Premium
Classification of Alzheimer’s disease dementia (ADD) by using features derived from resting‐state electroencephalography (rsEEG)
Author(s) -
Lizio Roberta
Publication year - 2020
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.045816
Subject(s) - dementia , receiver operating characteristic , medicine , electroencephalography , disease , dementia with lewy bodies , audiology , neuroscience , physical medicine and rehabilitation , psychology , psychiatry
Background Alzheimer’s disease (AD) is the most prevalent progressive neurodegenerative disease of the brain affecting the aged people, and the most common cause of dementia (ADD). The actual diagnostic biomarkers of AD are invasive and expensive (e.g., lumbar puncture for CSF sampling; the injection of radioactive tracers in PET procedures). Resting‐state electroencephalography (rsEEG) provide topographic markers useful to assess the neurophysiological changes of the brain that correlate with the cognitive decline and dementia in AD patients. Method In this retrospective study, we tested whether the cortical sources of rsEEG rhythms could classify with good performance ADD patients from healthy elderly (Nold) individuals and patients with other diseases. Clinical and rsEEG data of ADD, Parkinson’s disease with dementia (PDD), dementia with Lewy body (DLB), and Nold subjects were available in an international archive. eLORETA estimated the rsEEG cortical sources. Delta, theta, low alpha, high alpha, low beta, high beta, and gamma were the frequency bands of interest. The correct blind classifications (ADD vs Nold, ADD vs PDD, ADD vs DLB individuals) of these rsEEG source activities were performed by GraphPad‐Prism software to produce the receiver operating characteristic (ROC) curves. The area under the ROC curve (AUC) provided a measure of how well the rsEEG source activities distinguished the groups (AUC>0.7 as a threshold for a moderate classification rate). Results The posterior delta and alpha sources allowed good classification accuracy (AUC: 0.75‐0.90) in Nold (N=40) vs ADD (N=42), ADD (N=42) vs PDD (N=42), and ADD (N=42) vs DLB (N=38) individuals (Figure.1). We also obtained a good classification accuracy (AUC > 0.80) between the Nold (N = 100) and ADD (N = 100) individuals using bipolar parieto‐occipital delta/alpha and theta/alpha rsEEG features. Conclusion These results suggest that cortical sources of posterior rsEEG rhythms at different frequency bands and frequency band ratios can be used to discriminate ADD individuals with good performance (AUC>0.8). A single rsEEG marker from a point of care device would provide a cost‐effective, non‐invasive, and repeatable over time biomarker for the assessment of the AD status and progression directly at home.