Diagnostic classification of intrinsic functional connectivity highlights somatosensory, default mode, and visual regions in autism

Copy number variants (CNVs): a powerful tool for iPSC-based modelling of ASD

Patients diagnosed with chromosome microdeletions or duplications, known as copy number variants (CNVs), present a unique opportunity to investigate the relationship between patient genotype and cell phenotype. CNVs have high genetic penetrance and give a good correlation between gene locus and patient clinical phenotype. This is especially effective for the study of patients with neurodevelopmental disorders (NDD), including those falling within the autism spectrum disorders (ASD). A key question is whether this correlation between genetics and clinical presentation at the level of the patient can be translated to the cell phenotypes arising from the neurodevelopment of patient induced pluripotent stem cells (iPSCs).Here, we examine how iPSCs derived from ASD patients with an associated CNV inform our understanding of the genetic and biological mechanisms underlying the aetiology of ASD. We consider selection of genetically characterised patient iPSCs; use of appropriate control lines; aspects of human neurocellular biology that can capture in vitro the patient clinical phenotype; and current limitations of patient iPSC-based studies. Finally, we consider how future research may be enhanced to maximise the utility of CNV patients for research of pathological mechanisms or therapeutic targets.

Self-weighted adaptive structure learning for ASD diagnosis via multi-template multi-center representation

As a kind of neurodevelopmental disease, autism spectrum disorder (ASD) can cause severe social, communication, interaction, and behavioral challenges. To date, many imaging-based machine learning techniques have been proposed to address ASD diagnosis issues. However, most of these techniques are restricted to a single template or dataset from one imaging center. In this paper, we propose a novel multi-template multi-center ensemble classification scheme for automatic ASD diagnosis. Specifically, based on different pre-defined templates, we construct multiple functional connectivity (FC) brain networks for each subject based on our proposed Pearson's correlation-based sparse low-rank representation. After extracting features from these FC networks, informative features to learn optimal similarity matrix are then selected by our self-weighted adaptive structure learning (SASL) model. For each template, the SASL method automatically assigns an optimal weight learned from the structural information without additional weights and parameters. Finally, an ensemble strategy based on the multi- template multi-center representations is applied to derive the final diagnosis results. Extensive experiments are conducted on the publicly available Autism Brain Imaging Data Exchange (ABIDE) database to demonstrate the efficacy of our proposed method. Experimental results verify that our proposed method boosts ASD diagnosis performance and outperforms state-of-the-art methods.

Application of Supervised Machine Learning for Behavioral Biomarkers of Autism Spectrum Disorder Based on Electrodermal Activity and Virtual Reality

OBJECTIVE

Sensory processing is the ability to capture, elaborate, and integrate information through the five senses and is impaired in over 90% of children with autism spectrum disorder (ASD). The ASD population shows hyper-hypo sensitiveness to sensory stimuli that can generate alteration in information processing, affecting cognitive and social responses to daily life situations. Structured and semi-structured interviews are generally used for ASD assessment, and the evaluation relies on the examiner's subjectivity and expertise, which can lead to misleading outcomes. Recently, there has been a growing need for more objective, reliable, and valid diagnostic measures, such as biomarkers, to distinguish typical from atypical functioning and to reliably track the progression of the illness, helping to diagnose ASD. Implicit measures and ecological valid settings have been showing high accuracy on predicting outcomes and correctly classifying populations in categories.

METHODS

Two experiments investigated whether sensory processing can discriminate between ASD and typical development (TD) populations using electrodermal activity (EDA) in two multimodal virtual environments (VE): forest VE and city VE. In the first experiment, 24 children with ASD diagnosis and 30 TDs participated in both virtual experiences, and changes in EDA have been recorded before and during the presentation of visual, auditive, and olfactive stimuli. In the second experiment, 40 children have been added to test the model of experiment 1.

RESULTS

The first exploratory results on EDA comparison models showed that the integration of visual, auditive, and olfactive stimuli in the forest environment provided higher accuracy (90.3%) on sensory dysfunction discrimination than specific stimuli. In the second experiment, 92 subjects experienced the forest VE, and results on 72 subjects showed that stimuli integration achieved an accuracy of 83.33%. The final confirmatory test set (n = 20) achieved 85% accuracy, simulating a real application of the models. Further relevant result concerns the visual stimuli condition in the first experiment, which achieved 84.6% of accuracy in recognizing ASD sensory dysfunction.

CONCLUSION

According to our studies' results, implicit measures, such as EDA, and ecological valid settings can represent valid quantitative methods, along with traditional assessment measures, to classify ASD population, enhancing knowledge on the development of relevant specific treatments.

Attentional Connectivity-based Prediction of Autism Using Heterogeneous rs-fMRI Data from CC200 Atlas

Autism spectrum disorder (ASD) is a developmental syndrome characterized by obvious drawbacks in sociality and communication. It has crucial significance to exactly discern the individuals with ASD and typical controls (TC). Previous imaging studies on ASD/TC identification have made remarkable progress in the exploration of objective as well as crucial biomarkers associated with ASD. However, glaring deficiency is manifested by the investigation on solely homogeneous and small datasets. Thus, we attempted to unveil some replicable and robust neural patterns of autism using a heterogeneous multi-site brain imaging dataset from ABIDE (Autism Brain Imaging Data Exchange). Experiments were carried out with an attention mechanism based on Extra-Trees algorithm, taking the study object of brain connectivity measured with the resting-state functional magnetic resonance imaging (fMRI) data of CC200 atlas. With cross-validation strategy, our proposed method resulted in a mean classification accuracy of 72.2% (sensitivity=68.6%, specificity=75.4%). It raised the precision of ASD prediction by about 2% and specificity by 3.2% in comparison with the most competitive reported effort. Connectivity analysis on the optimal model highlighted informative regions strongly involved in the social cognition as well as interaction, and manifested lower correlation between the anterior and posterior default mode network (DMN) in autistic individuals than controls. This observation is concordant with previous studies, which enables our proposed method to effectively identify the individuals with risk of ASD.

