INTEGRATING U-NET CNN FOR MRI BRAIN TUMOR SEGMENTATION AND SURVIVAL PREDICTION:A DEEP LEARNING APPROACH

Abstract

Brain tumor segmentation is a difficult task in medical image processing that is essential for the detection and planning of brain cancer. Brain tumor imaging frequently uses magnetic resonance imaging (MRI), but manually segmenting tumors from MRI data is a laborious and subjective operation. In this research, I suggest employing a 3D U-Net convolutional neural network (CNN) architecture and deep learning to automatically segment brain tumors. I also incorporate the segmented tumor volume and patient clinical information into a Cox regression analysis survival prediction model. With a Dice coefficient of 0.88, a sensitivity of 0.84, and a specificity of 0.99, the experimental results show that our suggested approach achieves state-of-the-art performance on the BraTS 2020 dataset for brain tumor segmentation. With a concordance index of 0.83, the suggested survival prediction model has acceptable predictive accuracy. Additionally, I carry out ablation experiments to look into the significance of various elements in the suggested strategy, such as data augmentation, regularisation, and feature selection. The suggested method may help medical personnel make well-informed judgements about patient care and treatment plans while also greatly increasing the precision and efficacy of brain tumor diagnosis and treatment planning.