Deterioration from healthy to mild cognitive impairment and Alzheimer's disease mirrored in corresponding loss of centrality in directed brain networks

The Use of Generative Adversarial Network and Graph Convolution Network for Neuroimaging-Based Diagnostic Classification

Functional connectivity (FC) obtained from resting-state functional magnetic resonance imaging has been integrated with machine learning algorithms to deliver consistent and reliable brain disease classification outcomes. However, in classical learning procedures, custom-built specialized feature selection techniques are typically used to filter out uninformative features from FC patterns to generalize efficiently on the datasets. The ability of convolutional neural networks (CNN) and other deep learning models to extract informative features from data with grid structure (such as images) has led to the surge in popularity of these techniques. However, the designs of many existing CNN models still fail to exploit the relationships between entities of graph-structure data (such as networks). Therefore, graph convolution network (GCN) has been suggested as a means for uncovering the intricate structure of brain network data, which has the potential to substantially improve classification accuracy. Furthermore, overfitting in classifiers can be largely attributed to the limited number of available training samples. Recently, the generative adversarial network (GAN) has been widely used in the medical field for its generative aspect that can generate synthesis images to cope with the problems of data scarcity and patient privacy. In our previous work, GCN and GAN have been designed to investigate FC patterns to perform diagnosis tasks, and their effectiveness has been tested on the ABIDE-I dataset. In this paper, the models will be further applied to FC data derived from more public datasets (ADHD, ABIDE-II, and ADNI) and our in-house dataset (PTSD) to justify their generalization on all types of data. The results of a number of experiments show the powerful characteristic of GAN to mimic FC data to achieve high performance in disease prediction. When employing GAN for data augmentation, the diagnostic accuracy across ADHD-200, ABIDE-II, and ADNI datasets surpasses that of other machine learning models, including results achieved with BrainNetCNN. Specifically, in ADHD, the accuracy increased from 67.74% to 73.96% with GAN, in ABIDE-II from 70.36% to 77.40%, and in ADNI, reaching 52.84% and 88.56% for multiclass and binary classification, respectively. GCN also obtains decent results, with the best accuracy in ADHD datasets at 71.38% for multinomial and 75% for binary classification, respectively, and the second-best accuracy in the ABIDE-II dataset (72.28% and 75.16%, respectively). Both GAN and GCN achieved the highest accuracy for the PTSD dataset, reaching 97.76%. However, there are still some limitations that can be improved. Both methods have many opportunities for the prediction and diagnosis of diseases.

Cognitive Normal Older Adults with APOE-2 Allele Show a Distinctive Functional Connectivity Pattern in Response to Cerebral Aβ Deposition

The ε2 allele of apolipoprotein E (ε2) has neuroprotective effects against beta-amyloid (Aβ) pathology in Alzheimer's disease (AD). However, its impact on the functional connectivity and hub efficiency in cognitively normal older adults (CN) with ε2 is unclear. We investigated the functional connectivity differences in the default mode network (DMN), salience network, and central executive network (CEN) between A-PET-negative (N = 29) and A-PET-positive (N = 15) CNs with ε2/ε2 or ε2/ε3 genotypes. The A-PET-positive CNs exhibited a lower anterior DMN functional connectivity, higher posterior DMN functional connectivity, and increased CEN functional connectivity compared to the A-PET-negative CNs. Cerebral Aβ retention was negatively correlated with anterior DMN functional connectivity and positively correlated with posterior DMN and anterior CEN functional connectivity. A graph theory analysis showed that the A-PET-positive CNs displayed a higher betweenness centrality in the middle frontal gyrus (left) and medial fronto-parietal regions (left). The betweenness centrality in the middle frontal gyrus (left) was positively correlated with Aβ retention. Our findings reveal a reversed anterior-posterior dissociation in the DMN functional connectivity and heightened CEN functional connectivity in A-PET-positive CNs with ε2. Hub efficiencies, measured by betweenness centrality, were increased in the DMN and CEN of the A-PET-positive CNs with ε2. These results suggest unique functional connectivity responses to Aβ pathology in CN individuals with ε2.

Topologically convergent and divergent large scale complex networks among Alzheimer's disease spectrum patients: A systematic review

Alzheimer's disease (AD) is associated with disruption at the level of a large-scale complex network. To explore the underlying mechanisms in the progression of AD, graph theory was used to quantitatively analyze the topological properties of structural and functional connections. Although an increasing number of studies have shown altered global and nodal network properties, little is known about the topologically convergent and divergent patterns between structural and functional networks among AD-spectrum patients. In this review, we summarized the topological patterns of the large-scale complex networks using multimodal neuroimaging graph theory analysis in AD spectrum patients. Convergent deficits in the connectivity characteristics were primarily in the default mode network (DMN) itself both in the structural and functional networks, while a divergent changes in the neighboring regions of the DMN were also observed between the patient groups. Together, the application of graph theory to large-scale complex brain networks provides quantitative insights into topological principles of brain network organization, which may lead to increasing attention in identifying the underlying neuroimaging pathological changes and predicting the progression of AD.