Improving the detection of autism spectrum disorder by combining structural and functional MRI information

•

We present an approach for autism classification based on neuroimaging MRI.

•

The pipeline relies on connectivity matrices and machine learning techniques.

•

Accuracy is 85.06  ±  3.52% evaluated in more than 800 cases of the ABIDE I dataset.

•

The most important correlations for autism classification are highlighted.

•

Merging functional and structural information outperforms the monomodal pipelines.

Autism Spectrum Disorder (ASD) is a brain disorder that is typically characterized by deficits in social communication and interaction, as well as restrictive and repetitive behaviors and interests. During the last years, there has been an increase in the use of magnetic resonance imaging (MRI) to help in the detection of common patterns in autism subjects versus typical controls for classification purposes. In this work, we propose a method for the classification of ASD patients versus control subjects using both functional and structural MRI information. Functional connectivity patterns among brain regions, together with volumetric correspondences of gray matter volumes among cortical parcels are used as features for functional and structural processing pipelines, respectively. The classification network is a combination of stacked autoencoders trained in an unsupervised manner and multilayer perceptrons trained in a supervised manner. Quantitative analysis is performed on 817 cases from the multisite international Autism Brain Imaging Data Exchange I (ABIDE I) dataset, consisting of 368 ASD patients and 449 control subjects and obtaining a classification accuracy of 85.06 ± 3.52% when using an ensemble of classifiers. Merging information from functional and structural sources significantly outperforms the implemented individual pipelines.

A network clustering based feature selection strategy for classifying autism spectrum disorder

BACKGROUND

Advanced non-invasive neuroimaging techniques offer new approaches to study functions and structures of human brains. Whole-brain functional networks obtained from resting state functional magnetic resonance imaging has been widely used to study brain diseases like autism spectrum disorder (ASD). Auto-classification of ASD has become an important issue. Existing classification methods for ASD are based on features extracted from the whole-brain functional networks, which may be not discriminant enough for good performance.

METHODS

In this study, we propose a network clustering based feature selection strategy for classifying ASD. In our proposed method, we first apply symmetric non-negative matrix factorization to divide brain networks into four modules. Then we extract features from one of four modules called default mode network (DMN) and use them to train several classifiers for ASD classification.

RESULTS

The computational experiments show that our proposed method achieves better performances than those trained with features extracted from the whole brain network.

CONCLUSION

It is a good strategy to train the classifiers for ASD based on features from the default mode subnetwork.

Multivariate models of brain volume for identification of children and adolescents with fetal alcohol spectrum disorder

Magnetic resonance imaging (MRI) studies of fetal alcohol spectrum disorder (FASD) have shown reductions of brain volume associated with prenatal exposure to alcohol. Previous studies consider regional brain volumes independently but ignore potential relationships across numerous structures. This study aims to (a) identify a multivariate model based on regional brain volume that discriminates children/adolescents with FASD versus healthy controls, and (b) determine if FASD classification performance can be increased by building classification models separately for each sex. Three‐dimensional T1‐weighted MRI from two independent childhood/adolescent datasets were used for training (79 FASD, aged 5.7–18.9 years, 35 males; 81 controls, aged 5.8–18.5 years, 32 males) and testing (67 FASD, aged 6.0–19.6 years, 38 males; 74 controls, aged 5.2–19.5 years, 42 males) a classification model. Using FreeSurfer, 87 regional brain volumes were extracted for each subject and were used as input into a support vector machine generating a classification model from the training data. The model performed moderately well on the test data with accuracy 77%, sensitivity 64%, and specificity 88%. Regions that contributed heavily to prediction in this model included temporal lobe and subcortical gray matter. Further investigation of two separate models for males and females showed slightly decreased accuracy compared to the model including all subjects (male accuracy 70%; female accuracy 67%), but had different regional contributions suggesting sex differences. This work demonstrates the potential of multivariate analysis of brain volumes for discriminating children/adolescents with FASD and provides indication of the most affected regions.

Biomarkers for Autism Spectrum Disorders (ASD): A Meta-analysis

OBJECTIVE

To compare the reported accuracy and sensitivity of the various modalities used to diagnose autism spectrum disorders (ASD) in efforts to help focus further biomarker research on the most promising methods for early diagnosis.

METHODS

The Medline scientific literature database was searched to identify publications assessing potential clinical ASD biomarkers. Reports were categorized by the modality used to assess the putative markers, including protein, genetic, metabolic, or objective imaging methods. The reported sensitivity, specificity, area under the curve, and overall agreement were summarized and analyzed to determine weighted averages for each diagnostic modality. Heterogeneity was measured using the I² test.

RESULTS

Of the 71 papers included in this analysis, each belonging to one of five modalities, protein-based followed by metabolite-based markers provided the highest diagnostic accuracy, each with a pooled overall agreement of 83.3% and respective weighted area under the curve (AUC) of 89.5% and 88.3%. Sensitivity provided by protein markers was highest (85.5%), while metabolic (85.9%) and protein markers (84.7%) had the highest specificity. Other modalities showed degrees of sensitivity, specificity, and overall agreements in the range of 73%-80%.

CONCLUSIONS

Each modality provided for diagnostic accuracy and specificity similar or slightly higher than those reported for the gold-standard Autism Diagnostic Observation Schedule (ADOS) instrument. Further studies are required to identify the most predictive markers within each modality and to evaluate biological pathways or clustering with possible etiological relevance. Analyses will also be necessary to determine the potential of these novel biomarkers in diagnosing pediatric patients, thereby enabling early intervention.

