fMRI classification method with multiple feature fusion based on minimum spanning tree analysis

Classification of major depression disorder via using minimum spanning tree of individual high-order morphological brain network

BACKGROUND

Major depressive disorder (MDD) is an overbroad and heterogeneous diagnosis with no reliable or quantifiable markers. We aim to combine machine-learning techniques with the individual minimum spanning tree of the morphological brain network (MST-MBN) to determine whether the network properties can provide neuroimaging biomarkers to identify patients with MDD.

METHOD

Eight morphometric features of each region of interest (ROI) were extracted from 3D T1 structural images of 106 patients with MDD and 97 healthy controls. Six feature distances of the eight morphometric features were calculated to generate a feature distance matrix, which was defined as low-order MBN. Further linear correlations of feature distances between ROIs were calculated on the basis of low-order MBN to generate individual high-order MBN. The Kruskal's algorithm was used to generate the MST to obtain the core framework of individual low-order and high-order MBN. The regional and global properties of the individual MSTs were defined as the feature. The support vector machine and back-propagation neural network was used to diagnose MDD and assess its severity, respectively.

RESULT

The low-order and high-order MST-MBN constructed by cityblock distance had the excellent classification performance. The high-order MST-MBN significantly improved almost 20 % diagnostic accuracy compared with the low-order MST-MBN, and had a maximum R² value of 0.939 between the predictive and true Hamilton Depression Scale score. The different group-level connectivity strength mainly involves the central executive network and default mode network (no statistical significance after FDR correction).

CONCLUSION

We proposed an innovative individual high-order MST-MBN to capture the cortical high-order morphological correlation and make an excellent performance for individualized diagnosis and assessment of MDD.

Ensemble graph neural network model for classification of major depressive disorder using whole-brain functional connectivity

Major depressive disorder (MDD) is characterized by impairments in mood and cognitive functioning, and it is a prominent source of global disability and stress. A functional magnetic resonance imaging (fMRI) can aid clinicians in their assessments of individuals for the identification of MDD. Herein, we employ a deep learning approach to the issue of MDD classification. Resting-state fMRI data from 821 individuals with MDD and 765 healthy controls (HCs) is employed for investigation. An ensemble model based on graph neural network (GNN) has been created with the goal of identifying patients with MDD among HCs as well as differentiation between first-episode and recurrent MDDs. The graph convolutional network (GCN), graph attention network (GAT), and GraphSAGE models serve as a base models for the ensemble model that was developed with individual whole-brain functional networks. The ensemble's performance is evaluated using upsampling and downsampling, along with 10-fold cross-validation. The ensemble model achieved an upsampling accuracy of 71.18% and a downsampling accuracy of 70.24% for MDD and HC classification. While comparing first-episode patients with recurrent patients, the upsampling accuracy is 77.78% and the downsampling accuracy is 71.96%. According to the findings of this study, the proposed GNN-based ensemble model achieves a higher level of accuracy and suggests that our model produces can assist healthcare professionals in identifying MDD.

Spontaneous brain activity, graph metrics, and head motion related to prospective post-traumatic stress disorder trauma-focused therapy response

Introduction

Trauma-focused psychotherapy for post-traumatic stress disorder (PTSD) is effective in about half of all patients. Investigating biological systems related to prospective treatment response is important to gain insight in mechanisms predisposing patients for successful intervention. We studied if spontaneous brain activity, brain network characteristics and head motion during the resting state are associated with future treatment success.

Methods

Functional magnetic resonance imaging scans were acquired from 46 veterans with PTSD around the start of treatment. Psychotherapy consisted of trauma-focused cognitive behavioral therapy (tf-CBT), eye movement desensitization and reprocessing (EMDR), or a combination thereof. After intervention, 24 patients were classified as treatment responders and 22 as treatment resistant. Differences between groups in spontaneous brain activity were evaluated using amplitude of low-frequency fluctuations (ALFF), while global and regional brain network characteristics were assessed using a minimum spanning tree (MST) approach. In addition, in-scanner head motion was assessed.

