Related Experiment Video
Updated: Jul 17, 2025
12:18
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
7.6K
Prediction of post-stroke cognitive impairment after acute ischemic stroke using machine learning
Minwoo Lee1, Na-Young Yeo2,3, Hyo-Jeong Ahn2,3
1Department of Neurology, Hallym University Sacred Heart Hospital, Hallym University, Anyang, South Korea.
Alzheimer'S Research & Therapy
|August 31, 2023
Summary
Machine learning models accurately predict post-stroke cognitive impairment (PSCI) in acute ischemic stroke (AIS) patients. Extreme gradient boost and artificial neural networks showed the highest predictive accuracy for cognitive outcomes.
Area of Science:
- Neurology
- Artificial Intelligence
- Medical Informatics
Background:
- Post-stroke cognitive impairment (PSCI) affects up to 50% of patients following acute ischemic stroke (AIS).
- Predicting cognitive outcomes in AIS is crucial for guiding treatment decisions.
- This study investigated the use of machine learning for PSCI prediction.
Purpose of the Study:
- To evaluate the efficacy of machine learning models in predicting PSCI after AIS.
- To compare the performance of different machine learning algorithms for cognitive outcome prediction.
Main Methods:
- Retrospective analysis of a prospective PSCI cohort of AIS patients.
- Inclusion of demographic, clinical, and neuroimaging variables.
- Development and comparison of four machine learning models: logistic regression, support vector machine, extreme gradient boost, and artificial neural network.
- Primary outcome: PSCI at 3-6 months, defined by VCIHS-NP criteria.
Main Results:
- 951 AIS patients were analyzed (mean age 65.7 years, 61.5% male).
- Extreme gradient boost (AUC 0.7919) and artificial neural network (AUC 0.7365) models demonstrated the highest predictive accuracy.
- Key predictors identified: cortical infarcts, mesial temporal lobe atrophy, initial stroke severity, stroke history, and strategic infarcts.
Conclusions:
- Machine learning algorithms, specifically extreme gradient boost and artificial neural networks, are effective in predicting cognitive outcomes post-ischemic stroke.
- These models offer a promising approach for identifying patients at risk of PSCI.
