BrainSegNeT: A Lightweight Brain Tumor Segmentation Model Based on U-Net and Progressive Neuron Expansion.

Brain tumor segmentation is a critical task in medical image analysis. In recent years, several deep learning-based models have been developed for brain tumor segmentation using magnetic resonance imaging (MRI) data. To address challenges such as high computational time and the requirement of huge storage and resources, we proposed BrainSegNet, which is a lightweight extension of U-Net with progressively expanded neurons that require fewer weights and less memory space. Unlike other DL models, our proposed model has the simplest architecture and is more accurate than other state-of-the-art methods for brain tumor segmentation. The proposed approach was extensively analyzed using the BraTS2020 benchmark dataset for segmenting brain tumors. The experimental findings demonstrate the effectiveness of the proposed system, producing a 96.01% Dice score and 95.89% mean IoU for brain tumor segmentation from brain MRI images.