Image Retrieval for Alzheimer's Disease Detection

A project is described with the aim to develop a Computer-Aided Retrieval and Diagnosis of Alzheimer's disease. The domain of focus is Alzheimer's disease A manually curated MRI data set from the Alzheimer's Disease Neuroimaging initiative (ADNI) project (http://www.loni.ucla.edu/ADNI/) was used for training and validation. The system's main function is to generate accurate matches for any given visual or textual query. The system gives an option to perform the matching based on a variety of feature-sets, extracted using an adaptation of a discrete cosine transform algorithm. Classification is conducted using Support Vector Machines. Finally, ranking of most accurate matches are generated by applying an Euclidean distance score. The overall system architecture follows a multi-level model, permitting performance analysis of components independently. Experimental results demonstrate that the system can produce effective results.