Enhanced Multiclass Brain Tumor Detection in MRI Using Similarity-Navigated Graph Neural Networks with Banyan Tree Growth Optimization

The diagnosis of brain tumors is admittedly difficult because of their varied and complicated nature. Therefore, one has to get the right diagnosis and categorization for it to be treated well. This is a situation where identification of multiple classes such as; meningioma, pituitary, glioma and no tumor is crucial in taking the right course of action. Therefore, a new deep learning architecture is suggested in this work to increase the specificity between different types of tumors by detecting brain tumors with high accuracy in MRI scans. The proposed methodology starts with the Cumulative Curve Fitting Approximation (CCFA) preprocessing algorithm to improve the quality of MRI images. Preprocessed images are used as inputs for feature extraction based on geometric algebra transformers. These transformers include geometric relationships within the data, ensuring more distinctive selection. The categorization is further carried out with assistance of a Similarity-Navigated Graph Neural Network (SNGNN) incorporating the Banyan Tree Growth Optimizer (BTGO) model, which is constructed for efficient detection of the subtlest patterns and their connection between classes of brain tumors. It integrates very well in terms of the accuracy shown by its performance metrics. It shows effectiveness in a multiclass scenario of detecting brain tumors, which helps in the precise and early diagnosis and even tailors proper treatment plans.