Related Experiment Videos
Classification of quantitative EEG data by an artificial neural network: a preliminary study
L A Riquelme1, B S Zanuto, M G Murer
1Departamento de Fisiología y Biofísica, Facultad de Medicina, Universidad de Buenos Aires, Argentina.
Neuropsychobiology
|January 1, 1996
Summary
Quantitative electroencephalography (qEEG) effectively distinguishes dementia from other conditions. Both statistical analysis and neural networks show high accuracy in classifying qEEG data for dementia diagnosis.
Area of Science:
- Neuroscience
- Medical Informatics
Background:
- Quantitative electroencephalography (qEEG) shows potential for differentiating dementia from normal cognition.
- Accurate classification of dementia subtypes in ambulatory settings remains a challenge.
Purpose of the Study:
- To evaluate qEEG's ability to distinguish dementia across various pathological conditions in ambulatory settings.
- To compare the efficacy of classical statistical analysis versus neural networks for qEEG data classification.
Main Methods:
- Development of a multiple discriminant function using a patient training set.
- Application of Kohonen's unsupervised learning artificial neural network for classification.
- Validation using an independent patient group.
Main Results:
- The discriminant function achieved over 91% correct classification of independent qEEG samples with <5% false positives.
- The unsupervised learning neural network demonstrated comparable classification accuracy to the discriminant function.
- Both methods proved effective in classifying qEEG data.
Conclusions:
- qEEG, analyzed with both statistical methods and neural networks, shows high diagnostic accuracy for dementia.
- Unsupervised learning algorithms offer a promising alternative for classifying psychiatric patient data, especially when clinical profiles are complex.