Radiomics models predict IDH mutation status and survival in glioma patients across discovery and validation cohortsCan scans predict glioma risks without surgery? New models show promise but need more proof

Frontiers in Medicine Published April 6, 2026 DOI ↗ Editorial oversight: Dr. Julia Lee, PhD · Oncology, Genomics & Drug Development

AI-generated summary of the cited source, checked by automated accuracy review. How we work

Key Takeaway

Consider radiomics models as exploratory tools for IDH prediction, noting reduced validation performance and lack of safety data.

This multicenter cohort study included 638 patients with gliomas, comprising 213 from a local discovery institution and 425 from a public validation dataset. The researchers compared radiomics-based machine learning models against standard clinical features, including age, sex, WHO grade, and IDH mutation status. The primary outcomes assessed were the prediction of IDH mutation status and overall survival.

In the discovery cohort, the radiomics model achieved an area under the receiver operating characteristic curve (AUC) of 0.90 for predicting IDH mutation status. For overall survival, median duration was 21 months for high-risk patients versus 30 months for low-risk patients, a difference that was statistically significant (P < 0.05). In the validation cohort, the IDH prediction AUC decreased to 0.68. Median overall survival for high-risk versus low-risk patients was 10 versus 19.5 months, also representing a statistically significant difference (P < 0.05).

Safety data, adverse events, and discontinuations were not reported. The study did not report specific limitations, funding sources, or conflicts of interest. Follow-up duration was not reported. While the models showed promise in the discovery setting, the reduced AUC in the validation cohort suggests potential variability in generalizability. Clinicians should interpret these findings as exploratory evidence rather than established diagnostic tools for routine practice.

Gliomas are serious brain tumors where knowing the genetic details helps doctors choose the right care. Often, getting this information requires a biopsy, which carries risks. This study looked at whether computer programs could read MRI scans to predict these genetic details and survival chances without invasive procedures. The team tested these tools on 638 tumors from two different groups of patients. In the first group, the computer model guessed the genetic status correctly most of the time. However, when tested on a second, larger group of patients from a public database, the model was much less accurate.

The study also looked at how long patients lived based on the computer's risk scores. In both groups, patients marked as high-risk had significantly shorter lives than those marked as low-risk. This shows the models can spot dangerous patterns. But because the accuracy dropped in the second group, we cannot say these tools are ready to replace standard tests. The results are a start, but they are not enough to change how doctors treat patients today.

This research is a step forward in using technology to help brain cancer patients. However, the drop in accuracy between the two groups means we must be careful. Until these tools work consistently in many different hospitals, they remain an interesting idea rather than a practical solution. Patients should continue to rely on standard medical advice while researchers keep improving these digital helpers.

What this means for you:

Computer models can predict some glioma risks from scans, but accuracy varied greatly between groups, so they are not ready for patient care yet.

Study Details

Study typeCohort

EvidenceLevel 3

PublishedMar 2026

View Original Abstract ↓

IntroductionGliomas are infiltrative primary intracranial tumors with marked biological and clinical heterogeneity. Prognosis varies widely and depends on tumor grade, histopathological characteristics, and molecular alterations. Isocitrate dehydrogenase mutation is a key prognostic biomarker and is associated with improved treatment response and longer overall survival. Radiomics enables the extraction of quantitative features from routinely acquired medical images. This study evaluated radiomics-based machine learning models for noninvasive prediction of isocitrate dehydrogenase mutation status and overall survival in glioma patients.MethodsFrom T2-weighted MRI scans of 638 gliomas (213 from a local institution (discovery), 425 from a public dataset (validation)), 1,820 radiomics features were extracted. Machine learning models were constructed and trained on the discovery cohort and externally validated to predict isocitrate dehydrogenase mutation status. A radiomics risk score was computed using Lasso regression, and patients were stratified into high- and low-risk groups using the median radiomics risk score for Kaplan-Meier analysis. Cox regression assessed the prognostic value of radiomics risk score along with clinical features (age, sex, WHO grade, isocitrate dehydrogenase mutation status). A nomogram incorporating independent predictors to estimate 1-, 2-, and 3-year overall survival was assessed using the concordance index and calibration curves.ResultsLogistic regression and random forest classifier models achieved area under the receiver operating characteristic curve of 0.90 and 0.68 in the discovery and validation cohorts, respectively, for isocitrate dehydrogenase mutation prediction using 12 top radiomic features. High-risk patients showed significantly shorter median overall survival than low-risk patients in both discovery and validation cohorts (21 vs. 30 months, 10 vs. 19.5 months, respectively; P

Radiomics models predict IDH mutation status and survival in glioma patients across discovery and validation cohortsCan scans predict glioma risks without surgery? New models show promise but need more proof

Study Details

Experimental agents show response rates up to 70% in relapsed or refractory diffuse large B-cell lymphoma

New cancer therapies double response rates for tough lymphoma

Clinical research that matters. Delivered to your inbox.

Radiomics models predict IDH mutation status and survival in glioma patients across discovery and validation cohortsCan scans predict glioma risks without surgery? New models show promise but need more proof

More on Glioma

Study Details

Experimental agents show response rates up to 70% in relapsed or refractory diffuse large B-cell lymphoma

New cancer therapies double response rates for tough lymphoma

Clinical research that matters. Delivered to your inbox.

Related in Oncology

From Other Specialties