this study was to evaluate the performance of CBCT-based radiomics features to predict patient outcomes of prostate cancer.

Methods: Nineteen prostate cancer patients from an IRB-approved database were retrospectively selected. CBCT imaging characteristics were kept consistent by selecting patients treated on the same modality and imaging platform. Patients were treated on dose fractionation schemes of 80-86 Gy/40 fractions or 76-91 Gy/37-38 fractions with a 12 Gy boost. Patient daily CBCT images were reconstructed using both iterative CBCT (iCBCT) and standard CBCT (sCBCT) algorithms. First and second order statistical radiomics features were extracted from daily CBCT prostate GTV and averaged over all fractions. Patient outcomes analyzed included genitourinary toxicities (GUT), gastrointestinal toxicities (GIT), combined GU/GIT, and Quality-of-Life (QOL) score. Logistic regression model and linear regression model were used to estimate effects of radiomics features on toxicity and QOL, respectively. Odds ratios (ORs), regression coefficients, 95% confidence intervals, and p-value were estimated. Area under the receiver operating characteristic curve (AUC) and R2 were obtained as a measure of models’ goodness-of-fit. Statistical significance was p<0.05. All tests were two-sided.

Results: Several radiomics features were significant predictors of GUT (3 features), GU/GIT (4 features), and QOL (2 features). Toxicity occurred in 4 patients for GIT, 10 patients for GUT, and 11 patients for GU/GIT. No radiomics features were significant predictors for GIT due to low occurrence. The best outcome predictor overall was for GUT using Global Variance with AUC of 0.82/0.81 (iCBCT/sCBCT), p-value of 0.03, and OR of 0.21. Statistically significant radiomics features often had similar performance for iCBCT and sCBCT.

Conclusion: CBCT-based radiomics features of prostate cancer show clinically relevant and promising results in predicting GU toxicities, GU/GI toxicities, and QOL.

Funding Support, Disclosures, and Conflict of Interest: This project was funded in part by a grant from the Varian Medical Systems, Inc.


Cone-beam CT, Texture Analysis


Not Applicable / None Entered.

Contact Email