A Deep Learning Model for Brain Tumor Segmentation with Anatomical Localization and Treatment Guidance-SL-UNet Approach

Brain tumors are unnatural clusters of cells within the brain, classified as either noncancerous or cancerous.Timely identification and precise categorization are essential for effective treatment and better patient results. Anticipating brain tumors enhances survival rates and minimizes the need for invasive diagnostic methods. To predict brain tumors, it is important to gather and preprocess MRI or CT scan data to improve image quality. Current algorithms used for brain tumor prediction and segmentation include Convolutional Neural Networks, VGG, ResNet, Efficient Net, SVM, Random Forests, SegNet, Unet, S-Net and Attention UNet. The primary limitations of these algorithms are their requirement for large datasets, significant computational power, lengthy training times, and high resource consumption. Therefore, we suggest utilizing the SL-UNet model for brain tumor prediction because of its efficiency in medical image segmentation. SL-UNet is specifically built to capture both high-resolution spatial details and contextual features, making it very accurate in identifying exact tumor boundaries and delineating lobe regions such as the Frontal Lobe, Parietal Lobe, Temporal Lobe, and Occipital Lobe and recommend the treatment planning for the tumor by using the user details, like height, weight, sugar levels and BP levels. The benefits of SL-UNet include its capability to work effectively with smaller datasets, efficient training with fewer parameters, and robust performance even when dealing with intricate and noisy medical images.