Objective To investigate the predictive value of artificial intelligence-based Alberta Stroke Program Early CT Score (ASPECTS) combined with diffusion-weighted imaging (DWI) cerebral infarct volume for poor prognosis in wake-up stroke (WUS) patients.

Methods A total of 100 patients with acute ischemic stroke after waking up with unknown time window admitted to Kaifeng Central Hospital from September 2022 to June 2023 were selected as the research objects. All patients underwent emergency non-contrast-enhanced cranial CT and magnetic resonance imaging (MRI) scan, followed by reperfusion therapy. The patients were followed up for 3 months after treatment, and were divided into good prognosismodified Rankin Scale(mRS) ≤ 2 and poor prognosis groups mRS >2 according to the mRS score. The baseline data, artificial intelligence ASPECTS, and DWI cerebral infarct volumes were compared between the two groups. Multivariate logistic regression analysis was used to identify prognostic factors, and receiver operating characteristic (ROC) curves were employed to evaluate the diagnostic efficacy of artificial intelligence ASPECTS combined with DWI cerebral infarct volume.

Results After 3 months of follow-up, the poor prognosis rate of patients was 32.00% (32/100). The artificial intelligence ASPECTS at admission in the poor prognosis group was lower than that in the good prognosis group, and the DWI cerebral infarction volume at admission was larger than that in the good prognosis group, with statistically significant differences (P < 0.05). The results of multivariate logistics analysis showed that age (OR=2.190; 95%CI, 1.412 to 3.398), blood pressure variability (OR=1.726; 95%CI, 1.192 to 2.500), homocysteine (OR=1.902; 95%CI, 1.268 to 2.854), D-dimer (OR=2.275; 95%CI, 1.274 to 4.064), white blood cell count (OR=2.614; 95%CI, 1.484 to 4.606), neutrophil-to-lymphocyte ratio (OR=2.921; 95%CI, 1.350 to 6.323), National Institutes of Health Stroke Scale score (OR=3.171; 95%CI, 1.754 to 5.731), and DWI infarct volume (OR=3.586; 95%CI, 1.634 to 7.869) were identified as factors affecting poor prognosis (P < 0.05), while high artificial intelligence ASPECTS was identified as a protective factor (OR=0.534; 95%CI, 0.352 to 0.810; P < 0.05). The sensitivity, specificity and area under the curve of the combined prediction model were 96.88%, 85.29% and 0.947, respectively. The sensitivity and AUC of the combined prediction model were higher than that of the single prediction (P < 0.05), and the specificity was similar to that of the single prediction.

Conclusion The combined application of artificial intelligence ASPECTS and DWI infarct volume significantly enhances predictive efficacy for poor prognosis in WUS patients, providing a more accurate prognostic evaluation tool for clinical decision-making, and it has the value of guiding personalized treatment.