An optimal framework for alzheimer’s disease diagnosis

Abstract

Alzheimer’s disease (AD) is a kind of progressive dementia that is frequently accompanied by brain shrinkage. With the use of the morphological characteristics of MRI brain scans, this paper proposed a method for diagnosing moderate cognitive impairment (MCI) and AD. The anatomical features of 818 subjects were calculated using the FreeSurfer software, and the data were taken from the ADNI dataset. These features were first removed from the dataset after being preprocessed with an age correction algorithm using linear regression to estimate the effects of normal aging. With these preprocessed characteristics, the extreme learning machine served as a classifier for the diagnosis of AD and MCI. For determining accuracy, sensitivity, specificity, and area under the curve, ten-fold cross validation was used. The accuracy of AD diagnosis was 87.62 percent on average after 100 runs, while the area under curve was 94.25 percent. The sensitivity of theMCI diagnosis was 83.88 percent, while the accuracy was 73.38 percent. The age correction can help diagnose MCI more accurately. The outcomes showed that the proposed strategy for diagnosing AD and MCI was more effective than existing methods.