An SFFS technique for EEG feature classification to identify sub-groups

Pattern recognition techniques can be applied to problems in medicine to aid diagnostic accuracy and uncover patterns associated with disease states that are not always obvious to the clinician. In this work, a sequential forward floating search technique (SFFS) was applied to the problem of classification of patients with Alzheimer´s disease (AD), mild cognitive impairment (MCI) and normal controls. The technique resulted in superior classification rates over statistical methods, as described in the paper. The advantage of SFFS may lie in the technique´s ability to identify subgroups within diagnostic categories, and to correctly select features that identify those sub-groups.