A study originating from a final cycle project at the Institut Roger de Llúria of Barcelona, with collaboration from professionals at IDI·Metropolitana Sud, has analyzed the possibilities of radiomics combined with artificial intelligence to improve the characterization and monitoring of brain tumors. The review, published in the Journal of Medical Imaging and Radiation Sciences, highlights the role of medical imaging technicians and points out the pending challenges for its clinical application.
Radiomics is a technique that extracts hundreds of quantitative parameters from medical images (MRI, CT, PET) that are imperceptible to the human eye but can provide valuable information about tumor heterogeneity. The work compiles published evidence from 2012 to 2024 and emphasizes the importance of medical imaging technicians' knowledge to ensure the quality and reproducibility of subsequent analyses.
According to the reviewed literature, this technique could facilitate the classification of gliomas before surgery and the non-invasive characterization of key molecular alterations. Although previous studies have reported high accuracies in differentiating molecular subtypes, the authors stress that these results come from research models not yet widely validated externally. If their performance is confirmed, they could reduce the need for biopsies and promote more personalized treatments.
The review also addresses how machine learning models applied to imaging could help distinguish true tumor progression from pseudoprogression, thus avoiding unnecessary therapeutic modifications. The potential for anticipating treatment response and survival, as well as radiogenomics, is also explored, although these applications are still in the research phase and are not yet a reliable alternative to histopathological confirmation.
The authors insist that radiomics is not yet a consolidated clinical practice. They identify methodological and ethical challenges such as the lack of standardized protocols, variability in segmentation, and the scarcity of validated multicenter studies. They recommend harmonizing procedures, working with open datasets, and establishing governance frameworks to balance information exchange and confidentiality. Prospective multicenter studies and external validation are needed before its routine incorporation.
