Brain Tumor Detection Using Deep Learning With Synthetic Data Augmentation

Early brain tumor identification is one of the most critical challenges that neurologists and radiologists face. Effective segmentation and classification still need to be improved despite numeroussignificant efforts and encouraging outcomes. Images of various sorts are employed for tumor segmentation, categorization, and diagnosis. For a noninvasive and more accurate image of the interior anatomy of the tumor, magnetic resonance imaging (MRI) is preferred above all other imaging techniques. Nevertheless, manually identifying and differentiating brain tumors from magnetic resonance Imaging (MRI) scans are difficult and error-prone, and it calls for the need for an automated brain tumor detection system for early tumordetection. This study suggests a deep-learning approach for analyzing MRI data in order to detect brain tumors. The suggested approach comprises three key phases: pre-processing, segmentation, adopting k means clustering, and finally, tumor classification lastly MRI data using a customized VGG19 (19 layered Visual Geometric Group) model. Besides that, the synthetic data augmentation idea is adopted to enhance the amount of data accessible for classifier training to improve classification accuracy. The outcomes support the efficacy of the suggested strategy and demonstrate that it is more accurate than already available methodologies.