Ensemble learning with 3D convolutional neural networks for functional connectome-based prediction

The specificity and sensitivity of resting state functional MRI (rs-fMRI) measurements depend on preprocessing choices, such as the parcellation scheme used to define regions of interest (ROIs). In this study, we critically evaluate the effect of brain parcellations on machine learning models applied to rs-fMRI data. Our experiments reveal an intriguing trend: On average, models with stochastic parcellations consistently perform as well as models with widely used atlases at the same spatial scale. We thus propose an ensemble learning strategy to combine the predictions from models trained on connectivity data extracted using different (e.g., stochastic) parcellations. We further present an implementation of our ensemble learning strategy with a novel 3D Convolutional Neural Network (CNN) approach. The proposed CNN approach takes advantage of the full-resolution 3D spatial structure of rs-fMRI data and fits non-linear predictive models. Our ensemble CNN framework overcomes the limitations of traditional machine learning models for connectomes that often rely on region-based summary statistics and/or linear models. We showcase our approach on a classification (autism patients versus healthy controls) and a regression problem (prediction of subject's age), and report promising results.

Functional Connectivities Are More Informative Than Anatomical Variables in Diagnostic Classification of Autism

Machine learning techniques have been implemented to reveal brain features that distinguish people with autism spectrum disorders (ASDs) from typically developing (TD) peers. However, it remains unknown whether different neuroimaging modalities are equally informative for diagnostic classification. We combined anatomical magnetic resonance imaging (aMRI), diffusion weighted imaging (DWI), and functional connectivity MRI (fcMRI) using conditional random forest (CRF) for supervised learning to compare how informative each modality was in diagnostic classification. In-house data (N = 93) included 47 TD and 46 ASD participants, matched on age, motion, and nonverbal IQ. Four main analyses consistently indicated that fcMRI variables were significantly more informative than anatomical variables from aMRI and DWI. This was found (1) when the top 100 variables from CRF (run separately in each modality) were combined for multimodal CRF; (2) when only 19 top variables reaching >67% accuracy in each modality were combined in multimodal CRF; and (3) when the large number of initial variables (before dimension reduction) potentially biasing comparisons in favor of fcMRI was reduced using a less granular region of interest scheme. Consistent superiority of fcMRI was even found (4) when 100 variables per modality were randomly selected, removing any such potential bias. Greater informative value of functional than anatomical modalities may relate to the nature of fcMRI data, reflecting more closely behavioral condition, which is also the basis of diagnosis, whereas brain anatomy may be more reflective of neurodevelopmental history.

The Heterogeneity Problem: Approaches to Identify Psychiatric Subtypes

The imprecise nature of psychiatric nosology restricts progress towards characterizing and treating mental health disorders. One issue is the 'heterogeneity problem': different causal mechanisms may relate to the same disorder, and multiple outcomes of interest can occur within one individual. Our review tackles this heterogeneity problem, providing considerations, concepts, and approaches for investigators examining human cognition and mental health. We highlight the difficulty of pure dimensional approaches due to 'the curse of dimensionality'. Computationally, we consider supervised and unsupervised statistical approaches to identify putative subtypes within a population. However, we emphasize that subtype identification should be linked to a particular outcome or question. We conclude with novel hybrid approaches that can identify subtypes tied to outcomes, and may help advance precision diagnostic and treatment tools.

Characterizing and Predicting Autism Spectrum Disorder by Performing Resting-State Functional Network Community Pattern Analysis

Growing evidence indicates that autism spectrum disorder (ASD) is a neuropsychological disconnection syndrome that can be analyzed using various complex network metrics used as pathology biomarkers. Recently, community detection and analysis rooted in the complex network and graph theories have been introduced to investigate the changes in resting-state functional network community structure under neurological pathologies. However, the potential of hidden patterns in the modular organization of networks derived from resting-state functional magnetic resonance imaging to predict brain pathology has never been investigated. In this study, we present a novel analysis technique to identify alterations in community patterns in functional networks under ASD. In addition, we design machine learning classifiers to predict the clinical class of patients with ASD and controls by using only community pattern quality metrics as features. Analyses conducted on six publicly available datasets from 235 subjects, including patients with ASD and age-matched controls revealed that the modular structure is significantly disturbed in patients with ASD. Machine learning algorithms showed that the predictive power of our five metrics is relatively high (~85.16% peak accuracy for in-site data and ~75.00% peak accuracy for multisite data). These results lend further credence to the dysconnectivity theory of this pathology.

Enhancing Diagnosis of Autism With Optimized Machine Learning Models and Personal Characteristic Data

Autism spectrum disorder (ASD) is a developmental disorder, affecting about 1% of the global population. Currently, the only clinical method for diagnosing ASD are standardized ASD tests which require prolonged diagnostic time and increased medical costs. Our objective was to explore the predictive power of personal characteristic data (PCD) from a large well-characterized dataset to improve upon prior diagnostic models of ASD. We extracted six personal characteristics (age, sex, handedness, and three individual measures of IQ) from 851 subjects in the Autism Brain Imaging Data Exchange (ABIDE) database. ABIDE is an international collaborative project that collected data from a large number of ASD patients and typical non-ASD controls from 17 research and clinical institutes. We employed this publicly available database to test nine supervised machine learning models. We implemented a cross-validation strategy to train and test those machine learning models for classification between typical non-ASD controls and ASD patients. We assessed classification performance using accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Of the nine models we tested using six personal characteristics, the neural network model performed the best with a mean AUC (SD) of 0.646 (0.005), followed by k-nearest neighbor with a mean AUC (SD) of 0.641 (0.004). This study established an optimal ASD classification performance with PCD as features. With additional discriminative features (e.g., neuroimaging), machine learning models may ultimately enable automated clinical diagnosis of autism.

Machine learning identifies "rsfMRI epilepsy networks" in temporal lobe epilepsy

OBJECTIVES

Experimental models have provided compelling evidence for the existence of neural networks in temporal lobe epilepsy (TLE). To identify and validate the possible existence of resting-state "epilepsy networks," we used machine learning methods on resting-state functional magnetic resonance imaging (rsfMRI) data from 42 individuals with TLE.

METHODS

Probabilistic independent component analysis (PICA) was applied to rsfMRI data from 132 subjects (42 TLE patients + 90 healthy controls) and 88 independent components (ICs) were obtained following standard procedures. Elastic net-selected features were used as inputs to support vector machine (SVM). The strengths of the top 10 networks were correlated with clinical features to obtain "rsfMRI epilepsy networks."