Results

No differences in spontaneous brain activity and global network characteristics were observed between the responder and non-responder group. The right inferior parietal lobule, right putamen and left superior parietal lobule had a more central position in the network in the responder group compared to the non-responder group, while the right dorsolateral prefrontal cortex (DLPFC), right inferior frontal gyrus and left inferior temporal gyrus had a less central position. In addition, responders showed less head motion.

Discussion

These results show that areas involved in executive functioning, attentional and action processes, learning, and visual-object processing, are related to prospective PTSD treatment response in veterans. In addition, these findings suggest that involuntary micromovements may be related to future treatment success.

Depression Classification Using Frequent Subgraph Mining Based on Pattern Growth of Frequent Edge in Functional Magnetic Resonance Imaging Uncertain Network

The brain network structure is highly uncertain due to the noise in imaging signals and evaluation methods. Recent works have shown that uncertain brain networks could capture uncertain information with regards to functional connections. Most of the existing research studies covering uncertain brain networks used graph mining methods for analysis; for example, the mining uncertain subgraph patterns (MUSE) method was used to mine frequent subgraphs and the discriminative feature selection for uncertain graph classification (DUG) method was used to select discriminant subgraphs. However, these methods led to a lack of effective discriminative information; this reduced the classification accuracy for brain diseases. Therefore, considering these problems, we propose an approximate frequent subgraph mining algorithm based on pattern growth of frequent edge (unFEPG) for uncertain brain networks and a novel discriminative feature selection method based on statistical index (dfsSI) to perform graph mining and selection. Results showed that compared with the conventional methods, the unFEPG and dfsSI methods achieved a higher classification accuracy. Furthermore, to demonstrate the efficacy of the proposed method, we used consistent discriminative subgraph patterns based on thresholding and weighting approaches to compare the classification performance of uncertain networks and certain networks in a bidirectional manner. Results showed that classification performance of the uncertain network was superior to that of the certain network within a defined sparsity range. This indicated that if a better classification performance is to be achieved, it is necessary to select a certain brain network with a higher threshold or an uncertain brain network model. Moreover, if the uncertain brain network model was selected, it is necessary to make full use of the uncertain information of its functional connection.

Construction and Multiple Feature Classification Based on a High-Order Functional Hypernetwork on fMRI Data

Resting-state functional connectivity hypernetworks, in which multiple nodes can be connected, are an effective technique for diagnosing brain disease and performing classification research. Conventional functional hypernetworks can characterize the complex interactions within the human brain in a static form. However, an increasing body of evidence demonstrates that even in a resting state, neural activity in the brain still exhibits transient and subtle dynamics. These dynamic changes are essential for understanding the basic characteristics underlying brain organization and may correlate significantly with the pathological mechanisms of brain diseases. Therefore, considering the dynamic changes of functional connections in the resting state, we proposed methodology to construct resting state high-order functional hyper-networks (rs-HOFHNs) for patients with depression and normal subjects. Meanwhile, we also introduce a novel property (the shortest path) to extract local features with traditional local properties (cluster coefficients). A subgraph feature-based method was introduced to characterize information relating to global topology. Two features, local features and subgraph features that showed significant differences after feature selection were subjected to multi-kernel learning for feature fusion and classification. Compared with conventional hyper network models, the high-order hyper network obtained the best classification performance, 92.18%, which indicated that better classification performance can be achieved if we needed to consider multivariate interactions and the time-varying characteristics of neural interaction simultaneously when constructing a network.

Special Patterns of Dynamic Brain Networks Discriminate Between Face and Non-face Processing: A Single-Trial EEG Study

Face processing is a spatiotemporal dynamic process involving widely distributed and closely connected brain regions. Although previous studies have examined the topological differences in brain networks between face and non-face processing, the time-varying patterns at different processing stages have not been fully characterized. In this study, dynamic brain networks were used to explore the mechanism of face processing in human brain. We constructed a set of brain networks based on consecutive short EEG segments recorded during face and non-face (ketch) processing respectively, and analyzed the topological characteristic of these brain networks by graph theory. We found that the topological differences of the backbone of original brain networks (the minimum spanning tree, MST) between face and ketch processing changed dynamically. Specifically, during face processing, the MST was more line-like over alpha band in 0–100 ms time window after stimuli onset, and more star-like over theta and alpha bands in 100–200 and 200–300 ms time windows. The results indicated that the brain network was more efficient for information transfer and exchange during face processing compared with non-face processing. In the MST, the nodes with significant differences of betweenness centrality and degree were mainly located in the left frontal area and ventral visual pathway, which were involved in the face-related regions. In addition, the special MST patterns can discriminate between face and ketch processing by an accuracy of 93.39%. Our results suggested that special MST structures of dynamic brain networks reflected the potential mechanism of face processing in human brain.

