Accuracy evaluation of using Graph Theory metrics for functional connectivity analysis MCI classification

Detta är en Kandidat-uppsats från KTH/Datavetenskap

Författare: Pablo Urquijo Martínez; [2022]

Nyckelord: ;

Sammanfattning: Mild Cognitive Impairment (MCI) is an intermediate stage between the physiological cognitive deterioration due to ageing and a nonphysiological stage of dementia. MCI has morbidity between 12% and 18% of the population aged over 65 [1]. Even though MCI is not considered dementia and the symptomology of the patients is not severe, the patients with this disease are at higher risk of subsequently developing dementia and other neurodegenerative illnesses like Alzheimer’s (AD). Functional connectivity analysis has proven helpful in enhancing our knowledge of cognitive disorders such as MCI. The functional connectivity analysis is often combined with Graph Theory to extract metrics about the brain state and improve the classification of MCI patients using Machine Learning. In this report, I studied the classification accuracy improvement of using the Graph Theory metrics alongside the single functional connections for a specific dataset. Additionally, I explored the features that reported higher differences between MCI and healthy patients. I found that using only Graph Theory metrics for this dataset does not provide enough accuracy but that using them alongside the individual functional connections enhances the accuracy value. In addition, regarding the Graph Theory metrics, the node features proved to be more relevant for the classification than the whole matrix features. Additionally, I observed a significantly altered behaviour in some brain regions.

  HÄR KAN DU HÄMTA UPPSATSEN I FULLTEXT. (följ länken till nästa sida)