RESULTS

SVM could classify individuals with epilepsy with 97.5% accuracy (sensitivity = 100%, specificity = 94.4%). Ten networks with the highest ranking were found in the frontal, perisylvian, cingulo-insular, posterior-quadrant, thalamic, cerebello-thalamic, and temporo-thalamic regions. The posterior-quadrant, cerebello-thalamic, thalamic, medial-visual, and perisylvian networks revealed significant correlation (r > 0.40) with age at onset of seizures, the frequency of seizures, duration of illness, and a number of anti-epileptic drugs.

CONCLUSIONS

IC-derived rsfMRI networks contain epilepsy-related networks and machine learning methods are useful in identifying these networks in vivo. Increased network strength with disease progression in these "rsfMRI epilepsy networks" could reflect epileptogenesis in TLE.

KEY POINTS

• ICA of resting-state fMRI carries disease-specific information about epilepsy. • Machine learning can classify these components with 97.5% accuracy. • "Subject-specific epilepsy networks" could quantify "epileptogenesis" in vivo.

Predicting Autism Spectrum Disorder Using Domain-Adaptive Cross-Site Evaluation

The advances in neuroimaging methods reveal that resting-state functional fMRI (rs-fMRI) connectivity measures can be potential diagnostic biomarkers for autism spectrum disorder (ASD). Recent data sharing projects help us replicating the robustness of these biomarkers in different acquisition conditions or preprocessing steps across larger numbers of individuals or sites. It is necessary to validate the previous results by using data from multiple sites by diminishing the site variations. We investigated partial least square regression (PLS), a domain adaptive method to adjust the effects of multicenter acquisition. A sparse Multivariate Pattern Analysis (MVVPA) framework in a leave one site out cross validation (LOSOCV) setting has been proposed to discriminate ASD from healthy controls using data from six sites in the Autism Brain Imaging Data Exchange (ABIDE). Classification features were obtained using 42 bilateral Brodmann areas without presupposing any prior hypothesis. Our results showed that using PLS, SVM showed poorer accuracies with highest accuracy achieved (62%) than without PLS but not significantly. The regions occurred in two or more informative connections are Dorsolateral Prefrontal Cortex, Somatosensory Association Cortex, Primary Auditory Cortex, Inferior Temporal Gyrus and Temporopolar area. These interrupted regions are involved in executive function, speech, visual perception, sense and language which are associated with ASD. Our findings may support early clinical diagnosis or risk determination by identifying neurobiological markers to distinguish between ASD and healthy controls.

Characterization of Autism Spectrum Disorder across the Age Span by Intrinsic Network Patterns

Autism spectrum disorder (ASD) is characterized by abnormal functional organization of brain networks, which may underlie the cognitive and social impairments observed in affected individuals. The present study characterizes unique intrinsic connectivity within- and between- neural networks in children through to adults with ASD, relative to controls. Resting state fMRI data were analyzed in 204 subjects, 102 with ASD and 102 age- and sex-matched controls (ages 7-40 years), acquired on a single scanner. ASD was assessed using the autism diagnostic observation schedule (ADOS). BOLD correlations were calculated between 47 regions of interest, spanning seven resting state brain networks. Partial least squares (PLS) analyses evaluated the association between connectivity patterns and ASD diagnosis as well as ASD severity scores. PLS demonstrated dissociable connectivity patterns in those with ASD, relative to controls. Similar patterns were observed in the whole cohort and in a subgroup analysis of subjects under 18 years of age. Greater inter-network connectivity was seen in ASD with greater intra-network connectivity in controls. In conclusion, stronger inter-network and weaker intra-network resting state-fMRI BOLD correlations characterize ASD and may differentiate control and ASD cohorts. These findings are relevant to understanding ASD as a disruption of network topology.

Topological Properties of Resting-State fMRI Functional Networks Improve Machine Learning-Based Autism Classification

Automatic algorithms for disease diagnosis are being thoroughly researched for use in clinical settings. They usually rely on pre-identified biomarkers to highlight the existence of certain problems. However, finding such biomarkers for neurodevelopmental disorders such as Autism Spectrum Disorder (ASD) has challenged researchers for many years. With enough data and computational power, machine learning (ML) algorithms can be used to interpret the data and extract the best biomarkers from thousands of candidates. In this study, we used the fMRI data of 816 individuals enrolled in the Autism Brain Imaging Data Exchange (ABIDE) to introduce a new biomarker extraction pipeline for ASD that relies on the use of graph theoretical metrics of fMRI-based functional connectivity to inform a support vector machine (SVM). Furthermore, we split the dataset into 5 age groups to account for the effect of aging on functional connectivity. Our methodology achieved better results than most state-of-the-art investigations on this dataset with the best model for the >30 years age group achieving an accuracy, sensitivity, and specificity of 95, 97, and 95%, respectively. Our results suggest that measures of centrality provide the highest contribution to the classification power of the models.

Revisiting Abnormalities in Brain Network Architecture Underlying Autism Using Topology-Inspired Statistical Inference

A large body of evidence relates autism with abnormal structural and functional brain connectivity. Structural covariance magnetic resonance imaging (scMRI) is a technique that maps brain regions with covarying gray matter densities across subjects. It provides a way to probe the anatomical structure underlying intrinsic connectivity networks (ICNs) through analysis of gray matter signal covariance. In this article, we apply topological data analysis in conjunction with scMRI to explore network-specific differences in the gray matter structure in subjects with autism versus age-, gender-, and IQ-matched controls. Specifically, we investigate topological differences in gray matter structure captured by structural correlation graphs derived from three ICNs strongly implicated in autism, namely the salience network, default mode network, and executive control network. By combining topological data analysis with statistical inference, our results provide evidence of statistically significant network-specific structural abnormalities in autism.