The association between White matter microstructure alterations detected by Diffusional kurtosis imaging in Neural circuit and post-stroke depression

AIM

In order to study the mechanism of post-stroke depression (PSD), we used diffusion kurtosis imaging (DKI) to describe the changes in white matter (WM) microstructure in PSD patients, to investigate the association between WM damage in limbic-cortical-striatal-pallidal-thalamic (LCSPT) circuit and PSD, and the utility of DKI in the diagnosis of PSD.

METHODS

Fifty-eight participants were divided into different groups: control group (n=20), stroke patients without depression (Non-PSD, n=21) and PSD group (n=17). All were taken DKI scans. The WM of bilateral superior frontal gyrus, middle frontal gyrus, inferior frontal gyrus, temporal lobe, parietal lobe, occipital lobe, the anterior and posterior limb of internal capsule, the genu and splenium of corpus callosum were selected as the regions of interest (ROI) and selected mean kurtosis (MK), radial kurtosis (RK), axial kurtosis (AK) as the DKI parameters.

RESULTS

Compared with control and Non-PSD, MK of PSD group in bilateral superior frontal gyrus, middle frontal gyrus, temporal lobe and the genu of corpus callosum were decreased significantly, as well as the RK in left superior frontal gyrus, bilateral middle frontal gyrus and temporal lobe. But there was no significant difference in AK. Besides, the decrease in MK and RK in frontal and temporal lobe was negatively associated with the severity of PSD.

CONCLUSION

Our research indicated that the damage to WM microstructure in the frontal lobe, temporal lobe and the genu of corpus callosum may be related toPSD. DKI explores the microstructural changes of WM in PSD patients and may be an auxiliary diagnostic method for PSD.

Disrupted pathways from limbic areas to thalamus in schizophrenia highlighted by whole-brain resting-state effective connectivity analysis

BACKGROUND

Numerous neuroimaging studies have revealed that schizophrenia was characterized by wide-spread dysconnection among brain regions during rest measured by functional connectivity (FC). In contrast with FC, effective connectivity (EC) provides information about directionality of brain connections and is thus valuable in mechanistic investigation of schizophrenic brain. However, a systematic characterization of whole-brain resting-state EC (rsEC) and how it captures different information compared with resting-state FC (rsFC) in schizophrenia are still lacking.

AIMS

To systematically characterize the abnormalities of rsEC, compared with rsFC, in schizophrenia, and to test its discriminative power as a neuroimaging marker for schizophrenia diagnosis.

METHOD

Whole-brain rsEC and rsFC networks were constructed using resting-state fMRI data and compared between 103 patients with schizophrenia and 110 healthy participants. Pattern classifications between patients and controls based on whole-brain rsEC and rsFC were further performed using multivariate pattern analysis.

RESULTS

We identified 17 rsEC significantly disrupted (mostly decreased) in patients, among which all were associated with the thalamus and 15 were from limbic areas (including hippocampus, parahippocampus and cingulate cortex) to the thalamus. In contrast, abnormal rsFC were widely distributed in the whole brain. The classification accuracies for distinguishing patients and controls using whole-brain rsEC and rsFC patterns were 78.6% and 82.7%, respectively, and was further improved to 84.5% when combining rsEC and rsFC.

CONCLUSIONS

Schizophrenia is featured by disrupted 'limbic areas-to-thalamus' rsEC, in contrast with diffusively altered rsFC. Moreover, both rsEC and rsFC contain valuable and complementary information which may be used as diagnostic markers for schizophrenia.

