Classification of Schizophrenia using Intrinsic Connectivity Networks and Incremental Boosting Convolution Neural Networks.

Background: In functional magnetic resonance imaging (fMRI), functional network connectivity (FNC) captures temporal coupling among intrinsic connectivity networks (ICNs). Traditional FNC analyses often rely on linear models, which may overlook complex nonlinear interactions. We propose a multi-layered neural network that generates nonlinear heatmaps from FNC matrices, which we visualize at multiple layers, enabling us to better characterize multi-level interactions and improve interpretability. Methods: Our approach consists of two training stages. In the first, a deep convolutional neural network (DCNN) is trained to produce heatmaps from multiple convolution layers. In the second, a t-test-based feature selection identifies relevant heatmaps that help distinguish different groups. In addition, we introduce ‘source-based features’ which summarize the multi-layer model output using an independent component analysis-based procedure that provides valuable, interpretable insights into the specific layer outputs. We tested this approach on a large dataset of schizophrenia patients and healthy controls, split into training and validation sets. Furthermore, this method clarifies how underlying neural mechanisms differ between schizophrenia patients and healthy controls, revealing crucial patterns in the default mode and visual networks. Results: The results indicate increased default mode network connectivity with itself and cognitive control regions in patients, while controls showed stronger visual and sensorimotor connectivity. Our DCNN approach achieved 92.8% cross-validated classification accuracy, outperforming competing methods. We also separated individuals into three cognitive performance groups based on cognitive scores and showed that the model can accurately predict the cognitive level using the FNC data. Conclusion: Our novel approach demonstrates the advantage of employing more sophisticated models in characterizing complex brain connectivity patterns while enhancing the interpretability of results. These findings underscore the significance of modeling nonlinear dynamics in fMRI analysis, shedding new light on the intricate interplays underlying cognitive and psychiatric phenomena.

The explorations of brain functional connectivity network (FCN) using resting-state functional magnetic resonance imaging can provide crucial insights into discriminative analysis of neuropsychiatric disorders, such as schizophrenia (SZ). Pearson's correlation (PC) is widely used to construct a densely connected FCN which may overlook some complex interactions of paired regions of interest (ROIs) under confounding effect of other ROIs. Although the method of sparse representation takes into account this issue, it penalizes each edge equally, which often makes the FCN look like a random network. In this paper, we establish a new framework, called convolutional neural network with sparsity-guided multiple functional connectivity, for SZ classification. The framework consists of two components. (1) The first component constructs a sparse FCN by integrating PC and weighted sparse representation (WSR). The FCN retains the intrinsic correlation between paired ROIs, and eliminates false connection simultaneously, resulting in sparse interactions among multiple ROIs with the confounding effect regressed out. (2) In the second component, we develop a functional connectivity convolution to learn discriminative features for SZ classification from multiple FCNs by mining the joint spatial mapping of FCNs. Finally, an occlusion strategy is employed to explore the contributive regions and connections, to derive the potential biomarkers in identifying associated aberrant connectivity of SZ. The experiments on SZ identification verify the rationality and advantages of our proposed method. This framework also can be used as a diagnostic tool for other neuropsychiatric disorders.

In functional magnetic resonance imaging (fMRI) studies, it is common to evaluate the brain's functional network connectivity (FNC) which captures the temporal coupling between hemodynamic signals. FNC has been linked to various psychological phenomena. However, current FNCs mainly represent linear statistical relationships, which may not capture the fully complexity of the interactions among brain intrinsic connectivity networks (ICNs). Therefore, it is crucial to explore approaches that can better account for possible intricate nonlinear interactions involved in cognitive operations and the changes observed in psychiatric conditions such as schizophrenia. This exploration can lead to a better understanding of brain function and provide crucial insights into the underlying mechanisms of various psychological and psychiatric conditions. In this paper, we present an innovative approach which utilizes a deep convolutional neural network (DCNN) to extract nonlinear heatmaps from functional network connectivity matrices. By analyzing the heatmaps, multi-level nonlinear interactions can be derived from the corresponding input FNC data. Our results show these networks represent a significant improvement over previous approaches and offer a robust framework for understanding the complex inter-actions between brain regions. By incorporating two stages in the training process, our method ensures optimal efficiency and effectiveness. In the initial stage, a deep convolutional neural network is trained to create heatmaps from various convolution layers of the network. In the next stage, by utilizing a t-test-based feature selection method, we can effectively analyze each heatmap from different convolution layers. This approach ensures that we are able to extract linear and nonlinear functional connectivity from the heatmaps that play an important role in distinguishing different groups. We used a large dataset consisting of both schizophrenia patients and healthy controls, which were divided into separate training and validation sets to evaluate this approach. Our system shows the potential to accurately distinguish differences between the schizophrenia and healthy control groups with high accuracy.