Combination of rs-fMRI and sMRI Data to Discriminate Autism Spectrum Disorders in Young Children Using Deep Belief Network

In recent years, the use of advanced magnetic resonance (MR) imaging methods such as functional magnetic resonance imaging (fMRI) and structural magnetic resonance imaging (sMRI) has recorded a great increase in neuropsychiatric disorders. Deep learning is a branch of machine learning that is increasingly being used for applications of medical image analysis such as computer-aided diagnosis. In a bid to classify and represent learning tasks, this study utilized one of the most powerful deep learning algorithms (deep belief network (DBN)) for the combination of data from Autism Brain Imaging Data Exchange I and II (ABIDE I and ABIDE II) datasets. The DBN was employed so as to focus on the combination of resting-state fMRI (rs-fMRI), gray matter (GM), and white matter (WM) data. This was done based on the brain regions that were defined using the automated anatomical labeling (AAL), in order to classify autism spectrum disorders (ASDs) from typical controls (TCs). Since the diagnosis of ASD is much more effective at an early age, only 185 individuals (116 ASD and 69 TC) ranging in age from 5 to 10 years were included in this analysis. In contrast, the proposed method is used to exploit the latent or abstract high-level features inside rs-fMRI and sMRI data while the old methods consider only the simple low-level features extracted from neuroimages. Moreover, combining multiple data types and increasing the depth of DBN can improve classification accuracy. In this study, the best combination comprised rs-fMRI, GM, and WM for DBN of depth 3 with 65.56% accuracy (sensitivity = 84%, specificity = 32.96%, F1 score = 74.76%) obtained via 10-fold cross-validation. This result outperforms previously presented methods on ABIDE I dataset.

Enhancing the representation of functional connectivity networks by fusing multi-view information for autism spectrum disorder diagnosis

Functional connectivity network provides novel insights on how distributed brain regions are functionally integrated, and its deviations from healthy brain have recently been employed to identify biomarkers for neuropsychiatric disorders. However, most of brain network analysis methods utilized features extracted only from one functional connectivity network for brain disease detection and cannot provide a comprehensive representation on the subtle disruptions of brain functional organization induced by neuropsychiatric disorders. Inspired by the principles of multi-view learning which utilizes information from multiple views to enhance object representation, we propose a novel multiple network based framework to enhance the representation of functional connectivity networks by fusing the common and complementary information conveyed in multiple networks. Specifically, four functional connectivity networks corresponding to the four adjacent values of regularization parameter are generated via a sparse regression model with group constraint ( l_2,1 -norm), to enhance the common intrinsic topological structure and limit the error rate caused by different views. To obtain a set of more meaningful and discriminative features, we propose using a modified version of weighted clustering coefficients to quantify the subtle differences of each group-sparse network at local level. We then linearly fuse the selected features from each individual network via a multi-kernel support vector machine for autism spectrum disorder (ASD) diagnosis. The proposed framework achieves an accuracy of 79.35%, outperforming all the compared single network methods for at least 7% improvement. Moreover, compared with other multiple network methods, our method also achieves the best performance, that is, with at least 11% improvement in accuracy.

Classification and Prediction of Brain Disorders Using Functional Connectivity: Promising but Challenging

Brain functional imaging data, especially functional magnetic resonance imaging (fMRI) data, have been employed to reflect functional integration of the brain. Alteration in brain functional connectivity (FC) is expected to provide potential biomarkers for classifying or predicting brain disorders. In this paper, we present a comprehensive review in order to provide guidance about the available brain FC measures and typical classification strategies. We survey the state-of-the-art FC analysis methods including widely used static functional connectivity (SFC) and more recently proposed dynamic functional connectivity (DFC). Temporal correlations among regions of interest (ROIs), data-driven spatial network and functional network connectivity (FNC) are often computed to reflect SFC from different angles. SFC can be extended to DFC using a sliding-window framework, and intrinsic connectivity states along the time-varying connectivity patterns are typically extracted using clustering or decomposition approaches. We also briefly summarize window-less DFC approaches. Subsequently, we highlight various strategies for feature selection including the filter, wrapper and embedded methods. In terms of model building, we include traditional classifiers as well as more recently applied deep learning methods. Moreover, we review representative applications with remarkable classification accuracy for psychosis and mood disorders, neurodevelopmental disorder, and neurological disorders using fMRI data. Schizophrenia, bipolar disorder, autism spectrum disorder (ASD), attention deficit hyperactivity disorder (ADHD), Alzheimer's disease and mild cognitive impairment (MCI) are discussed. Finally, challenges in the field are pointed out with respect to the inaccurate diagnosis labeling, the abundant number of possible features and the difficulty in validation. Some suggestions for future work are also provided.

Electrophysiological signatures of atypical intrinsic brain connectivity networks in autism

OBJECTIVE

Abnormal local and long-range brain connectivity have been widely reported in autism spectrum disorder (ASD), yet the nature of these abnormalities and their functional relevance at distinct cortical rhythms remains unknown. Investigations of intrinsic connectivity networks (ICNs) and their coherence across whole brain networks hold promise for determining whether patterns of functional connectivity abnormalities vary across frequencies and networks in ASD. In the present study, we aimed to probe atypical intrinsic brain connectivity networks in ASD from resting-state electroencephalography (EEG) data via characterizing the whole brain network.

APPROACH

Connectivity within individual ICNs (measured by spectral power) and between ICNs (measured by coherence) were examined at four canonical frequency bands via a time-frequency independent component analysis on high-density EEG, which were recorded from 20 ASD and 20 typical developing (TD) subjects during an eyes-closed resting state.

MAIN RESULTS

Among twelve identified electrophysiological ICNs, individuals with ASD showed hyper-connectivity in individual ICNs and hypo-connectivity between ICNs. Functional connectivity alterations in ASD were more severe in the frontal lobe and the default mode network (DMN) and at low frequency bands. These functional connectivity measures also showed abnormal age-related associations in ICNs related to frontal, temporal and motor regions in ASD.

SIGNIFICANCE

Our findings suggest that ASD is characterized by the opposite directions of abnormalities (i.e. hypo- and hyper-connectivity) in the hierarchical structure of the whole brain network, with more impairments in the frontal lobe and the DMN at low frequency bands, which are critical for top-down control of sensory systems, as well as for both cognition and social skills.

