Previous work has suggested that evoked potential analysis might allow the detection of subjects with new onset Alzheimer's Disease, which would be useful clinically and personally. Here, it is described how new-onset Alzheimer's disease subjects have been differentiated from healthy, normal subjects to 100% accuracy, based on the back-projected independent components (BICs) of the P300 peak at the EEG electrodes in the response to an oddball, auditory evoked potential paradigm. After artefact removal, clustering, selection, and normalisation processes the BICs were classified using a neural network, a Bayes classifier, and a voting strategy. The technique is general and might be applied for pre-symptomatic detection, and to other conditions and evoked potentials, although further validation with more subjects, preferably in multi-centre studies is recommended.