An Efficient of Artificial Intelligence based Brain Tumor Diagnosis and Classification: An Advance Medical Diagnosis Approach

Abstract

As a central regulatory system of the body, the human brain is susceptible to the abnormal proliferation of cells that may cause brain tumors and become a significant threat to health and life. The accurate classification and early identification of these tumors are critical to attaining effective treatment schemes and better prognosis of these patients. Although the traditional diagnostic methods, including biopsies and radiological (CT, PET and MRI) are still good, they have some shortcomings, such as invasive nature, subjective and interpreter’s variability. As a solution to these shortcomings, machine learning (ML)-based and deep learning (DL)-based computer-aided diagnostic systems have become potent tools in automated tumor classification, detection and segmentation across different medical imaging modalities. The present review discusses the current trends of ML and DL models, focusing especially on MRI-based segmentation and classification. It compares the conventional ML systems, which include Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Random Forests, and Extreme Learning Machines, with the contemporary DL systems, which include Convolutional Neural Networks (CNNs), ResNet, and transformer-based systems. Moreover, it discusses popular datasets, metrics of evaluation, and segmentation, starting with simple thresholding and clustering to more sophisticated DL-based architectures. The paper makes a comparative evaluation of these methodologies, research gaps and emphasizes the increasing significance of 3D modeling techniques and attention mechanisms in enhancing diagnostic performance. The results showed that DL-based methods especially CNN-based models are always better at tumor detection and segmentation than traditional ML methods and provide more reliable, efficient, and clinically useful solutions to brain tumor diagnosis.