• Abstract

This article is dedicated to modern approaches for detecting glial tumors, particularly astrocytoma, oligodendroglioma, ependymoma, and glioblastoma, using Magnetic Resonance Imaging (MRI). MRI is one of the most essential tools in medical diagnostics, providing high-resolution visualization of tissue structures. Due to the unclear and infiltrative nature of glial tumors, fully detecting them remains a complex task. The paper analyzes MRI modalities such as T1, T2, FLAIR, and DWI, as well as AI-based automated detection models (CNN, U-Net), 3D spline interpolation, and radiomics technologies. Challenges such as difficulty in detecting low-grade gliomas, MRI contrast limitations, and segmentation errors are discussed. At the end of the article, modern technological suggestions are presented to address these challenges. The study aims to automate MRI-based diagnostics and improve accuracy, contributing significantly to clinical medicine.

The article also explores approaches for automatically identifying the location and grade of tumors using deeply studied machine learning models. In particular, it highlights methods for determining the structure and volume of gliomas from MRI images using Convolutional Neural Networks (CNN) and the U-Net segmentation architecture. Through radiomics, it becomes possible to evaluate the malignancy level of tumors based on hundreds of statistical and texture features extracted from MRI images. Furthermore, the advantages of using models based on 3D spline interpolation to improve segmentation accuracy and reduce edge detection errors are discussed. This article provides both theoretical and practical foundations for automating diagnostic processes and developing clinical decision support systems.