AI model training based on MRI of 1,029 patients with brain metastases from a single institutesingle MRI brand was conducted. A benchmark algorithm of 2D Mask R-CNN was usedresulted in an initial model (DeepMets®). Model generalization of DeepMets®was then carried out over a nationwide population-based dataset via deep active learning on 559 patients (randomized from 3,125) from National Health Insurance Administration (NHIA) medi-cloud, Taiwan. Iterative refining process using the ResNext50 U-Net architecture with attention mechanisms were undertakenresulted in a newer version model (DeepMets®-Plus).

"DeepMets was established on an enhanced U-Net, which combined convolutionself-attention mechanisms to explore localglobal contents for better segmentation performance. We also applied deep active learning to a nationwide population-based database of brain metastases. 
 
 
 
 The advantage was twofold. First, the access to a national-scale dataset opened the opportunity to generalize our system to accommodate MR images of various devices. Second, the use of active learning ensured that despite working on a huge collection of MR images, data annotation was done only on those that DeepMets performed unsatisfactorily. Our method provided an effective solution to achieve model generalizationyield significant performance gains."

"DeepMets®-plus is designed to speed up tumor detectionfacilitating healthcare qualitythroughput, which has assisted in diagnosis of thousands of cases since 2019,reduced the time physicians required to review the medical images from 10 minutes to only 30 seconds. Furthermore, Patient＇s medical service process is feasibly shortened from 2 weeks to a half day by the setting. The application has facilitated decision-making for treatmentalleviates patient＇s anxiety along the process.
 
 
 
 DeepMets®-plus can be widely applicable in tumor cares, assisting neurosurgeons for initial contouring of brain metastases on MRIto plan radiosurgery. It can also improve the reading performancehelp fast triage of abnormal case