Inference of structural alterations in AD from resting-state EEG and whole-brain network model
Poster No:
1417
Submission Type:
Abstract Submission
Authors:
Yunier Prieur-Coloma1, Felipe Torres1, Mónica Otero Ferreiro2,1, Alejandro Weinstein1, WAEL EL-DEREDY1,3,4
Institutions:
1Brain Dynamics Laboratory, Universidad de Valparaíso, Valparaíso, Chile, 2Universidad San Sebastián, Santiago de Chile, Chile, 3ValgrAI: Valencian Graduate School and Research Network of Artificial Intelligence, Valencia, Spain, 4Department of Electronic Engineering, School of Engineering, Universitat de València, Valencia, Spain
First Author:
Co-Author(s):
Mónica Otero Ferreiro
Universidad San Sebastián|Brain Dynamics Laboratory, Universidad de Valparaíso
Santiago de Chile, Chile|Valparaíso, Chile
Universidad San Sebastián|Brain Dynamics Laboratory, Universidad de Valparaíso
Santiago de Chile, Chile|Valparaíso, Chile
WAEL EL-DEREDY
Brain Dynamics Laboratory, Universidad de Valparaíso|ValgrAI: Valencian Graduate School and Research Network of Artificial Intelligence|Department of Electronic Engineering, School of Engineering, Universitat de València
Valparaíso, Chile|Valencia, Spain|Valencia, Spain
Brain Dynamics Laboratory, Universidad de Valparaíso|ValgrAI: Valencian Graduate School and Research Network of Artificial Intelligence|Department of Electronic Engineering, School of Engineering, Universitat de València
Valparaíso, Chile|Valencia, Spain|Valencia, Spain
Introduction:
Alzheimer's disease (AD) causes 60-80% of dementia cases and is the most common neurodegenerative disorder (Zheng et al., 2023; Miltiadous et al., 2023). Identifying AD using non-invasive neuroimaging methods like electroencephalogram (EEG) requires selecting distinctive features. Zheng et al., 2023, found differences in spectrum features, complexity, and synchronization in resting-state EEG (rsEEG) of AD in contrast to Control subjects. Here, we relate these dynamic features to biophysical parameters in a whole-brain Kuramoto model.
Methods:
We fit each subject EEG spectrum to a point in the model's parameter space by correlating the EEG spectrum with model spectra. We calculate spectra using Welch's method with a Hann window of 5 seconds, achieving a frequency resolution of 0.2 Hz. We correlate the spatial average spectrum in the frequency range from 0.5 Hz to 44 Hz.
EEG recordings: We use recordings from AD and control subjects provided by Miltiadous et al. 2023. The EEG recordings were collected using 19 electrodes while the subjects had their eyes closed. Each record lasted around 13.5 minutes at a sampling rate of 500 Hz. They removed artifacts and filtered the signal between 0.5 Hz and 45 Hz. We removed the EEG's aperiodic component by multiplying by f^∝, where∝ is the logarithmic slope of the average spectrum between 1 Hz and 35 Hz.
Model: We use a whole-brain model that consists of a network of delayed Kuramoto oscillators with coupling corresponding to an anatomical brain atlas. The used atlas is the Automated Anatomical Atlas with 90 regions, and we employed 40 Hz as the intrinsic frequency for all the oscillators (Cabral et al., 2014). Our simulations last 200 seconds for each set of parameters with a time step of 1 × 10^-3 s.
In addition, Miltiadous et al. 2023 also performed a cognitive and neuropsychological assessment. They employed the Mini-Mental State Examination (MMSE) (Creavin et al., 2016). Starting at 30 points, as lower the MMSE score, the more severe the cognitive decline. The MMSE score for the AD group was 17.75±4.5 and for the control group was 30.
The point of best correlation between rsEEG and model spectra corresponds to a value of global coupling K and mean delay MD in the parameters space. For each pair of {K, MD} also corresponds a value of spectral entropy (SE) and Kuramoto Order Parameter (KOP). SE is a metric of metastability of the model signals dependent on the spectrum, while KOP is a metric of global synchrony.
EEG recordings: We use recordings from AD and control subjects provided by Miltiadous et al. 2023. The EEG recordings were collected using 19 electrodes while the subjects had their eyes closed. Each record lasted around 13.5 minutes at a sampling rate of 500 Hz. They removed artifacts and filtered the signal between 0.5 Hz and 45 Hz. We removed the EEG's aperiodic component by multiplying by f^∝, where∝ is the logarithmic slope of the average spectrum between 1 Hz and 35 Hz.
Model: We use a whole-brain model that consists of a network of delayed Kuramoto oscillators with coupling corresponding to an anatomical brain atlas. The used atlas is the Automated Anatomical Atlas with 90 regions, and we employed 40 Hz as the intrinsic frequency for all the oscillators (Cabral et al., 2014). Our simulations last 200 seconds for each set of parameters with a time step of 1 × 10^-3 s.
