Enhancing Brain Tumor Assessment: A Comprehensive Approach using Computerized Diagnostic Tool and Advanced MRI Techniques

Abstract

Assessment of brain tumour using Three-Dimensional Magnetic Resonance Imaging (3D MRI) is computationally multifaceted task. Currently, hospitals employ 2D MRI scans, followed by manual evaluation by experienced doctors, aided by a Computerized Diagnostic Tool (CDT). This research aims to develop an advanced CDT to significantly enhance the accuracy of brain tumor assessment. The CDT presented in this study evaluates Axial-View (AV), Coronal-View (CV), and Sagittal-View (SV) MRI images. It encompasses a comprehensive pipeline, including pre-processing, post-processing, feature extraction, feature selection, and categorization phases. Various tumor segmentation techniques, including active contour, level-set, watershed, and region growing, are thoroughly explored. Additionally, a comparative analysis of classification methods such as SVM, ANFIS, k-NN, Random Forest, and Adaboost is conducted. Experimental validation using the BRATS 2016 dataset and real-time 2D MRI data demonstrates that the proposed CDT consistently achieves an average classification accuracy exceeding 95% in tumor-based categorization. This research represents a significant advancement in brain tumor assessment, leveraging machine learning and advanced MRI techniques to improve diagnostic precision.