Improved life expectancy has led to a significant increase in the number of people in the high-risk age groups that will develop Alzheimer's disease and other dementia. Efforts are being made to develop treatments that slow the progress of these diseases. However, unless a sufferer is diagnosed in the early stages the treatments cannot give the maximum benefit. Therefore, there is an urgent need for a practical, decision support tool that will enable the earliest possible detection of dementia within the large at-risk population. Current techniques such as Magnetic Resonance Imaging (MRI) that are used to diagnose and assess neurological disorders require specialist equipment and expert clinicians to interpret results. Such techniques are inappropriate as a method of detecting individual subjects with early dementia within the large at-risk population, because everyone within the at-risk group would need to be tested regularly and this would carry a very high cost. Therefore, it is desirable to develop a low cost method of assessment. This thesis describes research into the use of automated EEG analysis to provide the required testing for dementia. The research begins with a review of previous automated EEG analysis, particularly fractal dimension measures. Initial investigation into the nature of the fractal dimension of the EEG are conducted, including problems encountered when applying fractal measures in affine space. More appropriate fractal methods were evaluated and the most promising of these methods was blind tested using an independent clinical data set. This method was estimated to achieve 67% sensitivity to probable early Alzheimer's disease and 17% sensitivity to vascular dementia (as confirmed by a clinical neurophysiologist from the EEG) with a specificity of 99.9%.

Document Type


Publication Date