Separation of Healthy Individuals and Patients with Alzheimer's Disease Using the Effective Communication of Brain Signals

Alzheimer's disease (AD) is a degenerative and progressive disease of the brain that causes the deterioration of intellectual abilities. Approximately 5% of people over 70 years old and 20% of people over 80 years old suffer from this disease. So far, many tools and methods have been provided to diagnose AD. However, in most of these methods, the interactions and connections of different parts of the brain are not considered. Since AD can affect different structures of the brain, damage to any part of the brain disrupts its interaction with other regions.

The indexes of effective communication between different parts of the brain in two groups of healthy people and subjects with AD were extracted using Granger causality analysis. Following statistical comparisons between the quantitative values of the indexes in different EEG channels, we examined effective communication. Then, we used linear differential analysis to separate the two groups of participants. The data used in the research include EEG signals from 10 healthy subjects and 8 patients with AD (mild and severe).

With the correct diagnosis of all patients and only one wrong diagnosis of a healthy subject, an accuracy of 83.33%, an accuracy of 90%, a sensitivity of 100%, and a diagnosis of 80% were obtained for the test data. The effective communication rate of Fz and Cz channels for healthy people is higher than the effective communication of the Pz channel, while for patients with AD, the lowest effective brain communication was observed in the Fz channel, and the highest communication was observed in the Pz channel and sometimes in the Cz channel.

The results of Granger features are far better than the results of linear features, despite the fact that the number of extracted linear features was more than Granger features. Therefore, the effectiveness of Granger causality has been proven once again, and it can be said that the interaction indices between EEG channels provided valuable information for classification and led to better identification of patients with AD.