Abstract
Alzheimer’s disease (AD) is the most prevalent form of dementia in the world. Symptoms include progressive memory, cognitive and behavioural changes before death, caused by amyloid plaques and hyperphosphorated tau in the brain. The cause of AD is currently unknown and current interventions only slow the decline. Diagnosis is based on patient and familial history, interviews with close family and friends, cognitive, mental and physical tests. The electroencephalogram (EEG) records the electrical signals of the brain, which AD, as a cortical dementia, is known to directly affect. Non-linear signal processing has shown that these changes in the EEG can be identified with complementary findings to linear methods. This thesis aimed to explore these changes with novel univariate and bivariate methods using both synthetic signals and resting, eyes-closed EEGs recorded from 22 subjects, 11 AD patients (MMSE=13.1±5.9 (mean SD)) and 11 age-matched controls (30±0). Permutation entropy showed statistically significant increased complexity in control EEGs for electrodes at the front of the head. Bivariate analysis was novel for this EEG database so coherence was used to create a comparison results set. With the success of Lempel-Ziv Complexity (LZC), distance based bivariate forms were applied (dLZC03). Novel normalisation methods based on that for univariate LZC showed a greater representation of the signal patterns in the results. Volume conduction was shown to significantly impact the results, both of coherence and dLZC03, though this was greater with coherence. Lastly, volume conduction mitigated, bandwidth based pre filtering with dLZC03 was calculated, producing the most significant p values ever recorded with this EEG database. Thus this PhD shows increased distinction between the two subject groups with dLZC03 over LZC and increased distinction with limited but targeted bandwidths from those subject signals.