Alterations in resting state connectivity along the autism trait continuum: a twin study

Autism spectrum disorder (ASD) has been found to be associated with alterations in resting state (RS) functional connectivity, including areas forming the default mode network (DMN) and salience network (SN). However, insufficient control for confounding genetic and environmental influences and other methodological issues limit the generalizability of previous findings. Moreover, it has been hypothesized that ASD might be marked by early hyper-connectivity followed by later hypo-connectivity. To date, only a few studies have explicitly tested age-related influences on RS connectivity alterations in ASD. Using a within-twin pair design (N=150 twins; 8-23 years), we examined altered RS connectivity between core regions of the DMN and SN in relation to autistic trait severity and age in a sample of monozygotic (MZ) and dizygotic (DZ) twins showing typical development, ASD or other neurodevelopmental conditions. Connectivity between core regions of the SN was stronger in twins with higher autistic traits compared to their co-twins. This effect was significant both in the total sample and in MZ twins alone, highlighting the effect of non-shared environmental factors on the link between SN-connectivity and autistic traits. While this link was strongest in children, we did not identify differences between age groups for the SN. In contrast, connectivity between core hubs of the DMN was negatively correlated with autistic traits in adolescents and showed a similar trend in adults but not in children. The results support hypotheses of age-dependent altered RS connectivity in ASD, making altered SN and DMN connectivity promising candidate biomarkers for ASD.

The Diagnosis of Autism Spectrum Disorder Based on the Random Neural Network Cluster

As the autism spectrum disorder (ASD) is highly heritable, pervasive and prevalent, the clinical diagnosis of ASD is vital. In the existing literature, a single neural network (NN) is generally used to classify ASD patients from typical controls (TC) based on functional MRI data and the accuracy is not very high. Thus, the new method named as the random NN cluster, which consists of multiple NNs was proposed to classify ASD patients and TC in this article. Fifty ASD patients and 42 TC were selected from autism brain imaging data exchange (ABIDE) database. First, five different NNs were applied to build five types of random NN clusters. Second, the accuracies of the five types of random NN clusters were compared to select the highest one. The random Elman NN cluster had the highest accuracy, thus Elman NN was selected as the best base classifier. Then, we used the significant features between ASD patients and TC to find out abnormal brain regions which include the supplementary motor area, the median cingulate and paracingulate gyri, the fusiform gyrus (FG) and the insula (INS). The proposed method provides a new perspective to improve classification performance and it is meaningful for the diagnosis of ASD.

Load matters: neural correlates of verbal working memory in children with autism spectrum disorder

Background

Autism spectrum disorder (ASD) is a pervasive neurodevelopmental disorder characterised by diminished social reciprocity and communication skills and the presence of stereotyped and restricted behaviours. Executive functioning deficits, such as working memory, are associated with core ASD symptoms. Working memory allows for temporary storage and manipulation of information and relies heavily on frontal-parietal networks of the brain. There are few reports on the neural correlates of working memory in youth with ASD. The current study identified the neural systems underlying verbal working memory capacity in youth with and without ASD using functional magnetic resonance imaging (fMRI).

Methods

Fifty-seven youth, 27 with ASD and 30 sex- and age-matched typically developing (TD) controls (9–16 years), completed a one-back letter matching task (LMT) with four levels of difficulty (i.e. cognitive load) while fMRI data were recorded. Linear trend analyses were conducted to examine brain regions that were recruited as a function of increasing cognitive load.

Results

We found similar behavioural performance on the LMT in terms of reaction times, but in the two higher load conditions, the ASD youth had lower accuracy than the TD group. Neural patterns of activations differed significantly between TD and ASD groups. In TD youth, areas classically used for working memory, including the lateral and medial frontal, as well as superior parietal brain regions, increased in activation with increasing task difficulty, while areas related to the default mode network (DMN) showed decreasing activation (i.e., deactivation). The youth with ASD did not appear to use this opposing cognitive processing system; they showed little recruitment of frontal and parietal regions across the load but did show similar modulation of the DMN.

Conclusions

In a working memory task, where the load was manipulated without changing executive demands, TD youth showed increasing recruitment with increasing load of the classic fronto-parietal brain areas and decreasing involvement in default mode regions. In contrast, although they modulated the default mode network, youth with ASD did not show the modulation of increasing brain activation with increasing load, suggesting that they may be unable to manage increasing verbal information. Impaired verbal working memory in ASD would interfere with the youths’ success academically and socially. Thus, determining the nature of atypical neural processing could help establish or monitor working memory interventions for ASD.

Distributed Intrinsic Functional Connectivity Patterns Predict Diagnostic Status in Large Autism Cohort

Diagnosis of autism spectrum disorder (ASD) currently relies on behavioral observations because brain markers are unknown. Machine learning approaches can identify patterns in imaging data that predict diagnostic status, but most studies using functional connectivity MRI (fcMRI) data achieved only modest accuracies of 60-80%. We used conditional random forest (CRF), an ensemble learning technique protected against bias from feature correlation (which exists in fcMRI matrices). We selected 252 low-motion resting-state functional MRI scans from the Autism Brain Imaging Data Exchange, including 126 typically developing (TD) and 126 ASD participants, matched for age, nonverbal IQ, and head motion. A matrix of functional connectivities between 220 functionally defined regions of interest was used for diagnostic classification. In several runs, we achieved accuracies of 92-99% for classifiers with >300 features (most informative connections). Features, including pericentral somatosensory and motor regions, were disproportionately informative. Findings differed partially from a previous study in the same sample that used feature selection with random forest (which is biased by feature correlations). External validation in a smaller in-house data set, however, achieved only 67-71% accuracy. The large number of features in optimal models can be attributed to etiological heterogeneity under the clinical ASD umbrella. Lower accuracy in external validation is expected due to differences in unknown composition of ASD variants across samples. High accuracy in the main data set is unlikely due to noise overfitting, but rather indicates optimized characterization of a given cohort.

