FDG and PIB biomarker PET analysis for the Alzheimer's disease detection using Association Rules

This work shows an Association Rule (AR)-based approach in order to design a computer aided diagnosis (CAD) system for the Alzheimer's disease (AD) detection with a 18F-FDG and Pittsburg Compound B (PiB) PET (Positron Emission Tomography) biomarker analysis. The AD Neuroimaging Initiative (ADNI) 1 dataset is used for testing and a comparison is conducted in two different sets: controls versus AD or controls versus Mild Cognitive Impairment (MCI) subjects. 3D Regions of Interest (ROls) are obtained with an activation estimation method for both biomarkers considered independently or in a combined way without repetition. These ROls are used as input for an Apriori algorithm in order to obtain ARs from controls. Classification is performed taking into account the percentage of verified rules by each subject. Although PiB is distinguished from FDG in providing differential information, the combination of biomarkers shows positive results for both sets: while in the first, FDG results get better reaching an accuracy of 94.74%, in the second set both FDG and PIB rates are improved reaching a maximum accuracy of 92.86%; outperforming the results of other recently reported methods.