Related Experiment Video
Updated: Jun 16, 2026
09:59
A Magnetic Resonance Imaging Protocol for Stroke Onset Time Estimation in Permanent Cerebral Ischemia
Published on: September 16, 2017
Using radiological brain health to predict recurrence after ischemic stroke and transient ischemic attack: A
David Robinson1, Pooja Khatri2, Heidi Sucharew3
1Department of Neurology, University of Cincinnati School of Medicine, Cincinnati, OH.
Neurology Open Access
|June 15, 2026
Summary
Neuroimaging data modestly improved prediction of recurrent stroke. Advanced models incorporating brain health markers like atrophy and white matter hyperintensities showed moderate gains in identifying patients at risk for stroke recurrence.
Area of Science:
- Neurology
- Medical Imaging
- Data Science
Background:
- Recurrent strokes pose significant morbidity and mortality risks.
- Existing clinical models have limited ability to predict stroke recurrence.
- Neuroimaging may capture subtle indicators of stroke risk.
Purpose of the Study:
- To evaluate if neuroimaging markers improve stroke recurrence prediction beyond clinical factors.
- To apply advanced machine learning for enhanced predictive modeling.
Main Methods:
- Ischemic strokes and TIAs were identified in a US population.
- Standard-of-care MRI data were collected and analyzed for brain health metrics.
- Random survival forests were used to build prediction models at 90 days and 3 years.
Main Results:
- Models with clinical and imaging data outperformed clinical data alone for 90-day and 3-year recurrence prediction.
- C-statistics improved significantly (P<0.001 at 3 years).
- Key imaging predictors included global cortical atrophy, microbleed count, and white matter hyperintensities.
Conclusions:
- Neuroimaging data offer a moderate improvement in predicting stroke recurrence.
- Further strategies, including dynamic models and non-imaging biomarkers, are needed for optimal risk identification.
