Automatic Identification of Tumor in 3D Brain Magnetic Resonance Images using U-Net

Brain tumor causes early unconsciousness of the human and lead to death by false findings and wrong treatment. The prompt identification and precise segmenting the brain medical imaging are crucial for diagnosis, planning of treatment, and patient results. Manual tumor detection in 3D brain Magnetic Resonance (MR) scans is time-taking and relatively unpredictable. Accurate detection and classification is main aim in prediction of the tumor size and type. This research proposes a customized CNN-based tumor segmented model which is an automated procedure for identifying and segmenting brain tumors in glioblastoma 3D Flair MRI as input based on a modified U-Net architecture. U-Net, a prominent deep learning model noted for its efficacy in picture segmentation tasks, has been modified to handle the complexities of three-dimensional medical data. These tumors, by nature, can take on almost any shape or size and appear in any region of the brain. These reasons force us to look into an artificial intelligence system that is very efficient and makes use of a flexible, high-capacity neural network. The recommended technique is tested on a publicly accessible datasets, proving its capacity to correctly segment diverse tumor types. The experimental results reveal a significant improvement in segmentation accuracy over traditional approaches, with high DSC, accurate tumor stage finding and other key performance measures. The proposed method is a promising tool for medical experts to help them identify brain tumors quickly and reliably, resulting in better patient care and more efficient diagnostic procedures.