The neuronal activity in the brain of demented subject becomes non-uniform over the
cortical surface and this non-uniformity is sensitively detected by properly analyzing observed
scalp potentials. The present invention is related to the
computer algorithm and apparatus for numerically detecting the degree of such
neuronal degeneration. A proper frequency component of EEG is selected so that its
scalp potential becomes smooth in a normal subject. When the
brain function is impaired, non-uniformity of neuronal activity is detected in spatial as well as temporal fluctuations in the observed
scalp potential of properly filtered EEG (usually the alpha component). The statistical parameters derived from dipolarity values of such a properly selected EEG component determine the state of the brain, in which whether the subject is normal or demented is Judged in the early stage of
dementia by comparing the obtained parameters with their threshold values. This invention presents a reliable, inexpensive, easy-to-
handle and non-invasive method for sensitive screening of
dementia and monitoring progress of
dementia which also allows optimizing medication and treatment for dementia.