Objective      To establish a diagnostic model of Alzheimer’s disease and mild cognitive impairment.       Methods      Receiver operator curve (ROC) curve was used to screen the indices

discriminant analysis was carried out for diagnosis simulation

and the discriminant function with the highest accuracy was selected.       Results      Four indices were selected by ROC curve. After 100 times of random sampling for diagnosis simulation in the objects including normal people

patients with Alzheimer’s disease and patients with mild cognitive impairment

the average diagnosis accuracy of linear discriminant analysis was 71.42% and that of nonlinear discriminant analysis was 74.75%. The highest accuracy was 82.05%

and the corresponding discriminant function was obtained.       Conclusion      Nonlinear discriminant analysis can be used to diagnose Alzheimer’s disease and mild cognitive impairment.