Structural brain network correlations with amyloid burden in elderly individuals at risk of Alzheimer's disease

Alzheimer's disease (AD) has a long preclinical phase during which beta-amyloid accumulates in the brain without cognitive impairment. However, the pattern of brain network alterations in this early stage of the disease remains to be clarified. In this study we examined the relationships between regional brain network indices and beta-amyloid deposits. Twenty-four elderly subjects with the APOE4 allele underwent both a 1.5-Tesla magnetic resonance imaging (MRI) scan and a positron emission tomography (PET) scan using [¹⁸F]Florbetapir. We computed network metrics such as the degree, betweenness centrality, and clustering coefficient, and examined the relationships between the beta-amyloid accumulation and these regional brain network connectivity metrics. We found a significant positive correlation between the global standardized uptake value ratio (SUVR) of [¹⁸F]Florbetapir and the betweenness centrality in the left parietal region. However, there were no significant correlations between the SUVR score and other network indices or the regional gray matter volume. Our data suggest a relationship between the beta-amyloid accumulation and the regional brain network connectivity in subjects at risk of AD. The brain connectome may provide an adjunct biomarker for the early detection of AD.

Directed Network Defects in Alzheimer's Disease Using Granger Causality and Graph Theory

BACKGROUND

Few works studied the directed whole-brain interaction between different brain regions of Alzheimer's disease (AD). Here, we investigated the whole-brain effective connectivity and studied the graph metrics associated with AD.

METHODS

Large-scale Granger causality analysis was conducted to explore abnormal whole-brain effective connectivity of patients with AD. Moreover, graph-theoretical metrics including smallworldness, assortativity, and hierarchy, were computed from the effective connectivity network. Statistical analysis identified the aberrant network properties of AD subjects when compared against healthy controls.

RESULTS

Decreased small-worldness, and increased characteristic path length, disassortativity, and hierarchy were found in AD subjects.

CONCLUSION

This work sheds insight into the underlying neuropathological mechanism of the brain network of AD individuals such as less efficient information transmission and reduced resilience to a random or targeted attack.

A Novel Connectome-Based Electrophysiological Study of Subjective Cognitive Decline Related to Alzheimer's Disease by Using Resting-State High-Density EEG EGI GES 300

Aim: To investigate for the first time the brain network in the Alzheimer’s disease (AD) spectrum by implementing a high-density electroencephalography (HD-EEG - EGI GES 300) study with 256 channels in order to seek if the brain connectome can be effectively used to distinguish cognitive impairment in preclinical stages. Methods: Twenty participants with AD, 30 with mild cognitive impairment (MCI), 20 with subjective cognitive decline (SCD) and 22 healthy controls (HC) were examined with a detailed neuropsychological battery and 10 min resting state HD-EEG. We extracted correlation matrices by using Pearson correlation coefficients for each subject and constructed weighted undirected networks for calculating clustering coefficient (CC), strength (S) and betweenness centrality (BC) at global (256 electrodes) and local levels (29 parietal electrodes). Results: One-way ANOVA presented a statistically significant difference among the four groups at local level in CC [F (3, 88) = 4.76, p = 0.004] and S [F (3, 88) = 4.69, p = 0.004]. However, no statistically significant difference was found at a global level. According to the independent sample t-test, local CC was higher for HC [M (SD) = 0.79 (0.07)] compared with SCD [M (SD) = 0.72 (0.09)]; t (40) = 2.39, p = 0.02, MCI [M (SD) = 0.71 (0.09)]; t (50) = 0.41, p = 0.004 and AD [M (SD) = 0.68 (0.11)]; t (40) = 3.62, p = 0.001 as well, while BC showed an increase at a local level but a decrease at a global level as the disease progresses. These findings provide evidence that disruptions in brain networks in parietal organization may potentially represent a key factor in the ability to distinguish people at early stages of the AD continuum. Conclusions: The above findings reveal a dynamically disrupted network organization of preclinical stages, showing that SCD exhibits network disorganization with intermediate values between MCI and HC. Additionally, these pieces of evidence provide information on the usefulness of the 256 HD-EEG in network construction.