There has been a surge of interest in artificial intelligence and machine learning (AI/ML)-based medical devices. However, it is poorly understood how and which AI/ML-based medical devices have been approved in the USA and Europe. We searched governmental and non-governmental databases to identify 222 devices approved in the USA and 240 devices in Europe. The number of approved AI/ML-based devices has increased substantially since 2015, with many being approved for use in radiology. However, few were qualified as high-risk devices. Of the 124 AI/ML-based devices commonly approved in the USA and Europe, 80 were first approved in Europe. One possible reason for approval in Europe before the USA might be the potentially relatively less rigorous evaluation of medical devices in Europe. The substantial number of approved devices highlight the need to ensure rigorous regulation of these devices. Currently, there is no specific regulatory pathway for AI/ML-based medical devices in the USA or Europe. We recommend more transparency on how devices are regulated and approved to enable and improve public trust, efficacy, safety, and quality of AI/ML-based medical devices. A comprehensive, publicly accessible database with device details for Conformité Européene (CE)-marked medical devices in Europe and US Food and Drug Administration approved devices is needed.

Segmentation of anatomical structures and pathologies is inherently ambiguous. For instance, structure borders may not be clearly visible or different experts may have different styles of annotating. The majority of current state-of-the-art methods do not account for such ambiguities but rather learn a single mapping from image to segmentation. In this work, we propose a novel method to model the conditional probability distribution of the segmentations given an input image. We derive a hierarchical probabilistic model, in which separate latent variables are responsible for modelling the segmentation at different resolutions. Inference in this model can be efficiently performed using the variational autoencoder framework. We show that our proposed method can be used to generate significantly more realistic and diverse segmentation samples compared to recent related work, both, when trained with annotations from a single or multiple annotators. The code for this paper is freely available at https://github.com/baumgach/PHiSeg-code.

Background Recent studies have reported the additive value of combined gallium 68 (Ga)labeled Glu-urea-Lys (Ahx)-HBED-CC ligand targeting the prostate-specific membrane antigen (PSMA) (hereafter called Ga-PSMA-11) PET/MRI for the detection and localization of primary prostate cancer compared with multiparametric MRI. Purpose To compare the diagnostic accuracy and interrater agreement of multiparametric MRI and Ga-PSMA-11 PET/MRI for the detection of extracapsular extension (ECE) and seminal vesicle infiltration (SVI) in patients with prostate cancer. Materials and Methods Retrospective analysis of 40 consecutive men who underwent multiparametric MRI and Ga-PSMA-11 PET/MRI within 6 months for suspected prostate cancer followed by radical prostatectomy between April 2016 and July 2018. Four readers blinded to clinical and histopathologic findings rated the probability of ECE and SVI at multiparametric MRI and PET/MRI by using a five-point Likert-type scale. The prostatectomy specimen served as the reference standard. Accuracy was assessed with a multireader multicase analysis and by calculating reader-average areas under the receiver operating characteristics curve (AUCs), sensitivity, and specificity for ordinal and dichotomized data in a region-specific and patient-specific approach. Interrater agreement was assessed with the Fleiss multirater. Results For multiparametric MRI versus PET/MRI in ECE detection, respectively, AUC, sensitivity, and specificity in the region-specific analysis were 0.67 and 0.75 (.07), 28% (21 of 76) and 47% (36 of 76) (.09), and 94% (529 of 564) and 90% (509 of 564) (.007). For the patient-specific analysis, AUC, sensitivity, and specificity were 0.66 and 0.73 (.19), 46% (22 of 48) and 69% (33 of 48) (.04), and 75% (84 of 112) and 67% (75 of 112) (.19), respectively. For multiparametric MRI versus PET/MRI in SVI detection, respectively, AUC, sensitivity, and specificity of the region-specific analysis were 0.66 and 0.74 (.21), 35% (seven of 20) and 50% (10 of 20) (.25), and 98% (295 of 300) and 94% (282 of 300) (< .001). For the patient-specific analysis, AUC, sensitivity, and specificity were 0.65 and 0.79 (.25), 35% (seven of 20) and 55% (11 of 20) (.20), and 98% (137 of 140) and 94% (131 of 140) (.07), respectively. Interrater reliability for multiparametric MRI versus PET/MRI did not differ for ECE (, 0.46 vs 0.40; = .24) and SVI (, 0.23 vs 0.33; = .39). Conclusion Our results suggest that gallium 68 (Ga)-labeled Glu-urea-Lys (Ahx)-HBED-CC ligand targeting the prostate-specific membrane antigen (PSMA) (Ga-PSMA-11) PET/MRI and multiparametric MRI perform similarly for local staging of prostate cancer in patients with intermediate-to-high-risk prostate cancer. The increased sensitivity of Ga-PSMA-11 PET/MRI for the detection of extracapsular disease comes at the cost of a slightly reduced specificity.

The intense accumulation of prostate-specific membrane antigen (PSMA) radioligands in salivary glands is still not well understood. It is of concern for therapeutic applications of PSMA radioligands, because therapeutic radiation will damage these glands. A better understanding of the uptake mechanism is, therefore, crucial to find solutions to reduce toxicity. The aim of this study was to investigate whether the accumulation of PSMA-targeting radioligands in submandibular glands (SMGs) can be explained with PSMA expression levels using autoradiography (ARG) and immunohistochemistry (IHC). Methods: All patients gave written informed consent for further utility of the biologic material. The SMG of 9 patients, pancreatic tissue of 4 patients, and prostate cancer (PCA) lesions of 9 patients were analyzed. Tissue specimens were analyzed by means of PSMA-IHC (using an anti-PSMA-antibody and an immunoreactivity score system [IRS]) and ARG using 177 Lu-PSMA-617 (with quantification of the relative signal intensity compared with a PSMA-positive standard). The SUV max in salivary glands, pancreas, and PCA tissues were quantified in 60 clinical 68 Ga-PSMA-11 PET scans for recurrent disease as well as the 9 primary tumors selected for ARG and IHC. Results: PCA tissue samples revealed a wide range of PSMA staining intensity on IHC (IRS 5 70-300) as well as in ARG (1.3%-22% of standard). This variability on PCA tissue could also be observed in 68 Ga-PSMA-11 PET (SUV max , 4.4-16) with a significant correlation between ARG and SUV max (P , 0.001, R 2 5 0.897). On IHC, ARG, and 68 Ga-PSMA-11 PET, the pancreatic tissue was negative (IRS 5 0, ARG 5 0.1% ± 0.05%, SUV max of 3.1 ± 1.1). The SMG tissue displayed only focal expression of PSMA limited to the intercalated ducts on IHC (IRS 5 10-15) and a minimal signal on ARG (1.3% ± 0.9%). In contrast, all SMG showed a high 68 Ga-PSMA-11 accumulation on PET scans (SUV max 23.5 ± 5.2). Conclusion: Our results indicate that the high accumulation of PSMA radioligands in salivary glands does not correspond to high PSMA expression levels determined using ARG and IHC. These findings provide evidence, that the significant accumulation of PSMA radioligands in SMG is not primarily a result of PSMA-mediated uptake.