Loading...

Alzheimer's Disease Diagnosis Using Brain Source Localiztion, Based on Realistic Head Model

Aghajani, Haleh | 2012

699 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 42870 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Vosoughi Vahdat, Bijan; Jalili, Mahdi
  7. Abstract:
  8. Dementia is one of the most common disorders among the elderly population. Statistical analyses show that among several subtypes of dementia, Alzheimer’s disease (AD) is the most frequent cause of dementia and this number is projected to increase. AD not only results in impairment of learning and memory but also in the moderate stages of illness, motor functions are profoundly disturbed, and ultimately will affect the patient's lifetime. The proposed drug treatments for this disease only reduce its progress probability. Therefore, early diagnosis for effective treatment of Alzheimer's disease is one of the critical issues in the field of dementia. This project is an effort to extract discriminant features from sources of localization based on realistic head model to classify AD and healthy subjects. For brain source localization, sLORETA due to linearity, statistical approach, and being unbiased in no noise and high SNR conditions has been considered in many studies and applications other than AD. Therefore, in this project access to the location of brain sources is based on sLORETA. Because of its simplicity and lack of structural fitness of spherical models with head, in this project a realistic head model based on MNI152 atlas is entered to problem solvation prosedure by use of standardized BEM and by this means, we were relieved the problem of unavailable anatomical images for each subject and the high computational volume. An important step in obtaining reliable results in source localization has taken place in this project, by using the high spatial resolution EEG data. Statistical analysis results on the logarithmic transformed and normalized power spectrum has shown acceptable differences (P <0.05) between median values in different subbands and ROIs between two groups of healthy subjects and patients. Therefore, it can be inferred that Alzheimer's disease leads to power decline in high frequencies and power increase in low frequencies or in other word slowing EEG signals. By using optimized features, classification is done using various supervised methods. The best result belongs to SVM with accuracy of 0.912, sensitivity of 0.941, and specificity of 0.882. These results in comparison with the best results obtained from previous studies with Alzheimer's disease classification approach based on brain source localization, with an accuracy of 0.840, and sensitivity of 0.900 have increased. Based on the results, stated processing framework has a significant improvement in diagnosis of AD patients with source localization approach.
  9. Keywords:
  10. Electroencephalography ; Alzheimer ; Disease Diagnosis ; Head Model ; Brain Source Localization ; Standardized Low Resolution Brain Electromagnetic Tomography (SLORETA)

 Digital Object List

  • محتواي پايان نامه
  •   view

 Bookmark

No TOC