Fig. 4

Receiver operating characteristic (ROC) curves classifying diagnostic groups in Alzheimer’s disease using gait parameters. Binary logistic regression models were constructed for each classification. Statistical significance was assessed using bias-corrected confidence intervals of the area under the ROC curve (AUROC), based on 10 000 bootstrap samples. Model A incorporated four gait parameters in the path analysis, with the path models involving these parameters demonstrating satisfactory or good fit. Models B, C, and D additionally employed base width, gait variability (step length and double support time standard deviations), and both parameters, respectively, all demonstrating significant mean differences. Red lines in legends denote significant differences in AUROC values between the models. CU, cognitively unimpaired; AD, Alzheimer’s disease; AUROC, area under receiver operating characteristic curve; CI, confidence interval; SD, standard deviation.