The early diagnosis of brain tumors is critical for augmenting treatment options and bolstering patient survival rates. While magnetic resonance imaging (MRI) generates high-contrast and detailed images, the manual segmentation of brain lesions from MRI scans remains a time-consuming and challenging task. Deep learning, particularly via convolutional neural networks (CNNs), has shown considerable promise in automating medical picture interpretation, which might contribute to decreased errors and improved results. This study covers recent advancements in employing CNNs for the identification and categorization of brain malignancies in MRI imaging, concentrating on their capacity to recognize aberrant cell proliferation in the brain. By allowing correct diagnosis, medical imaging plays a key role in treatment planning and illness monitoring. Our study proposes a unique segmentation strategy for brain lesions incorporating CNNs and inception modules, seeking to increase segmentation accuracy. We present a hybrid deep learning model, TransAddAttUnet, which attained a performance accuracy of 98.1%. When compared to current models, our technique exhibited enhanced performance, suggesting promise to enhance the accuracy and efficacy of brain tumor detection.