Author: Ozan Cem Guler, William Silva Mendes, Sangbo Oh, Cem Onal, Lei Ren, Apurva Singh, Phuoc Tran 👨🔬
Affiliation: University of Maryland School of Medicine, Baskent University Faculty of Medicine, Department of Radiation Oncology, Department of Radiation Oncology, University of Maryland School of Medicine 🌍
Purpose: To predict metastasis-free survival (MFS) for patients with prostate adenocarcinoma treated with androgen deprivation therapy and external radiotherapy using clinical factors and radiomics extracted from pre-treatment PSMA PET/CT scans.
Methods: Our cohort includes 134 prostate adenocarcinoma patients (60 patients having nodal involvement). Gross-tumor-volumes of primary tumor (GTVp) and nodes (GTVn) on CT, PET scans were segmented. Features were extracted and Z-score-normalized; dimensionality reduction performed using Principal-Components-Analysis. For patients with only primary tumor, we took three principal-components (PCs) from CT and PET, respectively. For patients with nodes, we calculated weighted-average (by volume) of radiomics from primary tumor and nodes, computed their first PC, combined with two PCs from GTVp to obtain three PCs from CT and PET, respectively. Radiomics PCs and clinical variables (age, Gleason score, virgin-prostate-specific-antigen (vPSA), PSA_relapse) formed the predictors. Due to imbalance in MFS data (metastasis-24, no metastasis-110), we performed 70:30 train-test split and applied imbalance correction to training data. Univariate Cox-regression was used to select top five predictors (logistic regression p < 0.05) for model 1. Multivariate Cox-regression analysis was performed on imbalance-corrected train data and fit on test data (using predictors selected from train data). Model 2 was built using clinical variables, radiomic PCs from primary tumors to assess improvement by adding node-radiomics.
Results: Results of time-to-event analysis (metastasis-free survival) were- Cox-regression c-scores: model1: train- 0.70 [0.64, 0.71]; test- 0.61 [0.56, 0.62]; model2: train- 0.67 [0.62, 0.68]; test- 0.58 [0.54, 0.59]. We observed that integration of node with primary tumor-radiomics improved performance of the prognostic model.
Conclusion: This is one of the first studies to explore the prognostic value of pre-treatment PSMA-PET, a relatively recent advancement in the care of prostate adenocarcinoma patients. Results showed that using PSMA PET/CT radiomics information from primary tumor and nodes improves MFS prediction, compared to using primary tumor-radiomics only.