Hypertension is one of the main risk factors for vascular dementia and Alzheimer disease. To predict the onset of these diseases, it is necessary to develop tools to detect the early effects of vascular risk factors on the brain. Resting-state functional magnetic resonance imaging can investigate how the brain modulates its resting activity and analyze how hypertension impacts cerebral function. Here, we used resting-state functional magnetic resonance imaging to explore brain functional-hemodynamic coupling across different regions and their connectivity in patients with hypertension, as compared to subjects with normotension. In addition, we leveraged multimodal imaging to identify the signature of hypertension injury on the brain. Our study included 37 subjects (18 normotensives and 19 hypertensives), characterized by microstructural integrity by diffusion tensor imaging and cognitive profile, who were subjected to resting-state functional magnetic resonance imaging analysis. We mapped brain functional connectivity networks and evaluated the connectivity differences among regions, identifying the altered connections in patients with hypertension compared with subjects with normotension in the (1) dorsal attention network and sensorimotor network; (2) dorsal attention network and visual network; (3) dorsal attention network and frontoparietal network. Then we tested how diffusion tensor imaging fractional anisotropy of superior longitudinal fasciculus correlates with the connections between dorsal attention network and default mode network and Montreal Cognitive Assessment scores with a widespread network of functional connections. Finally, based on our correlation analysis, we applied a feature selection to highlight those most relevant to describing brain injury in patients with hypertension. Our multimodal imaging data showed that hypertensive brains present a network of functional connectivity alterations that correlate with cognitive dysfunction and microstructural integrity. Registration- URL: https://www.clinicaltrials.gov; Unique identifier: NCT02310217.

Deep learning-based classification and mutation prediction from histopathological images of hepatocellular carcinoma.

As a neural fingerprint, functional connectivity networks (FCNs) have been used to identify subjects from group. However, a number of studies have only paid attention to cerebral cortex when constructing the brain FCN. Other areas of the brain also play important roles in brain activities. It is widely accepted that the human brain is composed of many highly complex functional networks of cortex. Moreover, recent studies have confirmed correlations between signals of cortex and white matter (WM) bundles. Therefore, it is difficult to reflect the functional characteristics of the brain through a single-layer FCN. In this paper, a multilayer FCN involving both cerebral cortex and WM was proposed to characterize different levels of FCNs based on resting-state functional Magnetic Resonance Imaging (fMRI) data. The multilayer FCN is regarded as a neural fingerprint to realize individual identification and the individual variability of which is related to individual differences in fluid intelligence and gender. In addition, the proposed multilayer FCN has been learned by a multimodal convolutional neural network (CNN) algorithm to achieve individual identification. Compared with the existing FCN that only involves the cerebral cortex, the accuracy of individual recognition illustrates the effectiveness of our method. Meanwhile, the proposed multilayer FCN also achieves better results than the traditional single-layer FCN in gender classification and fluid intelligence prediction.

With the development of resting-state functional magnetic resonance imaging (rs-fMRI) technology, the functional connectivity network (FCN) which reflects the statistical similarity of temporal activity between brain regions has shown promising results for the identification of neuropsychiatric disorders. Alteration in FCN is believed to have the potential to locate biomarkers for classifying or predicting schizophrenia (SZ) from healthy control. However, the traditional FCN analysis with stationary assumption, i.e., static functional connectivity network (SFCN) at the time only measures the simple functional connectivity among brain regions, ignoring the dynamic changes of functional connectivity and the high-order dynamic interactions. In this article, the dynamic functional connectivity network (DFCN) is constructed to delineate the characteristic of connectivity variation across time. A high-order functional connectivity network (HFCN) designed based on DFCN, could characterize more complex spatial interactions across multiple brain regions with the potential to reflect complex functional segregation and integration. Specifically, the temporal variability and the high-order network topology features, which characterize the brain FCNs from region and connectivity aspects, are extracted from DFCN and HFCN, respectively. Experiment results on SZ identification prove that our method is more effective (i.e., obtaining a significantly higher classification accuracy, 81.82%) than other competing methods. Post hoc inspection of the informative features in the individualized classification task further could serve as the potential biomarkers for identifying associated aberrant connectivity in SZ.

