Development of an EEG Diagnostic for Alzheimer's Disease

Patients with diagnoses of sporadic and late onset Alzheimer's disease dementia, MCI, and DLB evaluated by a dementia subspecialist will meet published diagnostic criteria, and EEGs will be obtained through written informed consent. Regarding presymptomatic patients, we have previously shown that the mean age of MCI diagnosis in our cohort is 73 years and that preclinical cognitive decline begins as much as 20 years before clinical diagnosis but is also affected by APOE genotype. Eligibility therefore will include unimpaired APOE e4/4 homozygotes age 65-75 and APOE e3/4 heterozygotes age 75-85 for the preclinical AD subset and age, sex, and education matched APOE e4 noncarriers for the unaffected controls. Biomarker confirmation for preclinical diagnosis will be utilized to the extent possible (a subset of 130 members of our cohort have undergone amyloid-PET resulting in approximately 45 who are amyloid positive). EEG data will be performed during wakefulness with 15 minutes of eyes open and 15 minutes of eyes closed. A 32 electrode cap will be applied following the 10-20 anatomical system by certified EEG technologists. Data will be recorded using a research-grade EEG system with FDA 510(k) clearance for use in medical contexts. Subjects will be seated in a testing room with minimal distractions. An EEG tech will fit the subject with a cap containing NN Ag/AgCl electrodes placed according to the international 10-10 system and ensure electrode impedances stay below 10 kΩ. Subjects will be instructed to minimize movements and remain in a relaxed but wakeful state. We will record fifteen minutes of eyes-open resting state EEG. Afterwards, the subjects will be instructed to close their eyes and reminded to stay relaxed but awake, and we will record fifteen minutes of eyes-closed resting state EEG. During the recording session, a researcher will monitor the subject's behavior and the EEG signal. The researcher will briefly prompt the subject to remain awake if the subject's behavior or EEG traces show signs of drowsiness. All data will be de-identified then transferred to SPARK Neuro's research and development team for analysis via secure encrypted methods. The data will be stored on password-protected computer systems, and only the necessary research and research-support staff at SPARK Neuro will have access to the data. The SPARK Neuro research and development team will analyze de-identified patient data to address the aims of this proposal. SPARK Neuro will use various techniques including those standard in EEG analysis (e.g. filtering, scaling, independent components analysis), particular approaches shown to be successful in the Alzheimer's disease EEG classification literature (e.g. coherence, relative power in standard frequency bands, functional connectivity, spectral entropy), as well as approaches from the broader machine-learning and EEG literature (e.g. spectral clustering, convolutional neural networks, cross-frequency coupling, non-linear kernels, Katz fractal dimension, beamforming).