Brain tumors constitute a tough medical challenge, needing early detection and exact classification for optimal patient outcomes. The unchecked growth of brain cells within tumors affects nearby tissues, with symptoms varying based on factors such as location, size, and cellular composition. Distinguishing between cancerous and non-cancerous tumors is crucial, as treatment techniques and prognoses differ appropriately. In response to this critical requirement, this research proposes a unique method using Convolutional Neural Networks (CNN) with the U-Net architecture. My methodology uses public The Cancer Genome Atlas (TCGA) and The Cancer Imaging Archive (TCIA) datasets that consist of MRI images of 455 patients, to train and test a novel variant of the U-Net system within the context of deep learning. The results of the experiment are remarkable, with the proposed model achieving an extraordinary test accuracy of 99. 83% on the dataset, underscoring its efficacy in brain tumor detection. A comparative analysis has been done with the existing literature to determine that this U-Net based technique is significantly superior to other deep learning techniques in this regard. This paper outlines the opportunities for using deep learning techniques in the development of imaging analysis for medical applications, and thus the continuous search for the enhancement of diagnosis and treatment of patients with brain tumors.
