Early Detection of Alzheimer's disease using EEG based on Time-Frequency Analysis and Artificial Intelligence Techniques

Alzheimer's disease (AD) is a chronic neurodegenerative disease that causes problems with memory, thinking, language and even self-care in the advanced stages. It is the most common form of dementia, constituting about 50% to 70% of all cases.

AD has irreversible and progressive process. Available treatments have been shown relatively small benefits in delaying or stopping its process. Studies have shown that early diagnosis of AD can be beneficial because (i) the patient and his/her family have enough time to think and decide for life and financial matters caused by the disease and the care needed in the advanced stages, (ii) early diagnosis of AD gives the patient the opportunity of benefiting from symptom delaying medications which are most useful in the early stages.

In recent years, electroencephalography (EEG) as a cheap, potentially mobile and easy to record measure of brain activity has gained increasing attention of researchers for early diagnosis of AD. Although the progress in exploring the capability of EEG for early diagnosis of AD is considerable, a lot of problems still remain unsolved in this area. According to literature from signal processing point of view the reported effects of AD on EEG are promising but not consistent and practical.

Moreover it seems there exists a lack of discriminative feature selection and efficient classification methods in order to report reliable classification error rate. The aim of this project is to fill this gap using signal processing methods in particular time-frequency analysis and artificial intelligence techniques.

The target group of the project, involves groups of patients in early stages of the disease. i.e., mild cognitive impairment (MCI) or mild AD.

The project will contribute to the investigating the potential of EEG for early diagnosis or even prognosis of AD. If the capability of EEG is confirmed in this respect, it can be included as a screening tool for AD in yearly check-ups of people after the age of 50.
The specific objectives of the project include:

• Analysis the capability of proposed features so far from the consistency and effectiveness point of view.
  
  
• Explore the potential of time-frequency methods in describing the effect of AD on EEG more in depth than existing literature.
  
  
• Find a proper feature selection method which can select a subset of features that describe the signal efficiently and robust to inter-patient variations.
  
  
• Compare the capability of different artificial intelligence based methods for classifying the extracted features in order to discriminate between AD patients', MCI patients' and healthy group's EEG.
  The methodology for feature extraction of EEG, used in the project, is not limited to time-frequency methods and existing literature. The main reason of selecting time-frequency analysis to explore the EEG is the ability of these techniques in describing and characterizing the signals with non-stationary nature. Other signal processing methods will be used as well if necessary.
  

• Data mining
• Time-frequncy analysis
• Artificial intelligence