0052 A Novel Biomarker of Alzheimer’s Disease Based on High-order Interactions of Low-density Electroencephalography

Abstract

Abstract Introduction Alzheimer’s disease (AD) causes disruptions in brain functions, which can be tracked by proper analysis of electroencephalography (EEG). Current efforts to understand the impact of AD on brain activity have leveraged the development of high-density EEG. Alternatively, we propose that AD can be discriminated from healthy controls (HC) by analyzing low-density EEG sleep recordings with nonlinear measures that capture functional interactions beyond pairwise such as the total correlation (TC), dual total correlation (DTC), O-information (O) and S-information (S). Methods We analyzed overnight polysomnography data with 4 EEG channels (left central, right central, left mastoid, and right mastoid) of 28 older men (14 with a self-reported history of clinical diagnosis of AD and 14 HC, matched in age and education) in the Osteoporotic Fractures in Men Study (MrOS) obtained from the National Sleep Research Resource (NSRR). Each 30-second epoch was scored as wake (W), N1, N2, N3, or rapid eye movement. Features based on low (2) and high order (3-4) interactions between electrodes (TC, DTC, O, and S) were extracted from raw and band-pass filtered data within each wake/sleep stage. Finally, a feature selection algorithm provided the optimal subset of features that maximizes the discrimination between conditions. Results AD was discriminated from HC with an average area under the receiver operating characteristic curve of 0.90±0.05 using just 3 features: O (triplet) measured within W stage in the high γ band (40-100 Hz), DTC (triplet) for raw N1 data, and TC (pairwise) in the δ band (0.5-4 Hz) within N stages. Conclusion State-of-the-art analysis of high-order interactions can maximize the discrimination of AD using a low-density sleep EEG set-up. The potential of these features to capture the physiopathology of AD and its impact on brain function should be further confirmed using larger samples. Support (if any) National Institutes of Health (RF1AG059867, RF1AG064312, R01AG057234); ANID/FONDECYT Regular (1210195 and 1210176 and 1220995); Alzheimer’s Association (SG-20-725707). MrOS Sleep Study was supported by R01 HL071194, R01HL070848, R01HL070847, R01HL070842, R01HL070841, R01HL070837, R01HL070838, and R01HL070839. The NSRR was supported by R24HL114473, 75N92019R002.