EEG Analytics for Early Detection of Autism Spectrum Disorder: A data-driven approach

Autism spectrum disorder (ASD) is a complex and heterogeneous disorder, diagnosed on the basis of behavioral symptoms during the second year of life or later. Finding scalable biomarkers for early detection is challenging because of the variability in presentation of the disorder and the need for simple measurements that could be implemented routinely during well-baby checkups. EEG is a relatively easy-to-use, low cost brain measurement tool that is being increasingly explored as a potential clinical tool for monitoring atypical brain development. EEG measurements were collected from 99 infants with an older sibling diagnosed with ASD, and 89 low risk controls, beginning at 3 months of age and continuing until 36 months of age. Nonlinear features were computed from EEG signals and used as input to statistical learning methods. Prediction of the clinical diagnostic outcome of ASD or not ASD was highly accurate when using EEG measurements from as early as 3 months of age. Specificity, sensitivity and PPV were high, exceeding 95% at some ages. Prediction of ADOS calibrated severity scores for all infants in the study using only EEG data taken as early as 3 months of age was strongly correlated with the actual measured scores. This suggests that useful digital biomarkers might be extracted from EEG measurements.

Distinguishing Adolescents With Conduct Disorder From Typically Developing Youngsters Based on Pattern Classification of Brain Structural MRI

Background: Conduct disorder (CD) is a mental disorder diagnosed in childhood or adolescence that presents antisocial behaviors, and is associated with structural alterations in brain. However, whether these structural alterations can distinguish CD from healthy controls (HCs) remains unknown. Here, we quantified these structural differences and explored the classification ability of these quantitative features based on machine learning (ML).

Materials and Methods: High-resolution 3D structural magnetic resonance imaging (sMRI) was acquired from 60 CD subjects and 60 age-matched HCs. Voxel-based morphometry (VBM) was used to assess the regional gray matter (GM) volume difference. The significantly different regional GM volumes were then extracted as features, and input into three ML classifiers: logistic regression, random forest and support vector machine (SVM). We trained and tested these ML models for classifying CD from HCs by using fivefold cross-validation (CV).

Results: Eight brain regions with abnormal GM volumes were detected, which mainly distributed in the frontal lobe, parietal lobe, anterior cingulate, cerebellum posterior lobe, lingual gyrus, and insula areas. We found that these ML models achieved comparable classification performance, with accuracy of 77.9 ∼ 80.4%, specificity of 73.3 ∼ 80.4%, sensitivity of 75.4 ∼ 87.5%, and area under the receiver operating characteristic curve (AUC) of 0.76 ∼ 0.80.

Conclusion: Based on sMRI and ML, the regional GM volumes may be used as potential imaging biomarkers for stable and accurate classification of CD.

Aberrant hemodynamic responses in autism: Implications for resting state fMRI functional connectivity studies

Functional MRI (fMRI) is modeled as a convolution of the hemodynamic response function (HRF) and an unmeasured latent neural signal. However, HRF itself is variable across brain regions and subjects. This variability is induced by both neural and non-neural factors. Aberrations in underlying neurochemical mechanisms, which control HRF shape, have been reported in autism spectrum disorders (ASD). Therefore, we hypothesized that this will lead to voxel-specific, yet systematic differences in HRF shape between ASD and healthy controls. As a corollary, we also hypothesized that such alterations will lead to differences in estimated functional connectivity in fMRI space compared to latent neural space. To test these hypotheses, we performed blind deconvolution of resting-state fMRI time series acquired from large number of ASD and control subjects obtained from the Autism Brain Imaging Data Exchange (ABIDE) database (N = 1102). Many brain regions previously implicated in autism showed systematic differences in HRF shape in ASD. Specifically, we found that precuneus had aberrations in all HRF parameters. Consequently, we obtained precuneus-seed-based functional connectivity differences between ASD and controls using fMRI as well as using latent neural signals. We found that non-deconvolved fMRI data failed to detect group differences in connectivity between precuneus and certain brain regions that were instead observed in deconvolved data. Our results are relevant for the understanding of hemodynamic and neurochemical aberrations in ASD, as well as have methodological implications for resting-state functional connectivity studies in Autism, and more generally in disorders that are accompanied by neurochemical alterations that may impact HRF shape.

The Neural Correlates of Emotional Lability in Children with Autism Spectrum Disorder

Autism spectrum disorder (ASD) is exceptionally heterogeneous in both clinical and physiopathological presentations. Clinical variability applies to ASD-specific symptoms and frequent comorbid psychopathology such as emotional lability (EL). To date, the physiopathological underpinnings of the co-occurrence of EL and ASD are unknown. As a first step, we examined within-ASD inter-individual variability of EL and its neuronal correlates using resting-state functional magnetic resonance imaging (R-fMRI). We analyzed R-fMRI data from 58 children diagnosed with ASD (5-12 years) in relation to the Conners' Parent Rating Scale EL index. We performed both an a priori amygdala region-of-interest (ROI) analysis, and a multivariate unbiased whole-brain data-driven approach. While no significant brain-behavior relationships were identified regarding amygdala intrinsic functional connectivity (iFC), multivariate whole-brain analyses revealed an extended functional circuitry centered on two regions: middle frontal gyrus (MFG) and posterior insula (PI). Follow-up parametric and nonparametric ROI-analyses of these regions revealed relationships between EL and MFG- and PI-iFC with default, salience, and visual networks suggesting that higher-order cognitive and somatosensory processes are critical for emotion regulation in ASD. We did not detect evidence of amygdala iFC underpinning EL in ASD. However, exploratory whole-brain analyses identified large-scale networks that have been previously reported abnormal in ASD. Future studies should consider EL as a potential source of neuronal heterogeneity in ASD and focus on multinetwork interactions.

