Combining clinical and imaging data for predicting functional outcomes after acute ischemic stroke: an automated
Hongju Jo1, Changi Kim2, Dowan Gwon3
1Department of CGMS Sensor, Sensor R&D Center, i-SENS, Seoul, Republic of Korea.
Scientific Reports
|October 7, 2023
Summary
An automated machine learning system accurately predicts 3-month functional outcomes for acute ischemic stroke patients. Combining clinical and neuroimaging data, this advanced model outperforms traditional risk scores for better stroke outcome prediction.
Area of Science:
- Neurology
- Artificial Intelligence
- Medical Informatics
Background:
- Predicting functional outcomes after acute ischemic stroke (AIS) is crucial for patient management.
- Traditional risk-scoring models have limitations in accurately forecasting long-term recovery.
Purpose of the Study:
- To develop and validate an automated machine learning (ML) system for predicting 3-month functional outcomes in AIS patients.
- To integrate clinical and neuroimaging features for enhanced predictive accuracy.
Main Methods:
- Developed three ML models: clinical features only (Model_A), neuroimaging data only (Model_B), and integrated features (Model_C).
- Utilized a multicenter stroke registry dataset of 4147 patients.
- Compared ML models against each other and traditional risk-scoring methods.
Main Results:
- The integrated ML model (Model_C) achieved the highest predictive performance with an area under the curve (AUC) of 0.786.
- Key predictors included age, initial National Institutes of Health Stroke Scale (NIHSS), and early neurologic deterioration.
- The integrated models demonstrated superior performance over traditional risk-scoring models.
Conclusions:
- An automated ML system combining clinical and neuroimaging data significantly improves the prediction of functional outcomes in AIS patients.
- This integrated approach offers a more accurate and robust tool compared to existing risk-scoring models.
