Value of 18F-PSMA-1007 PET/CT-based radiomics model for differential diagnosis between prostate cancer and benign prostatic hyperplasia
10.3760/cma.j.cn321828-20231115-00107
- VernacularTitle:基于 18F-PSMA-1007 PET/CT影像组学模型在前列腺癌与前列腺增生鉴别诊断中的价值
- Author:
Liang LUO
1
;
Ruxi CHANG
;
Yunxuan LI
;
Jungang GAO
;
Xiang LIU
;
Xiaoyi DUAN
Author Information
1. 西安交通大学第一附属医院PET/CT室，西安 710061
- Keywords:
Prostatic neoplasms;
Prostatic hyperplasia;
Prostate-specific membrane antigen;
Radiomics;
Positron-emission tomography;
Tomography, X-ray computed
- From:
Chinese Journal of Nuclear Medicine and Molecular Imaging
2024;44(2):80-85
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To evaluate the value of prostate specific membrane antigen (PSMA) PET/CT-based radiomics models in differentiation between prostate cancer and benign prostatic hyperplasia (BPH).Methods:Data from 50 patients with prostate cancer (age: (70.0±8.8) years) and 25 patients with BPH (age: (66.9±9.4) years) who underwent 18F-PSMA-1007 PET/CT imaging and prostate biopsy in the First Affiliated Hospital of Xi′an Jiaotong University from May 2020 to September 2022 were retrospectively collected. Patients were divided into the training set ( n=53) and test set ( n=22) in the ratio of 7∶3 by using random seed number. The ROIs were delineated based on PET and CT images, and radiomics features were extracted respectively. Feature selection was performed using the minimum redundancy and maximum relevance (mRMR) and the least absolute shrinkage and selection operator (LASSO) algorithm. PET and PET/CT radiomics models were generated using logistic regression. ROC curve analysis was employed for model evaluation. In addition, comparisons of the 2 radiomics models with parameters including the ratio of free prostate specific antigen (fPSA)/total prostate specific antigen (tPSA), PET metabolic parameters, as well as prostate cancer molecular imaging standardize evaluation (PROMISE) were conducted (Delong test). Results:A total of 7 features were included in the PET radiomics model, and 3 CT-based features and 4 PET-based features were included in the PET/CT radiomics model. The AUCs of PET and PET/CT radiomics models in the training set and test set were 0.941, 0.914 and 0.965, 0.914, respectively, which were higher than those of fPSA/tPSA (0.719 and 0.710), SUV max(0.748 and 0.800), peak of SUV (SUV peak, 0.722 and 0.771), metabolic tumor volume (MTV, 0.640 and 0.595), total lesion uptake (TLU, 0.525 and 0.476) and PROMISE (0.644 and 0.667)[ z values for the training set: from -6.26 to -3.13, all P<0.01; z values for the test set: from -3.16 to -1.08, P>0.05 (fPSA/tPSA, SUV max, SUV peak) or P<0.05 (MTV, TLU, PROMISE)]. The differential diagnostic accuracy, sensitivity and specificity of PET and PET/CT radiomics models in the test set were 86.36%(19/22), 13/15, 6/7 and 90.91%(20/22), 15/15, 5/7, respectively. Conclusion:Compared with the clinical and PET parameters, PSMA PET/CT-based radiomics model can further improve the efficiency of differential diagnosis between prostate cancer and BPH.