In addition, Miltiadous et al. 2023 also performed a cognitive and neuropsychological assessment. They employed the Mini-Mental State Examination (MMSE) (Creavin et al., 2016). Starting at 30 points, as lower the MMSE score, the more severe the cognitive decline. The MMSE score for the AD group was 17.75±4.5 and for the control group was 30.
The point of best correlation between rsEEG and model spectra corresponds to a value of global coupling K and mean delay MD in the parameters space. For each pair of {K, MD} also corresponds a value of spectral entropy (SE) and Kuramoto Order Parameter (KOP). SE is a metric of metastability of the model signals dependent on the spectrum, while KOP is a metric of global synchrony.
Results:
The results show the distributions of structural connectivity parameters from 36 AD subjects and 29 Control subjects across corresponding spectral entropy values.
AD subjects exhibit lower K [3.77 SD(1.39)] than Control subjects [4.74 SD(1.08)] (Wilcoxon rank-sum test, *p<0.01). However, both groups exhibit similar MD [AD: 22.5 ms SD(5.42); Control: 20.37 ms SD(4.58)]. The structural connectivity parameters of AD subjects correspond to lower metastability (p>0.09) and synchrony (*p<0.01) than the Control subjects. These findings align with Zheng et al., 2023, using the same database.
No significant correlation was found between K, MD, SE, or KOP and MMSE scores for AD subjects. This was expected as in the Control group all K, MD, SE, and KOP values correspond to MMSE = 30.
AD subjects exhibit lower K [3.77 SD(1.39)] than Control subjects [4.74 SD(1.08)] (Wilcoxon rank-sum test, *p<0.01). However, both groups exhibit similar MD [AD: 22.5 ms SD(5.42); Control: 20.37 ms SD(4.58)]. The structural connectivity parameters of AD subjects correspond to lower metastability (p>0.09) and synchrony (*p<0.01) than the Control subjects. These findings align with Zheng et al., 2023, using the same database.
No significant correlation was found between K, MD, SE, or KOP and MMSE scores for AD subjects. This was expected as in the Control group all K, MD, SE, and KOP values correspond to MMSE = 30.
·Fig1. Parameters position of EEG and model spectra corrlation, and model's metrics of dynamical features.
Conclusions:
The spectrum, synchrony, and metastability of AD rsEEG differ from Control. We provide quantitative metrics of the dynamic features by correlating the EEG's average spectrum with spectra from a whole-brain model. Fitting with the model, we also extract structural connectivity features, aligning with literature findings of connection loss in AD subjects.
In addition, Control subjects show less spread in the parameter space, and there are no AD subjects fitting close to the average point {K = 4.74, MD = 20.37 ms}. Therefore, we propose that fitting far from that point suggests a requirement for diagnosing brain disorders. Additional research is necessary to establish the extent of deviation from that point.
In addition, Control subjects show less spread in the parameter space, and there are no AD subjects fitting close to the average point {K = 4.74, MD = 20.37 ms}. Therefore, we propose that fitting far from that point suggests a requirement for diagnosing brain disorders. Additional research is necessary to establish the extent of deviation from that point.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s)
Modeling and Analysis Methods:
Classification and Predictive Modeling 1
EEG/MEG Modeling and Analysis 2
Keywords:
Computational Neuroscience
DISORDERS
Electroencephaolography (EEG)
Modeling
Open Data
Other - Alzheimer's disease; structural connectivity
1|2Indicates the priority used for review
Provide references using author date format
Exploring mechanisms of spontaneous functional connectivity in MEG: How delayed network interactions lead to
structured amplitude envelopes of band-pass filtered oscillations. NeuroImage, 90, 423–435. https://doi.org/10.1016/j.neuroimage.2013.11.047
Creavin ST, Wisniewski S, Noel-Storr AH, Trevelyan CM, Hampton T, Rayment D, Thom VM, Nash KJ, El-
hamoui H, Milligan R, Patel AS, Tsivos DV, Wing T, Phillips E, Kellman SM, Shackleton HL, Singleton GF, Neale
BE, Watton ME, Cullum S. (2016). Mini-Mental State Examination (MMSE) for the detection of dementia in clinically
unevaluated people aged 65 and over in community and primary care populations. Cochrane Database Syst Rev. doi: 10.1002/14651858.CD011145.pub2.
Miltiadous, A., Tzimourta, K.D., Afrantou, T., Ioannidis, P., Grigoriadis, N., Tsalikakis, D.G., Angelidis, P.,
Tsipouras, M.G., Glavas, E., Giannakeas, N. & Tzallas, A. (2023). A Dataset of Scalp EEG Recordings of Alzheimer’s Disease, Frontotemporal Dementia and Healthy Subjects from Routine EEG. Data, 8, 95. https://doi.org/10.3390/data8060095
Zheng, X., Wang, B., Liu, H., Wu, W., Sun, J., Fang, W., ... & Chen, S. S. C. (2023). Diagnosis of Alzheimer’s
disease via resting-state EEG: integration of spectrum, complexity, and synchronization signal features. Frontiers in
Aging Neuroscience, 15.