Identification of mild cognitive impairment using extreme learning machines model

Abstract

Alzheimer’s disease (AD) is very common disorder among the older people. Predicting mild cognitive impairment (MCI), an intermediate stage between normal cognition and dementia, in individuals with some symptoms of cognitive decline may have great influence on treatment choice and disease progression. In this work, a novel prediction model based on extreme learning machines algorithm was proposed to identify mild cognitive impairment using the information of some biomarkers, i.e., Position Emission Tomography (PET), Magnetic Resonance Imaging (MRI), and the cerebrospinal fluid (CSF). Firstly, each subject will be represented by three biomarkers using a vector with 201 dimensions after some essential pre-processing. Then, a prediction model based on extreme learning machine algorithm was trained to catch the difference between the patients and healthy persons. Finally, the model had been employed to identify identifying MCI participants from the normal people. Based on a dataset, including 99 MCI patients and 52 health controls, our proposed method achieved 70.86 % prediction accuracy with 67.68 % sensitivity at the precision of 76.92 %. Extensive experiments are performed to compare our method with state-of-the-art techniques, i.e., support vector machine (SVM) and SVM with fusion kernels. Experimental results demonstrate that proposed extreme learning machine is a powerful tool for predicting MCI with excellent performance and less time.