An AI-integrated Enhanced Anatomical Brain Tumor Segmentation in Substantia Nigra for Early Disease Identification using Deep Resnet50-CNN

The early identification of brain tumours is crucial for improving patient prognosis and treatment planning. Recent advancements in neuroimaging techniques have paved the way for the enhanced detection and characterization of these lesions through CT or MRI image scans. During medical image analysis, the feature analysis is important factor to know the object relations, entity resolution information and disease relational properties, from the prior analysis, machine learning models are mostly covered to segment the disease region by retaining non scaling factors. Due to lower precision class, poor segmentation leads low accuracy to identifying the disease regions to produce higher false rates. In this research explores a novel approach to brain tumour identification utilizing Neuromelanin-Sensitive Magnetic Resonance Imaging (NMS-MRI). We propose a novel preprocessing methodology leveraging the Weigner filter to enhance image normalization and overall quality, facilitating more accurate subsequent analyses. Following preprocessing, we employ Cascading Slice Window Volumetry Segmentation (CSWVM) to achieve refined tumor delineation in three-dimensional MRI scans. The segmentation results are further refined through Cross Modality Particle Swarm Optimization (CMPSO)-based feature selection, which effectively identifies the most relevant features across various imaging modalities to optimize the classification process. To classify the detected tumors, further to integrate a ResNet 50-Convolutional Neural Network (ResNet50-CNN), leveraging its deep learning capabilities for robust pattern recognition in neuroimaging data. The proposed system proves higher performance in precision, recall, f1 score by attaining feature selection, and deep learning algorithms, ultimately contributing to enhanced accuracy and effectiveness in brain tumor identification.