Parallel principal component neural networks for classification of event-related potential waveforms

J R Sveinsson; J A Benediktsson; S B Stefansson; K Davidsson

doi:10.1016/s1350-4533(96)00035-5

Parallel principal component neural networks for classification of event-related potential waveforms

Med Eng Phys. 1997 Jan;19(1):15-20. doi: 10.1016/s1350-4533(96)00035-5.

Authors

J R Sveinsson¹, J A Benediktsson, S B Stefansson, K Davidsson

Affiliation

¹ Engineering Research Institute, University of Iceland, Reykjavik, Iceland.

PMID: 9140869
DOI: 10.1016/s1350-4533(96)00035-5

Abstract

Artificial neural networks (ANNs) are discussed in terms of classification of brain auditory event-related potentials (ERPs). A new ANN architecture for the classification of ERPs is proposed. The new architecture is called the parallel principal component neural network (PPCNN). The use of the PPCNN for classification of ERP data obtained from both normal control subjects and chronic schizophrenic patients is discussed. Experimental results are given.

Publication types

Research Support, Non-U.S. Gov't

MeSH terms

Adult
Biomedical Engineering
Evaluation Studies as Topic
Evoked Potentials, Auditory, Brain Stem / physiology*
Humans
Male
Middle Aged
Neural Networks, Computer*
Schizophrenia / physiopathology