Background/Objectives: This study evaluates whether convolutional neural networks (CNNs) can be trained to determine the primary tumor origin from MRI images alone in patients with metastatic brain lesions. Methods: This retrospective, monocentric study involved the segmentation of 1175 brain lesions from MRI scans of 436 patients with histologically confirmed primary tumor origins. The four most common tumor types—lung adenocarcinoma, small cell lung cancer, breast cancer, and melanoma—were selected, and a class-balanced dataset was created through under-sampling. This resulted in 276 training datasets and 88 hold-out test datasets. Bayesian optimization was employed to determine the optimal CNN architecture, the most relevant imaging sequences, and whether the masking of images was necessary. We compared the performance of the CNN with that of two expert radiologists specializing in neuro-oncological imaging. Results: The best-performing CNN from the Bayesian optimization process used masked images across all available MRI sequences. It achieved Area-Under-the-Curve (AUC) values of 0.75 for melanoma, 0.65 for small cell lung cancer, 0.64 for breast cancer, and 0.57 for lung adenocarcinoma. Masked images likely improved performance by focusing the CNN on relevant regions and reducing noise from surrounding tissues. In comparison, Radiologist 1 achieved AUCs of 0.55, 0.52, 0.45, and 0.51, and Radiologist 2 achieved AUCs of 0.68, 0.55, 0.64, and 0.43 for the same tumor types, respectively. The CNN consistently showed higher accuracy, particularly for melanoma and breast cancer. Conclusions: Bayesian optimization enabled the creation of a CNN that outperformed expert radiologists in classifying the primary tumor origin of brain metastases from MRI.

The Prostate Imaging-Reporting and Data System (PI-RADS) calls for reporting the prostate index lesion and the location within the transition (TZ) or peripheral zone (PZ) and location on a corresponding sector map. The aim of this study was to train a deep learning DL-based algorithm for automatic prostate sector mapping and to validate its' performance. An automatic 24-sector grid-map (ASG) of the prostate was developed, based on an automatic zone-specific deep learning segmentation of the prostate. To evaluate the efficacy of the method, fiducials for random locations within the prostate were placed, and the corresponding sectors were determined for 50 mpMRI datasets. The reference standard was defined in a consensus read by two expert uroradiologists. Annotated fiducial locations were evaluated automatically by the ASG and by four radiologists in two reads with and without the help of a superimposed sector grid-map and the success rate was compared. The ASG algorithm identified the correct prostate sector of the annotated lesions in 80 % (40/50 reads) of the cases and outperformed readings of the four radiologists with 55 % (109/200), p < 0.0001. The added use of the 24 ASG map significantly improved the rate of correct sector annotation for the four radiologists to 71 % (141/200), p < 0.004. The 24 ASG map was effective for prostate sector segmentation and significantly improved location reporting of prostate lesions.