Detection of subclinical brain electrical activity changes in Huntingtonʹs disease using artificial neural networks

Objective: The aim of this study was to analyze EEG background activity in Huntingtonʹs disease (HD) patients and relatives at risk, in relation to CAG repeat size and clinical state, in order to detect an electrophysiological marker of early disease. Methods: We selected 13 patients and 7 subjects at risk. Thirteen normal subjects, sex- and age-matched, were also evaluated. Artifact-free epochs were selected and analyzed through Fast-Fourier Transform. EEG background activity was tested using both linear analysis and artificial neural network (ANN) classifier in order to evaluate whether EEG abnormalities were linked to functional changes preceding the onset of the disease. Results: The most important EEG classification pattern was the absolute α power not correlated with cognitive decline. The ANN correctly classified 11/13 patients and 12/13 normals. Moreover, the neural scores for subjects at risk seemed to be correlated to the expected time before the onset of the disease. Conclusions: ANN is a very powerful method to discriminate between normals and patients. It could be used as an automatic diagnostic tool. EEG changes in positive gene-carriers for HD confirm an early functional impairment which should be taken into account in the genetic counseling and in the management of the early stages of the disease.