Clinical Scorecard: AI may predict focal GBM recurrence

At a Glance

Key Highlights

• The random forest model achieved an AUC of 0.87 during 10-fold internal cross-validation and 0.80 in a separate 20-patient internal validation cohort.
• AI-derived infiltration alone had a mean AUC of 0.62; histopathologic infiltration alone had a mean AUC of 0.59.
• AI-derived infiltration was the leading predictor in 5 of 6 classifiers.
• A reduced random forest model using AI-derived infiltration, patient age, and resection cavity volume had a mean cross-validation AUC of 0.82.
• SRH imaging of unprocessed tissue took less than 60 seconds per sample.

Guideline-Based Recommendations

Diagnosis

• The study included patients with pathology-confirmed IDH-wild-type GBM; it evaluated recurrence at sampled margin sites against subsequent imaging.

Management

• The study evaluated an investigational AI-assisted approach and did not provide treatment recommendations.

Monitoring & Follow-up

• Follow-up imaging was assessed using Response Assessment in Neuro-Oncology 2.0 criteria.
• Spatial recurrence estimates within 5 mm and 10 mm of sampled sites were based on 10-fold internal cross-validation, not the separate validation cohort.

Risks

• The study did not standardize the number or distribution of margin samples and could not characterize infiltration between sampled locations.
• Functional mapping and proximity to eloquent brain regions constrained sampling.
• All development and validation patients underwent maximal resection; performance in unresectable GBM is unknown.
• The separate validation cohort came from the same health system as the development cohort.

Patient & Prescribing Data

Retrospective development cohort: 60 adults with 273 margin samples; 43 patients experienced recurrence. Separate prospective internal validation cohort: 20 patients with 98 margin samples.

All patients in the development and validation cohorts underwent tumor resection. The study assessed prediction of focal recurrence and did not report prescribing or treatment-effect data.

Clinical Best Practices

• Interpret the reported AUCs in the context of the study design: the separate validation cohort was internal and from the same health system.
• Recognize that margin sampling was not standardized and was limited by functional mapping and eloquent brain regions.
• Do not infer model performance in patients with unresectable GBM from these cohorts.

Related Resources & Content

• Science Advances study described in the source article