Prediction of outcomes following intravenous thrombolysis in patients with acute ischemic stroke using serum UCH-L1, S100β, and NSE: a multicenter prospective cohort study employing machine learning methods
Background: Acute ischemic stroke (AIS) is a leading cause of mortality and disability worldwide. Intravenous thrombolysis (IVT) improves recovery, but predicting outcomes remains challenging. Machine learning (ML) and biomarkers like ubiquitin carboxyl-terminal hydrolase L1 (UCH-L1), S100 calcium-b...
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SAGE Publishing
2025-06-01
|
| Series: | Therapeutic Advances in Neurological Disorders |
| Online Access: | https://doi.org/10.1177/17562864251342429 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Background: Acute ischemic stroke (AIS) is a leading cause of mortality and disability worldwide. Intravenous thrombolysis (IVT) improves recovery, but predicting outcomes remains challenging. Machine learning (ML) and biomarkers like ubiquitin carboxyl-terminal hydrolase L1 (UCH-L1), S100 calcium-binding protein β (S100β), and neuron-specific enolase (NSE) may enhance prognostic accuracy. Objectives: We aimed to assess the predictive value of serum brain injury biomarkers for 3-month outcomes in AIS patients treated with IVT, using an ML-based model. Design: A multicenter prospective cohort study was conducted, enrolling AIS patients treated with recombinant tissue plasminogen activator from 16 hospitals. Methods: Of 1580 patients, 1028 were included and divided into training ( n = 571), testing ( n = 243), and external validation ( n = 214) cohorts. Thirty-three variables, including demographics, clinical data, and biomarkers (UCH-L1, S100β, NSE), were analyzed. Least Absolute Shrinkage and Selection Operator regression was used for feature selection, and six ML algorithms were tested. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), F 1-score, calibration curve, and decision curve analysis. Results: The light gradient boosting machines (LightGBM) model achieved the best performance in the training dataset (AUC: 0.846; F 1-score: 0.789) and external validation dataset (AUC: 0.714). Eight critical predictors, including age, admission National Institutes of Health Stroke Scale (NIHSS) score, Trial of Org 10172 in Acute Stroke Treatment, white blood cell, finger blood glucose, UCH-L1, S100β, and NSE, were identified and incorporated into an ML model for clinical application. Shapley additive interpretation analysis enhances the interpretability of the model, with NIHSS score and NSE as top contributors. External validation confirmed good calibration and consistent net benefit across threshold probabilities (0.1–0.8). Conclusion: Integrating serum biomarkers (UCH-L1, S100β, NSE) with ML significantly improves 3-month outcome prediction in AIS patients. The LightGBM model offers robust performance and clinical interpretability for individualized treatment planning. |
|---|---|
| ISSN: | 1756-2864 |