Translational approach for dementia subtype classification using convolutional neural network based on EEG connectome dynamics

Abstract

Dementia spectrum disorders, characterized by progressive cognitive decline, pose a significant global health burden. Early screening and diagnosis are essential for timely and accurate treatment, improving patient outcomes and quality of life. This study investigated dynamic features of resting-state electroencephalography (EEG) functional connectivity to identify characteristic patterns of dementia subtypes, such as Alzheimer's disease (AD) and frontotemporal dementia (FD), and to evaluate their potential as biomarkers. We extracted distinctive statistical features, including mean, variance, skewness, and Shannon entropy, from brain connectivity measures, revealing common alterations in dementia, specifically a generalized disruption of Alpha-band connectivity. Distinctive characteristics were found, including generalized Delta-band hyperconnectivity with increased complexity in AD and disrupted phase-based connectivity in Theta, Beta, and Gamma bands for FD. We also employed a convolutional neural network model, enhanced with these dynamic features, to differentiate between dementia subtypes. Our classification models achieved a multiclass classification accuracy of 93.6% across AD, FD, and healthy control groups. Furthermore, the model demonstrated 97.8% and 96.7% accuracy in differentiating AD and FD from healthy controls, respectively, and 97.4% accuracy in classifying AD and FD in pairwise classification. These establish a high-performance deep learning framework utilizing dynamic EEG connectivity patterns as potential biomarkers, offering a promising approach for early screening and diagnosis of dementia spectrum disorders using EEG. Our findings suggest that analyzing brain connectivity dynamics as a network and during cognitive tasks could offer more valuable information for diagnosis, assessing disease severity, and potentially identifying personalized neurological deficits.

Keywords: Aging disorders; Alzheimer’s disease; Brain connectivity; Convolutional neural network; Electroencephalogram; Frontotemporal dementia.

Fig. 1

The workflow of EEG connectome profile-based convolutional neural network approach for dementia subtype classification. (A) Derivation of connectivity dynamics profile map from EEG recording. (B) The convolutional neural network-based architecture of our classification model.

Fig. 2

Mean connectivity between dementia subtypes and healthy control. (A) Mean connectivity significance profile. Each color represents the preference classification group from the significance of mean connectivity in a brain wave band and feature between an electrode pair. The color intensity corresponds to the p-value of difference from the post-hoc analysis in a negative logarithmic scale. (B) Proportion of significant connections. Each bar represents the proportion of electrode pairs with a higher value of mean connectivity in the brain wave band and the feature preferring a specific group minus the proportion of electrode pairs that suggest otherwise. The proportion of + 1 indicates that the mean connectivity for the group in the study is higher than other comparative groups in all connections, while the proportion of − 1 indicates that the value is lower than other comparative groups in all connections. The inter-frontal and far-frontal refer to frontal-frontal and frontal-occipital connections, respectively. AD = Alzheimer’s Disease; FD = Frontotemporal Dementia; HC = Healthy Control; NS = Not significant; F = Frontal electrodes; O = Occipital electrodes; P = Parietal electrodes; T = Temporal electrodes.

Fig. 3

Variance of connectivity between dementia subtypes and healthy control. (A) Variance of connectivity significance profile. Each color represents the preference classification group from the significance of variance of connectivity in a brain wave band and feature between an electrode pair. The color intensity corresponds to the p-value of difference from the post-hoc analysis in a negative logarithmic scale. (B) Proportion of significant connections. Each bar represents the proportion of electrode pairs with a higher value of the variance of connectivity in the brain wave band and the feature preferring a specific group minus the proportion of electrode pairs that suggest otherwise. The proportion of + 1 indicates that the variance of connectivity for the group in the study is higher than other comparative groups in all connections, while the proportion of − 1 indicates that the value is lower than other comparative groups in all connections. The inter-frontal and far-frontal refer to frontal-frontal and frontal-occipital connections, respectively. AD = Alzheimer’s Disease; FD = Frontotemporal Dementia; HC = Healthy Control; NS = Not significant; F = Frontal electrodes; O = Occipital electrodes; P = Parietal electrodes; T = Temporal electrodes.

Fig. 4

Skewness of connectivity between dementia subtypes and healthy control. (A) The skewness of connectivity significance profile. Each color represents the preference classification group from the significance of skewness of connectivity in a brain wave band and feature between an electrode pair. The color intensity corresponds to the p-value of difference from the post-hoc analysis in a negative logarithmic scale. (B) Proportion of significant connections. Each bar represents the proportion of electrode pairs with a higher value of skewness of connectivity in the brain wave band and the feature preferring a specific group minus the proportion of electrode pairs that suggest otherwise. The proportion of + 1 indicates that the skewness of connectivity for the group in the study is higher than other comparative groups in all connections, while the proportion of − 1 indicates that the value is lower than other comparative groups in all connections. The inter-frontal and far-frontal refer to frontal-frontal and frontal-occipital connections, respectively. AD = Alzheimer’s Disease; FD = Frontotemporal Dementia; HC = Healthy Control; NS = Not significant; F = Frontal electrodes; O = Occipital electrodes; P = Parietal electrodes; T = Temporal electrodes.

Fig. 5

Shannon entropy of connectivity between dementia subtypes and healthy control. (A) Shannon entropy of connectivity significance profile. Each color represents the preference classification group from the significance of Shannon entropy of connectivity in a brain wave band and feature between an electrode pair. The color intensity corresponds to the p-value of difference from the post-hoc analysis in a negative logarithmic scale. (B) Proportion of significant connections. Each bar represents the proportion of electrode pairs with a higher value of Shannon entropy of connectivity in the brain wave band and the feature preferring a specific group minus the proportion of electrode pairs that suggest otherwise. The proportion of + 1 indicates that the Shannon entropy of connectivity for the group in the study is higher than other comparative groups in all connections, while the proportion of − 1 indicates that the value is lower than other comparative groups in all connections. The inter-frontal and far-frontal refer to frontal-frontal and frontal-occipital connections, respectively. AD = Alzheimer’s Disease; FD = Frontotemporal Dementia; HC = Healthy Control; NS = Not significant; F = Frontal electrodes; O = Occipital electrodes; P = Parietal electrodes; T = Temporal electrodes.

Fig. 6

Two-dimensional t-SNE visualization of convolutional features of the instance-wise model for AD/FD/HC task. A uniform jitter of diameter 5 was added to provide a better view of overlapping data points. The color of the data point represents an actual labeled group of each trial. Data points in the test set and training set, corresponding to each fold, are displayed in diamonds and points, respectively. Outliers are also displayed in cross symbols. AD = Alzheimer’s Disease; FD = Frontotemporal Dementia; HC = Healthy Control.