Texture analysis and radiomics in the diagnosis of multiple sclerosis: a review

The clinical signs of multifocal brain lesions, including multiple sclerosis, are highly variable and largely depend on lesion site and size. Differential diagnosis of such changes may be challenging in certain cases. Vascular, inflammatory, infectious, and hereditary diseases may demonstrate similar magnetic resonance imaging patterns, whereas their assessment is limited by technical factors and human visual perception. In recent years, novel approaches such as texture analysis and radiomics have been increasingly integrated into radiological research, facilitating the acquisition of imaging details that would otherwise remain undetectable by the naked eye. These methods include first-order statistical analysis of signal intensities, gray-level co-occurrence and gray-level run-length matrices, fractal and wavelet analyses, and the development of predictive models using machine learning algorithms. Radiomics was initially developed for oncologic imaging; however, now its capabilities are also applied in the diagnosis of other conditions.

This article presents a review of the current scientific data on the use of texture analysis and radiomics in the differential diagnosis of demyelinating diseases, with a particular focus on multiple sclerosis. Data search was conducted in PubMed and eLibrary using the keywords “radiomics,” “digital image texture analysis,” “multiple sclerosis,” “радиомика” (radiomics), “текстурный анализ” (texture analysis), and “рассеянный склероз” (multiple sclerosis). The search period covered the last 9 years. Only original studies (n = 17) investigating the use of radiomics and digital image texture analysis in the diagnosis of demyelinating diseases were included in this review.

Texture analysis and radiomics represent promising adjunctive tools for the evaluation of multifocal brain lesions in demyelinating diseases. However, their implementation in clinical practice requires the development of optimized feature extraction algorithms, identification of the most informative texture parameters, and standardization and validation of the resulting imaging biomarkers.