Multimodal machine learning for early risk stratification of post-stroke cognitive impairment.
A machine-learning model combining brain imaging and clinical data shows strong ability to predict cognitive impairment after stroke, enabling early intervention for at-risk patients.
A stacking ensemble ML model trained on 1070 AIS patients integrating neuroimaging, clinical, and demographic features predicts post-stroke cognitive impairment with strong external validation AUC (0.9049), potentially enabling early identification of at-risk patients for targeted prevention. Notably, the internal AUC of 0.9972 likely reflects some overfitting, and independent multi-center replication is needed before clinical deployment.
What the study was
- Study design
- Retrospective cohort study with ML model development and external validation
- Population
- Acute ischemic stroke patients (AIS) admitted to Lianyungang First People's Hospital, Jan 2020–Aug 2023
- Sample size
- 1070
- Category
- AI/ML in Clinical Diagnostics
- Maturity
- Exploratory
- Journal
- Journal of Alzheimer's Disease
Why it surfaced
Multimodal ML approach with reasonable external validation for PSCI prediction in a clinically important problem; single-center with suspiciously high internal AUC (0.9972) signals potential overfitting; external AUC 0.9049 is more realistic but still strong.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.