Current functional magnetic resonance imaging (fMRI) data modeling techniques, such as independent component analysis and sparse coding methods, can effectively reconstruct dozens or hundreds of concurrent interacting functional brain networks simultaneously from the whole brain fMRI signals. However, such reconstructed networks have no correspondences across different subjects. Thus, automatic, effective, and accurate classification and recognition of these large numbers of fMRI-derived functional brain networks are very important for subsequent steps of functional brain analysis in cognitive and clinical neuroscience applications. However, this task is still a challenging and open problem due to the tremendous variability of various types of functional brain networks and the presence of various sources of noises. In recognition of the fact that convolutional neural networks (CNN) has superior capability of representing spatial patterns with huge variability and dealing with large noises, in this paper, we design, apply, and evaluate a deep 3-D CNN framework for automatic, effective, and accurate classification and recognition of large number of functional brain networks reconstructed by sparse representation of whole-brain fMRI signals. Our extensive experimental results based on the Human Connectome Project fMRI data showed that the proposed deep 3-D CNN can effectively and robustly perform functional networks classification and recognition tasks, while maintaining a high tolerance for mistakenly labeled training instances. This study provides a new deep learning approach for modeling functional connectomes based on fMRI data.

Decision letter: Disruption in structural–functional network repertoire and time-resolved subcortical fronto-temporoparietal connectivity in disorders of consciousness

Neuroimaging studies have shown that functional brain networks composed from select regions of interest have a modular community structure. However, the organization of functional network connectivity (FNC), comprising a purely data-driven network built from spatially independent brain components, is not yet clear. The aim of this study is to explore the modular organization of FNC in both healthy controls (HCs) and patients with schizophrenia (SZs). Resting state functional magnetic resonance imaging data of HCs and SZs were decomposed into independent components (ICs) by group independent component analysis (ICA). Then weighted brain networks (in which nodes are brain components) were built based on correlations between ICA time courses. Clustering coefficients and connectivity strength of the networks were computed. A dynamic branch cutting algorithm was used to identify modules of the FNC in HCs and SZs. Results show stronger connectivity strength and higher clustering coefficient in HCs with more and smaller modules in SZs. In addition, HCs and SZs had some different hubs. Our findings demonstrate altered modular architecture of the FNC in schizophrenia and provide insights into abnormal topological organization of intrinsic brain networks in this mental illness.

Dynamic functional network connectivity (dFNC) analysis is a widely used approach for studying brain function and offering insight into how brain networks evolve over time. Typically, dFNC studies utilize fixed spatial maps and evaluate transient changes in coupling among time courses estimated from independent component analysis (ICA). This manuscript presents a complementary approach that relaxes this assumption by spatially reordering the components dynamically at each time point to optimize for a smooth gradient in the FNC (i.e., a smooth gradient among ICA connectivity values). Several methods are presented to summarize dynamic FNC gradients (dFNGs) over time, starting with static FNC gradients (sFNGs), then exploring the reordering properties as well as the dynamics of the gradients themselves. We then apply this approach to a dataset of schizophrenia (SZ) patients and healthy controls (HCs). Functional dysconnectivity between different brain regions has been reported in SZ, yet the neural mechanisms behind it remain elusive. Using resting-state fMRI and ICA on a dataset consisting of 151 SZ patients and 160 age and gender-matched HCs, we extracted 53 intrinsic connectivity networks (ICNs) for each subject using a fully automated spatially constrained ICA approach. We develop several summaries of our functional network connectivity gradient analysis, both in a static sense, computed as the Pearson correlation coefficient between full time series, and a dynamic sense, computed using a sliding window approach followed by reordering based on the computed gradient, and evaluate group differences. Static connectivity analysis revealed significantly stronger connectivity between subcortical (SC), auditory (AUD), and visual (VIS) networks in patients, as well as hypoconnectivity in the sensorimotor (SM) network relative to controls. sFNG analysis highlighted distinctive clustering patterns in patients and HCs along cognitive control (CC)/default mode network (DMN), as well as SC/AUD/SM/cerebellar (CB) and VIS gradients. Furthermore, we observed significant differences in the sFNGs between groups in SC and CB domains. dFNG analysis suggested that SZ patients spend significantly more time in a SC/CB state based on the first gradient, while HCs favor the SM/DMN state. For the second gradient, however, patients exhibited significantly higher activity in CB domains, contrasting with HCs' DMN engagement. The gradient synchrony analysis conveyed more shifts between SM/SC networks and transmodal CC/DMN networks in patients. In addition, the dFNG coupling revealed distinct connectivity patterns between SC, SM, and CB domains in SZ patients compared to HCs. To recap, our results advance our understanding of brain network modulation by examining smooth connectivity trajectories. This provides a more complete spatiotemporal summary of the data, contributing to the growing body of current literature regarding the functional dysconnectivity in SZ patients. By employing dFNG, we highlight a new perspective to capture large-scale fluctuations across the brain while maintaining the convenience of brain networks and low-dimensional summary measures.

