Every year, 3 people in 100,000 are diagnosed with glioma, the most common primary malignant brain tumor in adults. Yet behind this single word lies a mosaic of very different diseases, each with its own genetic fingerprint, prognosis, and response to treatments.
What if MRI scanning could already know the molecular composition of tumors, before any surgery, before any biopsy? In neuro-oncology, the molecular profile of a glioma is not a scientific curiosity. It is a clinical necessity. The presence or absence of certain genetic mutations determines prognosis and guides therapeutic choices.
Yet obtaining this molecular information remains an invasive, time-consuming, and sometimes incomplete process. Tumor tissue must be surgically removed, processed, then sequenced. For patients who cannot be operated on, or whose tumor is located in functional brain areas, this step may simply be impossible.
A question arises: can advanced MRI techniques replace, or at least anticipate, the molecular biology laboratory? Recent research suggests the answer may be closer to “yes” than previously thought.
Why Is the Molecular Profile of Gliomas So Important?
Gliomas are today classified not only according to their appearance under the microscope, but also according to their genetic fingerprint. Among the most clinically significant markers are IDH (isocitrate dehydrogenase) gene mutations, MGMT promoter methylation, TERT promoter mutations, and co-deletion of chromosomal arms 1p and 19q.
Each of these alterations tells a different story. IDH mutations are associated with prolonged survival and better response to both chemotherapy and radiotherapy. MGMT methylation predicts sensitivity to a specific class of alkylating agents, the backbone of most chemotherapy protocols in glioblastoma. TERT mutations, conversely, are linked to more aggressive behavior and poorer outcomes in many glioma subtypes. Finally, 1p/19q co-deletion, combined with an IDH mutation, defines the oligodendroglioma subtype, a cancer with a notably favorable prognosis.
The clinical implications are direct: two patients presenting with radiologically indistinguishable tumors may require entirely different treatments. Knowing molecular status from the outset is not purely academic, it saves time, guides biopsy planning, and could one day enable true preoperative personalized decision-making.
Radiomics and Multiparametric MRI: Extracting What the Eye Cannot See
A first approach relies on the concept of radiomics, the extraction of hundreds of quantitative features from standard MRI sequences, invisible to the human eye: shape irregularities, subtle intensity variations, texture patterns at different spatial scales.
Comparing the predictive power of individual MRI sequences (T1-weighting, T2, diffusion, apparent diffusion coefficient, and contrast-enhanced T1) with multi-sequence models combining features from all five individual sequences shows positive results, with the combined model consistently outperforming any individual sequence for IDH, MGMT, and TERT prediction. However, no significant advantage is demonstrated for 1p/19q co-deletion. This finding underscores the particular challenge posed by this specific alteration with this type of imaging.
This radiomic approach translates into real clinical benefit in improving therapeutic decisions based on glioma biomarkers.
CEST MRI: Reading the Chemistry of the Tumor
While radiomic approaches explore the spatial texture of conventional sequences, a second technique poses a fundamentally different question: what is the biochemical microenvironment of the tumor, and can it be measured non-invasively?
Chemical exchange saturation transfer (CEST) MRI exploits the exchange of protons between water molecules and neighboring chemical groups, specifically amide bonds (–NH, resonating around 3.5 ppm) and amine groups (–NH₂, around 2 ppm). These signals, present in endogenous proteins and free amino acids, are extremely sensitive to local concentrations and pH. Since tumor metabolism differs markedly from normal brain tissue, and IDH-mutant gliomas produce an aberrant oncometabolite (2-hydroxyglutarate) that acidifies the cellular environment, CEST offers a window into molecular biology that structural imaging cannot provide.
The amide signal and amine signal prove to be a robust discriminator between IDH wild-type gliomas (gliomas without IDH mutation) and IDH-mutant gliomas, with a generally higher amide/amine ratio for IDH wild-type gliomas. The amine signal alone allows differentiation of tumors with intact 1p/19q from those with 1p/19q co-deletion, a comparison that radiomics could not achieve.
CEST imaging provides the molecular information necessary for diagnostic differentiation, in turn delivering clinical benefit in patient management.
Complementary Tools for a Common Goal
Read independently, each technique offers a compelling proof of concept. Read together, they illuminate a broader landscape.
Radiomic approaches exploiting standard clinical sequences are already accessible on present equipment with existing workflows. The computational burden is manageable, feature extraction pipelines are increasingly standardized, and the integration of clinical variables is straightforward.
CEST MRI operates in a different register. It requires specialized acquisition sequences, dedicated post-processing pipelines, and a level of physics expertise that remains concentrated in research centers. But it accesses information that is invisible to conventional imaging, and even to diffusion imaging.
What both approaches share is a commitment to the same vision: replacing or reducing dependence on surgical tissue sampling for molecular diagnosis, making MRI a smarter tool for tumor biology.
Toward the Growth of Advanced MRI Techniques in Clinical Practice
The question of whether MRI can predict the molecular identity of a glioma no longer seems theoretical. Current evidence argues for a nuanced “yes”: with the right combination of sequences, features, and models, non-invasive molecular characterization is within reach for the most clinically relevant biomarkers, with a nuance for 1p/19q co-deletions.
Further study work remains to transform this nuanced “yes” into a robust “yes” and to move these models from research tools to clinical instruments. However, the results are encouraging and offer strong potential for replacing biopsy with advanced MRI techniques in the future.
For radiologists, the practical implication is already taking shape: reporting on a glioma MRI will increasingly need to incorporate quantitative and molecular dimensions, not just morphological ones. For the industry, the opportunity lies in developing validated, regulatory authority-approved software capable of integrating radiomics and advanced quantitative MRI into routine neuro-oncology workflows.
Sources:
Mancini L, Casagranda S, Gautier G, Peter P, Lopez B, Thorne L, McEvoy A, Miserocchi A, Samandouras G, Kitchen N, Brandner S, De Vita E, Torrealdea F, Rega M, Schmitt B, Liebig P, Sanverdi E, Golay X, Bisdas S. CEST MRI provides amide/amine surrogate biomarkers for treatment-naïve glioma sub-typing. Eur J Nucl Med Mol Imaging. 2022 Jun;49(7):2377-2391. doi: 10.1007/s00259-022-05676-1. PMID: 35029738; PMCID: PMC9165287. https://pmc.ncbi.nlm.nih.gov/articles/PMC9165287/
He J, Ren J, Niu G, Liu A, Wu Q, Xie S, Ma X, Li B, Wang P, Shen J, Wu J, Gao Y. Multiparametric MR radiomics in brain glioma: models comparison to predict biomarker status. BMC Med Imaging. 2022 Aug 5;22(1):137. doi: 10.1186/s12880-022-00865-8. PMID: 35931979; PMCID: PMC9354364. https://pubmed.ncbi.nlm.nih.gov/35931979/
Les tumeurs gliales diffuses de l’adulte. 2012. https://www.medecinesciences.org/en/articles/medsci/full_html/2012/10/medsci20122810p813/medsci20122810p813.
html#:~:text=Les%20gliomes%20ou%20tumeurs%20gliales,personnes%20par%20ann%C3%A9e%20%5B1%5D.
