Prospects for routine detection of dementia using the fractal dimension ofthe human electroencephalogram

The paper details research which aims to improve the contribution made by e lectroencephalogram (EEG) analysis to the diagnosis and care of patients wi th brain disease; dementia in particular. Previous attempts to automate EEG analysis have concentrated on separating patient groups from control group s, often on the basis of a single neurophysiological index derived from a s hort, isolated segment of EEG. The authors seek to develop, and test, a nov el technique for the analysis of changes in serial EEG recordings on indivi duals (subject-specific analysis) which may serve as a basis fur routine ea rly detection of dementia. The objectives of the reported study were to exa mine the feasibility of applying appropriate fractal dimension (FD) (comple xity) measures to the human EEG. and to examine whether methods using the s ubject specific variability of these measures are likely to be useful for d etecting patients who develop dementia. The reason for undertaking the stud y was to establish a 'proof of concept' and determine whether research shou ld concentrate in this area. Existing EEG analysis methods were reviewed an d four FD measures suitable for EEG analysis were developed. These four mea sures were applied to a total of 21 EEG recordings (from seven subjects wit h various dementias, eight age matched controls and two young subjects who gave three recordings each). The results were analysed and the following co nclusions were drawn: it is possible to measure the complexity of the human EEC using the FD, and the subject specific variability of the FD is an imp ortant candidate method for identifying patients with dementia. Therefore, further work in this area is justified.