Diagnostic features of Alzheimer´s disease extracted from FDG PET images

FDG-PET images of patients suffering from Alzheimers disease (AD) were obtained from Paul Scherrer Institute, Villingen, Switzerland. The data were from a CTI/Siemens ECAT 933/04-16 scanner, comprising of 7 image slices 128 × 128 pixels. The study included 48 clinically diagnosed AD patients and 73 normal controls. Using an invariant feature extraction method features were extracted. The features are invariant to translation and rotation of object(s) within the image. The patients are separated into two groups one for training (24 AD and 37 normal controls) and one cross validation testing (24 AD and 36 normal controls). Discriminant function analysis yielded a classification accuracy of 88% sensitivity and 86% specificity, when these features were used.