Recently, unsupervised deep learning methods have shown great potential in image processing. Compared with a large-amount demand for paired training data of supervised methods with a specific task, unsupervised methods can learn a universal and explicit prior information on data distribution and integrate it into the reconstruction process. Therefore, it can be used in various image reconstruction environments without showing degraded performance. The importance of unsupervised learning in MRI reconstruction appears to be growing. Nevertheless, the establishment of prior formulation in unsupervised deep learning varies a lot depending on mathematical approximation and network architectures. In this work, we summarized basic concepts of unsupervised deep learning comprehensively and compared performances of several state-of-the-art unsupervised learning methods for MRI reconstruction.

【Objective】 To explore the application value of ¹⁸F-PSMA PET/CT on the detection of metastatic lesions of prostate cancer with serum total prostate specific antigen (tPSA) ≤20 ng/mL and the predictive variables affecting the imaging results, and to establish a predictive nomogram for the metastasis of prostate cancer. 【Methods】 The imaging, pathological, serum and clinical data of 175 pathologically confirmed prostate cancer patients who underwent ¹⁸F-PSMA PET/CT examination during Jan.2020 and Oct.2021 were retrospectively collected.The patients were divided into metastatic group and non-metastatic group according to PET/CT imaging results, and the positive detection rate of metastatic lesions was calculated.The independent influencing factors of ¹⁸F-PSMA PET/CT in the positive detection of metastatic lesions were determined with univariate and multivariate logistic regression analyses.The predictive nomogram was established. 【Results】 Of the 175 patients, metastatic lesions were detected in 78 cases and not detected in 97 cases, with a detection rate of 44.6% (78/175).There were statistically significant differences between the metastatic group and the non-metastatic group in urinary tract symptoms, androgen deprivation treatment (ADT) at the time of PET/CT examination and the risk level of Gleason score (GS) (P<0.05).Univariate logistic regression showed that urinary tract symptoms(OR=3.64, P<0.001), GS risk (OR=3.96, P<0.001) and concurrent ADT treatment (OR=3.71, P<0.001) were associated with the positive detection rate of metastatic lesions.Multivariate Logistic regression showed that urinary tract symptoms (OR=3.19, P=0.002), GS high-risk group (OR=2.95, P=0.005) and concurrent ADT treatment (OR=3.27, P=0.001) were independent predictors of positive detection rate. 【Conclusion】 The probability of metastasis in newly diagnosed prostate cancer patients with tPSA≤20 ng/mL is high.¹⁸F-PSMA PET/CT is of high value for the early detection of metastasis.Urinary tract symptoms, GS high-risk group and concurrent ADT treatment are independent predictors of metastatic lesions.The predictive nomogram can help assist clinical optimization of imaging examination path.