제목: Multi-stage classification of the clinical stage in Alzheimer;s Disease using EEG during Resting states and Memory Encoding states
발표자: 김슬기 (고려대 뇌공학과 박사과정 수료 및 한국과학기술연구원(KIST) 바이오닉스연구센터 연구원)
일시: 2023년 4월 27일 목요일 오후 4시
Background and objective: The number of patients with Alzheimer’s disease (AD) is increasing worldwide. AD is a progressive and life-threatening disease that is difficult to treat. Mild cognitive impairment (MCI) is considered a transitional phase between normal aging and AD, and research suggests that, each year, approximately 8 – 15% of patients diagnosed with MCI progress to AD. Therefore, early detection of MCI is essential for controlling disease progression and delaying the onset of intellectual decline. Many studies have been conducted to diagnose this condition in advance; however, most of them measured EEG signals only in the resting state. In our study, we identified whether obtaining electroencephalography(EEG) signals at rest and during cognitive functioning had a positive effect on the early diagnosis of Alzheimer's.
Methods: The resting and memory-encoding states of 58 patients (20 with SCD, 10 with non-amnestic MCI (naMCI), 18 with aMCI, and 10 with AD) were measured and classified into four groups. We extracted features that could reflect the phase, spectral, and temporal characteristics of the resting and memory-encoding states. For the classification, we compared nine machine learning models.
Results: In the nine models used, the memory-encoding states realized a higher classification performance than rest states. The cKNN model had a high reliability score of 0.77 and grouping accuracy of 91.38% when applied to EEG data obtained during memory encoding. The F1-scores of the groups were 89.47%, 90.00%, 91.89%, and 95.24%. In addition, we confirmed that each encoding session contributed differently to the classification of the stage of AD.
Conclusions: The clinical stages of AD were classified using EEG characteristics obtained when participants performed a memory-encoding task that required less time than the rest period. We found that EEG characteristics obtained during cognitive function is more helpful in diagnosing and predicting AD. In particular, the classification includes SCD and naMCI, which are stages between normal and mild MCI and non-memory and MCI.
Keywords: Alzheimer’s disease, mild cognitive impairment, memory-encoding states, EEG