Application of Artificial Neural Networks to Identify Alzheimer’s Disease Using Cerebral Perfusion SPECT Data

Abstract

The aim of this study was to demonstrate the usefulness of artificial neural networks in Alzheimer disease diagnosis (AD) using data of brain single photon emission computed tomography (SPECT). The results were compared with discriminant analysis. The study population consisted of 132 clinically diagnosed patients. There were 72 subjects with AD and 60 belonging to the normal control group. The artificial neural network used 36 numerical values being the count numbers obtained for each area of brain SPECT. These numbers determined the set of input data for the artificial neural network. The sensitivity of Alzheimer disease diagnosis detection by artificial neural network and discriminant analysis were 93.8% and 86.1%, respectively, and the corresponding specificity was 100% and 95%. We also used receiver operating characteristic curve (ROC) analysis and areas under receiver operating characteristics curves were correspondingly 0.97 (p < 0.0001) for the artificial neural networks (ANN) and 0.96 (p < 0.0001) for discriminant analysis. In conclusion, artificial neural networks and conventional statistics methods (discriminant analysis) are a useful tool in Alzheimer disease diagnosis.