Glenohumeral subluxation (GHS) is a common complication in stroke patients with hemiplegia, occurring in approximately 17-81% of cases. This study aims to evaluate the relationship between shoulder muscle thickness and the degree of subluxation using ultrasound imaging. A cross-sectional study of 61 stroke patients with hemiplegia was conducted, measuring supraspinatus muscle thickness, deltoid muscle thickness, and acromion-greater tuberosity (AGT). Logistic regression and ROC analyses were used. ROC curves, calibration plots, and decision curves were drawn on the training and validation sets. According to logistic regression analysis, the ratio of supraspinatus muscle thickness was statistically significant (OR: 0.80; 95% CI: 0.70-0.92; p < 0.01), and it was an independent factor for evaluating the presence or absence of GHS. An AUC of 0.906 (95% CI, 0.802-1.000) was found in the training set; meanwhile, the AUC in the validation set was 0.857 (95% CI, 0.669-1.000), indicating good performance. According to the training set ROC curve, the most effective statistical threshold was 93%, with a sensitivity of 84% and a specificity of 96%. The ratio of supraspinatus muscle thickness is a valuable criterion for evaluating GHS risk, supporting targeted rehabilitation interventions.