Non-linear analysis of the Human Electroencephalogram for the detection of Alzheimer's disease using Mutual Information

Alzheimer disease (AD) is the most prevalent neurodegenerative disease in the world. Its impact on patients, as well as its wide or its cost makes it one of the major challenges that the research has to take up. Indeed, it involves million of people and billion of dollars each year in care. Its wide is expected to be multiplied by 4 in the next 40 years. Nowadays, no medication can cure the disease. However, a good medication can considerably slow the progression of the disease, especially i f it is detected early. Today's detection consists on psychological tests and interviews. These kinds of detection cannot detect the disease in its earlier stage, where no memory deterioration has appeared. The electroencephalogram (EEG) is a measurement, which can be used to detect AD. Different methods are studies for the interpretation of the EEG. This study looks at the mutual information (MI), a statistical method that looks at linear as well as non-linear interactions between areas of the brain. As A D consists on destruction of neurons, we can expect this link to be weaker for patients with A D than for healthy patients. We have used a particular method to calculate the probability density function used in the calculation of the MI: the Parzen estimator. We have applied our method on eleven E EG recordings measured from patients with AD. We have compared them with eight E EG recordings from normal controls. We have first determined the importance of different parameters for the computation of the MI. We have shown a link between the MI and the power of the EEG. Once the power of the EEG have been normalised, the MI of E EG measured on ill patients is lower than the EEG measured on normal controls. This difference is more important when we look at distant electrodes than when we look at close electrodes. This may confirm the assumption that A D involves some troubles in the transmission of information through the brain.School of Computing, Communications and Electronic