In the field of image classification, the Convolutional Neural Networks (CNNs) are effective. Most of the work focuses on improving and innovating CNN’s network structure. However, using labeled data more effectively for training has also been an essential part of CNN’s research. Combining image disturbance and consistency regularization theory, this paper proposes a model training method (PairTraining) that takes image pairs as input and dynamically modify the training difficulty according to the accuracy of the model in the training set. According to the accuracy of the model in the training set, the training process will be divided into three stages: the qualitative stage, the fine learning stage and the strengthening learning stage. Contrastive learning images are formed using a progressively enhanced image disturbance strategy at different training stages. The input image and contrast learning image are combined into image pairs for model training. The experiments are tested on four public datasets using eleven CNN models. These models have different degrees of improvement in accuracy on the four datasets. PairTraining can adapt to a variety of CNN models for image classification training. This method can better improve the effectiveness of training and improve the degree of generalization of classification models after training. The classification model obtained by PairTraining has better performance in practical application.

Brain Connectivity: A Journal of Clinical Neurology, Neuroscience, and Neuroimaging Advancing the Field of Neurology.

Artificial neural networks (ANN) which include deep learning neural networks (DNN) have problems such as the local minimal problem of Back propagation neural network (BPNN), the unstable problem of Radial basis function neural network (RBFNN) and the limited maximum precision problem of Convolutional neural network (CNN). Performance (training speed, precision, etc.) of BPNN, RBFNN and CNN are expected to be improved. Main works are as follows: Firstly, based on existing BPNN and RBFNN, Wavelet neural network (WNN) is implemented in order to get better performance for further improving CNN. WNN adopts the network structure of BPNN in order to get faster training speed. WNN adopts the wavelet function as an activation function, whose form is similar to the radial basis function of RBFNN, in order to solve the local minimum problem. Secondly, WNN-based Convolutional wavelet neural network (CWNN) method is proposed, in which the fully connected layers (FCL) of CNN is replaced by WNN. Thirdly, comparative simulations based on MNIST and CIFAR-10 datasets among the discussed methods of BPNN, RBFNN, CNN and CWNN are implemented and analyzed. Fourthly, the wavelet-based Convolutional Neural Network (WCNN) is proposed, where the wavelet transformation is adopted as the activation function in Convolutional Pool Neural Network (CPNN) of CNN. Fifthly, simulations based on CWNN are implemented and analyzed on the MNIST dataset. Effects are as follows: Firstly, WNN can solve the problems of BPNN and RBFNN and have better performance. Secondly, the proposed CWNN can reduce the mean square error and the error rate of CNN, which means CWNN has better maximum precision than CNN. Thirdly, the proposed WCNN can reduce the mean square error and the error rate of CWNN, which means WCNN has better maximum precision than CWNN.

Alterations in fMRI-based brain functional network connectivity (FNC) are associated with schizophrenia (SCZ) and the genetic risk or subthreshold clinical symptoms preceding the onset of SCZ, which often occurs in early adulthood. Thus, age-sensitive FNC changes may be relevant to SCZ risk-related FNC. We used independent component analysis to estimate FNC from childhood to adulthood in 9,236 individuals. To capture individual brain features more accurately than single-session fMRI, we studied an average of three fMRI scans per individual. To identify potential familial risk-related FNC changes, we compared age-related FNC in first-degree relatives of SCZ patients mostly including unaffected siblings (SIB) with neurotypical controls (NC) at the same age stage. Then, we examined how polygenic risk scores for SCZ influenced risk-related FNC patterns. Finally, we investigated the same risk-related FNC patterns in adult SCZ patients (oSCZ) and young individuals with subclinical psychotic symptoms (PSY). Age-sensitive risk-related FNC patterns emerge during adolescence and early adulthood, but not before. Young SIB always followed older NC patterns, with decreased FNC in a cerebellar-occipitoparietal circuit and increased FNC in two prefrontal-sensorimotor circuits when compared to young NC. Two of these FNC alterations were also found in oSCZ, with one exhibiting reversed pattern. All were linked to polygenic risk for SCZ in unrelated individuals (R2 varied from 0.02 to 0.05). Young PSY showed FNC alterations in the same direction as SIB when compared to NC. These results suggest that age-related neurotypical FNC correlates with genetic risk for SCZ and is detectable with MRI in young participants.