Obituary: old-school glioblastoma immune profiling, age "surprisingly recent," passed away after a long struggle with invasive tissue sampling, partial snapshots, and the general inconvenience of needing actual tumor pieces every time biology changed its mind.
That is the setup for a new paper asking a very 2026 question: can an MRI do more than show where a glioblastoma is? Can it also hint at what the patient’s immune system is doing out in the bloodstream?
A team led by Johanna Heugenhauser and colleagues thinks the answer might be yes - at least enough to make the field sit up a little straighter and stop pretending scans are just fancy photographs [1].
Your MRI might be hiding a second job
Glioblastoma is the most common malignant primary brain tumor in adults, and it is a brutal operator. It grows fast, rewires its neighborhood, and builds an immune environment that feels less like "defense system activated" and more like "security team locked outside the building" [4,5].
Radiomics is the trick researchers use to squeeze extra information out of scans. Instead of a radiologist saying, "that edge looks thick" or "that area looks messy," radiomics turns the image into a giant spreadsheet of texture, shape, intensity, and pattern features. Basically, the MRI stops being a picture and becomes telemetry. Very software-product-launch energy [2,3].
Most radiomics work in glioma has focused on things like diagnosis, prognosis, molecular markers, or what is happening inside the tumor microenvironment [2,3]. This new study pushes the idea further. It asks whether imaging features from the tumor can reflect systemic immune activity - meaning signals in the blood, not just inside the tumor zip code [1].
The pitch deck, but for immune cells
The researchers looked at 34 glioblastoma patients with both MRI scans and detailed blood immune profiling. They extracted 321 radiomic features from three tumor regions: the necrotic core, the contrast-enhancing margin, and the T2/FLAIR-bright peritumoral zone. Then they matched those against 67 blood-based immune markers measured by flow cytometry and PCR [1].
This is where the paper gets fun, in the "scientists built a weird matchmaking app for MRI pixels and T cells" sense.
They found that a radiomics feature from the T2 hyperintense zone seemed to predict blood levels of T helper 17 cells. Features from the necrotic core were linked to RORγT, a transcription factor involved in Th17 biology, and to CD15+ myeloid cells in the blood. More importantly, multivariable models could delineate broader immune activation signals, including naive and activated CD8+ T cells, early-differentiated CD8+ T cells, CD56+ natural killer cells, and T-bet, a transcription factor associated with a type 1 T-cell response [1].
Translated into regular human language: the scan may contain clues about whether the immune system is geared up, worn down, or behaving in a particular style. That is a big deal, because blood immune states and tumor immune states both matter when you are trying to figure out who might benefit from immunotherapy and who might not [4,5].
Why this is actually interesting, not just "AI found a pattern"
The coolest part is not that a machine learning model found correlations. Models do that all day. The interesting part is the biological bridge.
Glioblastoma is not only a lump in the brain. It also pushes systemic immune changes across the rest of the body. Earlier work from some of the same investigators already hinted that conventional MRI features could track with blood immune markers [2]. This new paper adds a more granular radiomics layer and suggests that the tumor's visual fingerprint may echo broader immune activity [1].
If that holds up in larger studies, you can see the product roadmap immediately. A routine MRI could help flag immune states without another invasive procedure. That could make patient stratification smarter, especially in immunotherapy trials where picking the right patient is half the battle and the other half is glioblastoma acting like technical debt with blood supply [4,5].
It could also help with longitudinal monitoring. Tumors evolve. Immune responses evolve. Repeated brain biopsies are not exactly a casual errand between lunch and dentist. A noninvasive readout would be much easier to scale.
Before we declare victory and ring the NASDAQ bell
There is a reason the authors call this hypothesis-generating. The cohort was small. The study was retrospective. And radiomics as a field still has reproducibility headaches involving imaging protocols, segmentation choices, and the eternal machine-learning question: does this work somewhere else, or just in the building where it was born? [1-3]
That caution matters. Glioblastoma research has a graveyard full of "promising signals" that later ran into biology, statistics, or both. Immunotherapy in glioblastoma has been especially humbling, with many phase III efforts failing despite solid hype and good intentions [4,5].
Still, this paper lands on a genuinely useful idea. Instead of treating MRI and immunology like two coworkers who only email in emergencies, it suggests they might belong on the same dashboard.
And in a cancer that has spent years making our best treatments look underpowered, any tool that reads more from the data you already collect deserves attention.
References
-
Heugenhauser J, Visus C, Buchroithner J, Marosi C, Rössler K, Felzmann T, Widhalm G, Iglseder S, Nowosielski M, Erhart F. Glioblastoma radiomics can delineate systemic immune activity states like blood abundance of T cell populations or transcription factors. Journal of Neuroinflammation. 2026. DOI: 10.1186/s12974-026-03806-2
-
Khalili N, Fathi Kazerooni A, Familiar A, Haldar D, Kraya A, Foster J, et al. Radiomics for characterization of the glioma immune microenvironment. npj Precision Oncology. 2023;7:59. DOI: 10.1038/s41698-023-00413-9. PMCID: PMC10279670
-
Singh G, Manjila S, Sakla N, True A, Wardeh AH, Beig N, et al. Radiomics and radiogenomics in gliomas: a contemporary update. British Journal of Cancer. 2021;125:641-657. DOI: 10.1038/s41416-021-01387-w
-
Arrieta VA, Dmello C, McGrail DJ, Brat DJ, Lee-Chang C, Heimberger AB, et al. Immune checkpoint blockade in glioblastoma: from tumor heterogeneity to personalized treatment. Journal of Clinical Investigation. 2023;133(2):e163447. DOI: 10.1172/JCI163447
-
Liu Y, Zhou F, Ali H, Lathia JD, Chen P. Immunotherapy for glioblastoma: current state, challenges, and future perspectives. Cellular & Molecular Immunology. 2024;21:1354-1375. DOI: 10.1038/s41423-024-01226-x
Disclaimer: The image accompanying this article is for illustrative purposes only and does not depict actual experimental results, data, or biological mechanisms.