Brain tumours must be accurately identified and located on magnetic resonance imaging (MRI) scans for proper diagnoses and treatments. Accuracy when segmenting these areas ensures doctors have a clear understanding of how much of the tumour needs to be removed or treated. In our research, we propose using an Enhanced 3D U-net Model for accurately segmenting and analysing brain tumours. The model has been trained using a dataset of brain MRI scans that have been merged and optimized from 2D and 3D spatial information and labelled with tumour indications. The efficacy of the suggested system is determined by computing several metrics, including loss, accuracy, mean IOU, precision, sensitivity, specificity, and Dice coefficient. We looked into SegNet as another architecture to compare with UNet. Examining the results of our 3D UNet model against SegNet could provide us with a better understanding of how effective our solution is for categorizing and analyzing brain tumours. Our findings indicate that the 3D UNet model proposed has a higher accuracy in tumour segmentation than other approaches, with excellent precision. This model offers fast processing, making it appropriate for real-life medical applications. The updated 3D UNet architecture could result in more precise and successful segmentation of brain tumours, thus resulting in better diagnosis and treatment plans. To segregate brain tumours and evaluate 3D MRI data, this research provides a new Convolutional Neural Network model, which represents a significant advancement in medical image processing.