EEG Feature Extraction Based on a Bilevel Network: Minimum Spanning Tree and Regional Network

Feature extraction is essential for classifying different motor imagery (MI) tasks in a brain–computer interface (BCI). Although the methods of brain network analysis have been widely studied in the BCI field, these methods are limited by differences in network size, density, and standardization. To address this issue and improve classification accuracy, we propose a novel method, in which the hybrid features of the brain function based on the bilevel network are extracted. Minimum spanning tree (MST) based on electroencephalogram (EEG) signal nodes in different MIs is constructed as the first network layer to solve the global network connectivity problem. In addition, the regional network in different movement patterns is constructed as the second network layer to determine the network characteristics, which is consistent with the correspondence between limb movement patterns and cerebral cortex in neurophysiology. We attempt to apply MST to the classification of the MI EEG signals, and the bilevel network has better interpretability. Thereafter, a vector is formed by combining the MST fundamental features with the directional features of the regional network. Our method is validated using the BCI Competition IV Dataset I. Experimental results verify the feasibility of the bilevel network framework. Furthermore, the average classification performance of the proposed method reaches 89.50%, which is higher than that of other competing methods, thereby indicating that the bilevel network is effective for MI classification.

Resting-State Functional Network Scale Effects and Statistical Significance-Based Feature Selection in Machine Learning Classification

In recent years, functional brain network topological features have been widely used as classification features. Previous studies have found that network node scale differences caused by different network parcellation definitions significantly affect the structure of the constructed network and its topological properties. However, we still do not know how network scale differences affect the classification accuracy, performance of classification features, and effectiveness of the feature selection strategy using P values in terms of the machine learning method. This study used five scale parcellations, involving 90, 256, 497, 1003, and 1501 nodes. Three local properties of resting-state functional brain networks were selected (degree, betweenness centrality, and nodal efficiency), and the support vector machine method was used to construct classifiers to identify patients with major depressive disorder. We analyzed the impact of the five scales on classification accuracy. In addition, the effectiveness and redundancy of features obtained by the different scale parcellations were compared. Finally, traditional statistical significance (P value) was verified as a feature selection criterion. The results showed that the feature effectiveness of different scales was similar; in other words, parcellation with more regions did not provide more effective discriminative features. Nevertheless, parcellation with more regions did provide a greater quantity of discriminative features, which led to an improvement in the accuracy of the classification. However, due to the close distance between brain regions, the redundancy of parcellation with more regions was also greater. The traditional P value feature selection strategy is feasible with different scales, but our analysis showed that the traditional P < 0.05 threshold was too strict for feature selection. This study provides an important reference for the selection of network scales when applying topological properties of brain networks to machine learning methods.

Detection of EEG K-Complexes Using Fractal Dimension of Time Frequency Images Technique Coupled With Undirected Graph Features

K-complexes identification is a challenging task in sleep research. The detection of k-complexes in electroencephalogram (EEG) signals based on visual inspection is time consuming, prone to errors, and requires well-trained knowledge. Many existing methods for k-complexes detection rely mainly on analyzing EEG signals in time and frequency domains. In this study, an efficient method is proposed to detect k-complexes from EEG signals based on fractal dimension (FD) of time frequency (T-F) images coupled with undirected graph features. Firstly, an EEG signal is partitioned into smaller segments using a sliding window technique. Each EEG segment is passed through a spectrogram of short time Fourier transform (STFT) to obtain the T-F images. Secondly, the box counting method is applied to each T-F image to discover the FDs in EEG signals. A vector of FD features are extracted from each T-F image and then mapped into an undirected graph. The structural properties of the graphs are used as the representative features of the original EEG signals for the input of a least square support vector machine (LS-SVM) classifier. Key graphic features are extracted from the undirected graphs. The extracted graph features are forwarded to the LS-SVM for classification. To investigate the classification ability of the proposed feature extraction combined with the LS-SVM classifier, the extracted features are also forwarded to a k-means classifier for comparison. The proposed method is compared with several existing k-complexes detection methods in which the same datasets were used. The findings of this study shows that the proposed method yields better classification results than other existing methods in the literature. An average accuracy of 97% for the detection of the k-complexes is obtained using the proposed method. The proposed method could lead to an efficient tool for the scoring of automatic sleep stages which could be useful for doctors and neurologists in the diagnosis and treatment of sleep disorders and for sleep research.