PT - JOURNAL ARTICLE AU - Li, Hui AU - Liu, Chen AU - Yuan, Jianmin AU - Wang, Zhe AU - Meng, Xiangxi AU - Wang, Shujing AU - Zhang, Yan AU - Yu, Boqi AU - Zhu, Hua AU - Li, Nan AU - Yang, Zhi TI - PSMA combined with luminal water imaging for the diagnosis of primary prostate cancer: A study based on multiparameter PSMA PET/MR DP - 2021 May 01 TA - Journal of Nuclear Medicine PG - 1363--1363 VI - 62 IP - supplement 1 4099 - http://jnm.snmjournals.org/content/62/supplement_1/1363.short 4100 - http://jnm.snmjournals.org/content/62/supplement_1/1363.full SO - J Nucl Med2021 May 01; 62 AB - 1363Objectives: Multiparametric magnetic resonance imaging (mpMRI) of the prostate is widely used for the diagnosis and staging of prostate cancer (PCa). Recently, the value of prostate specific membrane antigen (PSMA) PET/CT in the diagnosis and staging of PCa is widely accept. The simultaneous PET/MR enables the possibility of combination of the advantages of these two imaging modalities for the management of PCa. Luminal water imaging (LWI), a quantitative multicomponent T2 mapping technique, has shown promise for PCa detection and characterization. In this study, we assessed the diagnostic performance of PSMA combined with LWI for the diagnosis of primary PCa.Methods: Nine patients with systematic biopsy proven PCa were enrolled in this study, all patients underwent pelvic PSMA PET/MR (uPMR 790, United-Imaging Healthcare, Shanghai, China). From the PSMA PET/MR date, twelve MR parameters (mean components, mean luminal water fraction (LWF), mean T2 short NNLS, mean T2 long NNLS, mean Area short, mean Area long, std components, std LWF, std T2 short NNLS, std T2 long NNLS, std Area short, std Area long) and three PET parameters (SUVmax, SUVmean, SUVmin) were generated. A paired t test was used to determine significant differences between PSMA PET/MR parameters in malignant and nonmalignant tissue. Subsequently, two lowest p-value to select the relevant features were determined through a cross-validation process, and the features thus selected were used in the subsequent machine learning analysis.Results: The eight MR parameters (mean LWF, mean T2 short NNLS, mean T2 long NNLS, mean Area long, std components, std LWF, std T2 long NNLS, std Area long) were significantly different between malignant and nonmalignant tissue (all P<0.05). The three PET parameters (SUVmax, SUVmean, SUVmin) were significantly different between malignant and nonmalignant tissue (all P<0.05). Mean T2 long NNLS and SUVmax were the two lowest p-value features. The univariate logistic regression model established with mean T2 long NNLS and SUVmax could all effectively predict the malignant tissue with sensitivity and specificity were 94.7% and 75% for mean T2 long NNLS (AUC=0.855), 78.9% and 100% for SUVmax (AUC=0.876). Subsequent machine learning analysis showed that the combination of mean T2 long NNLS and SUVmax improved the sensitivity and specificity than each single factor. Conclusion: Combining the molecular information of PSMA PET with LWI could improve the diagnostic performance of PCa than each single imaging modality.Acknowledgements:This work is supported by Beijing Municipal Administration of Hospitals-Yangfan Project (ZYLX201816).