Objective  To establish a differential diagnosis model between moyamoya disease (MMD) and non-MMD ischemic stroke by screening clinically accessible variables.

Methods  A total of 150 patients diagnosed with MMD ischemic stroke and 150 patients diagnosed with non-MMD ischemic stroke were included and divided into training group (210 cases) and validation group (90 cases) according to a ratio of 7 to 3. Logistic regression analysis, Lasso regression and support vector machine (SVM) were used to construct the diagnosis model; the optimal model was visualized with nomogram, and discriminant ability of the nomogram was tested in the training group and the validation group respectively.

Results  Binary Logistic regression showed the largest C statistics in the training group and the verification group (0.87 and 0.88). Multivariate logistic regression analysis showed statistically significant differences (P < 0.05) in systolic blood pressure, total cholesterol (TC), albumin (ALB), free triiodothyranine (FT₃), homocysteine (HCY) and age, and these six variables were included in the nomogram. Hosmer-Lemeshow test P values for the training group and the verification group were 0.28 and 0.19, respectively, and the calibration curve was well calibrated.The nomogram score < 168 was classified as a low risk of ischemic stroke in MMD, and the nomogram score ≥168 was classified as a high risk of ischemic stroke in MMD.

Conclusion The nomograms established in the study can be used to distinguish MMD from non-MMD ischemic stroke, and it has been verified that the model has good discriminative ability in both training group and verification group.