Introduction:
A significant portion of patients with acute minor stroke have poor functional outcome due to early neurological deterioration (END). The purpose of this study is to investigate the applicability of machine learning algorithms to predict END in patients with acute minor stroke.
Methods:
We collected clinical and neuroimaging information from patients with acute minor stroke with NIHHS score of 3 or less. Early neurological deterioration was defined as any worsening of NIHSS score within three days after admission. Poor functional outcome was defined as a modified Rankin Scale score of 2 or more. We also compared clinical and neuroimaging information between END and No END group. Four machine learning algorithms, i.e., Boosted trees, Bootstrap decision forest, Deep learning, and Logistic Regression, are selected and trained by our dataset to predict early neurological deterioration
Results:
A total of 739 patients were included in this study. Seventy-eight patients (10.6%) had early neurological deterioration. Among 78 patients with END, 61 (78.2%) had poor functional outcomes at 90 days after stroke onset. On multivariate analysis, NIHSS score on admission (P=0.003), hemorrhagic transformation(P=0.010), and stenosis (P=0.014) or occlusion (P=0.004) of a relevant artery were independently associated with END. Compared with four machine learning algorithms, Boosted trees, Deep learning, and Logistic Regression achieved an excellent prediction of END in patients with acute minor stroke (Boosted trees: accuracy = 0.966, F1 score = 0.8 and an area under the curve value = 0.934, Deep learning :0.966, 0.8, 0. 904, and Logistic Regression : 0.966, 0.8, 0.885).
Conclusions:
This study suggests that machine learning algorithms which integrate clinical and neuroimaging information accurately predict END in patients with acute minor ischemic stroke. Further studies based on an extensive data set are needed to predict END accurately for treatment strategies and better functional outcome.