Brain tumor is a major challenge of public health, and accurate and efficient detection and quantification of brain tumor are essential for planning effective treatment. In this research, a deep learning approach is proposed to detect and segment brain tumors based on 3D FLAIR MRI scans. A 3D U-Net architecture is proposed where the accuracy is 0.976 which is a good measure of high performance in tumor detection and segmentation. The system also measures tumor metrics in a detailed manner, including tumor volume, orientation, axis length, and location, which are both important in clinical decision making. Preprocessing, segmentation and morphological refinement are performed with the methodology to guarantee specificity in tumor delineation. Training and validation were performed on a large-scale public dataset of 484 patients. The interface of the system is convenient for the users and has the capability of real time processing that helps the seamless integration with the clinical workflows. This paper has three key contributions: high performance detection, real time segmentation, 3D Tumor volume and shape representation. However, the clinical relevance of this system is in improving treatment planning, monitoring and patient outcomes. The system will be improved in future work by adding multimodal imaging, transfer learning, and longitudinal studies.