A Nomogram Based on Ocular Hemodynamics for Predicting Ischemic Stroke

Abstract

PURPOSE

Ischemic stroke is a cerebrovascular disease with high mortality and disability. Due to similar physiological characteristics, ocular vascular characteristics are important indicators for monitoring cerebrovascular diseases. This study aimed to develop a nomogram prediction model for ischemic stroke based on ocular hemodynamic characteristics.

DESIGN

Retrospective clinical cohort study.

METHODS

A total of 337 patients were included in this study and randomly divided into 235 training and 102 validation cohorts. The general data were collected, and the hemodynamic parameters of ophthalmic artery, central retinal artery and posterior ciliary artery were detected by ultrasound. The retinal vascular diameter was extracted from the color fundus image, and the relevant laboratory indexes of the patients were collected. Logistic regression analysis was used to determine the risk factors of ischemic stroke. A nomogram was constructed based on the identified risk factors, and the accuracy and clinical applicability of the model were analyzed using the receiver operating curve (ROC), Hosmer-Lemeshow test, and decision curve analysis (DCA).

RESULTS

Independent risk factors for ischemic stroke including hypertension (OR 2.17, 95% confidence interval [CI] 1.16 to 4.08; P = .016), hyperlipidemia (OR 2.21, 95% CI 1.18 to 4.14; P = .013), and resistance index of ophthalmic artery (OR 5.98, 95% CI 3.27 to 10.93; P < .001) were identified by multivariate regression analysis. The area under the ROC curve of the training cohort was 0.790 (95% CI 0.733 to 0.847) and that of the validation cohort was 0.773 (95% CI 0.679 to 0.866), revealing the consistent ability of the nomogram to predict ischemic stroke. The mean absolute error of the training and validation cohorts were 0.020 and 0.013, respectively. In addition, the DCA curve showed good clinical benefit.

CONCLUSIONS

The nomogram combining traditional factors and ophthalmic artery resistance index has a preferable predictive performance for ischemic stroke. This suggests that the model combined with ocular hemodynamics can effectively promote the early diagnosis and intervention of ischemic stroke.