Deep Learning revolutionizes brain tumour segmentation in medical imaging, enabling automatic detection with high accuracy. A lot of software with emphasis on automatic classification and segmentation in the medical field was developed with Deep Learning algorithms as their base. This thesis investigates the use of deep learning for automatic brain tumor segmentation in MRI scans, employing a 2D U-Net convolutional neural network. The model was trained on the BraTS 2020 dataset using FLAIR and T1ce modalities to classify tumor subregions, including necrotic core, edema, and enhancing tumor. Achieving an accuracy of 99.23%, a Dice coefficient of 0.6471, and a Mean IoU of 0.7134, the model demonstrated strong performance in segmenting tumor structures.