Deep Learning-Based Brain Age Predicts Stroke Recurrence in Acute Ischemic Disease
Overview
A novel deep learning model, MBA Net, estimates brain age from non-infarcted brain regions in acute ischemic cerebrovascular disease (AICVD) patients. The brain age gap (BAG) derived from this model independently predicts stroke recurrence at 3 months and 5 years, outperforming chronological age and enhancing existing risk models.
Background
Acute ischemic cerebrovascular disease is characterized by high rates of stroke recurrence despite current therapies, underscoring the need for improved risk stratification biomarkers. Brain age, derived from MRI using AI algorithms, reflects brain health and has been linked to stroke risk but its role in predicting recurrence remains unclear. Acute infarcts complicate brain age estimation during the acute phase due to dynamic lesion changes. Thus, focusing on non-infarcted brain regions may yield more reliable prognostic information.
Data Highlights
Dataset
Sample Size
Mean Absolute Error (MAE)
R²
Training (Healthy)
4265
1.58 years
0.92
Test (Healthy)
544
3.27 years
0.89
AICVD Patients (Inference)
10,890
Not specified
Not specified
Key Findings
MBA Net accurately predicts contextual brain age (CBA) by masking acute infarcts on T2-FLAIR MRI, achieving low MAE and high R² in healthy subjects.
BAG, defined as CBA minus chronological age, independently predicts stroke recurrence at both 3 months and 5 years post-event.
BAG outperforms chronological age alone in predicting stroke recurrence risk.
Incorporating BAG into established stroke recurrence prediction models significantly improves their discriminative performance.
The model was validated in a large, multicenter cohort of 10,890 AICVD patients, demonstrating clinical applicability.
Clinical Implications
The use of MBA Net to estimate brain age gap provides clinicians with a novel, objective biomarker for stratifying stroke recurrence risk in AICVD patients. This approach may guide more personalized secondary prevention strategies by identifying high-risk individuals beyond traditional clinical factors. Early integration of BAG into clinical workflows could improve long-term outcomes through targeted interventions.
Conclusion
MBA Net-derived brain age gap is a promising biomarker that independently predicts stroke recurrence and enhances existing risk models in acute ischemic cerebrovascular disease. This AI-driven approach supports precision medicine efforts for secondary stroke prevention.
References
Wang et al. 2024 -- Utilizing Deep Learning to Estimate Brain Age as a Predictor of Stroke Recurrence in Acute Ischemic Cerebrovascular Disease