Network-specific sex differentiation of intrinsic brain function in males with autism

Background

The male predominance in the prevalence of autism spectrum disorder (ASD) has motivated research on sex differentiation in ASD. Multiple sources of evidence have suggested a neurophenotypic convergence of ASD-related characteristics and typical sex differences. Two existing, albeit competing, models provide predictions on such neurophenotypic convergence. These two models are testable with neuroimaging. Specifically, the Extreme Male Brain (EMB) model predicts that ASD is associated with enhanced brain maleness in both males and females with ASD (i.e., a shift-towards-maleness). In contrast, the Gender Incoherence (GI) model predicts a shift-towards-maleness in females, yet a shift-towards-femaleness in males with ASD.

Methods

To clarify whether either model applies to the intrinsic functional properties of the brain in males with ASD, we measured the statistical overlap between typical sex differences and ASD-related atypicalities in resting-state fMRI (R-fMRI) datasets largely available in males. Main analyses focused on two large-scale R-fMRI samples: 357 neurotypical (NT) males and 471 NT females from the 1000 Functional Connectome Project and 360 males with ASD and 403 NT males from the Autism Brain Imaging Data Exchange.

Results

Across all R-fMRI metrics, results revealed coexisting, but network-specific, shift-towards-maleness and shift-towards-femaleness in males with ASD. A shift-towards-maleness mostly involved the default network, while a shift-towards-femaleness mostly occurred in the somatomotor network. Explorations of the associated cognitive processes using available cognitive ontology maps indicated that higher-order social cognitive functions corresponded to the shift-towards-maleness, while lower-order sensory motor processes corresponded to the shift-towards-femaleness.

Conclusions

The present findings suggest that atypical intrinsic brain properties in males with ASD partly reflect mechanisms involved in sexual differentiation. A model based on network-dependent atypical sex mosaicism can synthesize prior competing theories on factors involved in sex differentiation in ASD.

Electronic supplementary material

The online version of this article (10.1186/s13229-018-0192-x) contains supplementary material, which is available to authorized users.

Revisiting subcortical brain volume correlates of autism in the ABIDE dataset: effects of age and sex

BACKGROUND

Autism spectrum disorders (ASD) are characterized by substantial clinical, etiological and neurobiological heterogeneity. Despite this heterogeneity, previous imaging studies have highlighted the role of specific cortical and subcortical structures in ASD and have forwarded the notion of an ASD specific neuroanatomy in which abnormalities in brain structures are present that can be used for diagnostic classification approaches.

METHOD

A large (N = 859, 6-27 years, IQ 70-130) multi-center structural magnetic resonance imaging dataset was examined to specifically test ASD diagnostic effects regarding (sub)cortical volumes.

RESULTS

Despite the large sample size, we found virtually no main effects of ASD diagnosis. Yet, several significant two- and three-way interaction effects of diagnosis by age by gender were found.

CONCLUSION

The neuroanatomy of ASD does not exist, but is highly age and gender dependent. Implications for approaches of stratification of ASD into more homogeneous subtypes are discussed.

Classification of Autism Spectrum Disorder Using Random Support Vector Machine Cluster

Autism spectrum disorder (ASD) is mainly reflected in the communication and language barriers, difficulties in social communication, and it is a kind of neurological developmental disorder. Most researches have used the machine learning method to classify patients and normal controls, among which support vector machines (SVM) are widely employed. But the classification accuracy of SVM is usually low, due to the usage of a single SVM as classifier. Thus, we used multiple SVMs to classify ASD patients and typical controls (TC). Resting-state functional magnetic resonance imaging (fMRI) data of 46 TC and 61 ASD patients were obtained from the Autism Brain Imaging Data Exchange (ABIDE) database. Only 84 of 107 subjects are utilized in experiments because the translation or rotation of 7 TC and 16 ASD patients has surpassed ±2 mm or ±2°. Then the random SVM cluster was proposed to distinguish TC and ASD. The results show that this method has an excellent classification performance based on all the features. Furthermore, the accuracy based on the optimal feature set could reach to 96.15%. Abnormal brain regions could also be found, such as inferior frontal gyrus (IFG) (orbital and opercula part), hippocampus, and precuneus. It is indicated that the method of random SVM cluster may apply to the auxiliary diagnosis of ASD.

Using Pattern Classification to Identify Brain Imaging Markers in Autism Spectrum Disorder

Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by deficits in social interaction and communication, as well as repetitive and restrictive behaviours. The etiological and phenotypic complexity of ASD has so far hindered the development of clinically useful biomarkers for the condition. Neuroimaging studies have been valuable in establishing a biological basis for ASD. Increasingly, neuroimaging has been combined with 'machine learning'-based pattern classification methods to make individual diagnostic predictions. Moving forward, the hope is that these techniques may not only facilitate the diagnostic process but may also aid in fractionating the ASD phenotype into more biologically homogeneous sub-groups, with defined pathophysiology, predictable outcomes and/or responses to targeted treatments and/or interventions. This review chapter will first introduce 'machine learning' and pattern recognition methods in general, with a focus on their application to diagnostic classification. It will highlight why such approaches to biomarker discovery may have advantages over more conventional analytical methods. Magnetic resonance imaging (MRI) findings of atypical brain structure, function and connectivity in ASD will be briefly reviewed before we describe how pattern recognition has been applied to generate predictive models for ASD. Last, we will discuss some limitations and pitfalls of pattern recognition analyses in ASD and consider how the field can advance beyond the prediction of binary outcomes.

Subtyping cognitive profiles in Autism Spectrum Disorder using a Functional Random Forest algorithm

DSM-5 Autism Spectrum Disorder (ASD) comprises a set of neurodevelopmental disorders characterized by deficits in social communication and interaction and repetitive behaviors or restricted interests, and may both affect and be affected by multiple cognitive mechanisms. This study attempts to identify and characterize cognitive subtypes within the ASD population using our Functional Random Forest (FRF) machine learning classification model. This model trained a traditional random forest model on measures from seven tasks that reflect multiple levels of information processing. 47 ASD diagnosed and 58 typically developing (TD) children between the ages of 9 and 13 participated in this study. Our RF model was 72.7% accurate, with 80.7% specificity and 63.1% sensitivity. Using the random forest model, the FRF then measures the proximity of each subject to every other subject, generating a distance matrix between participants. This matrix is then used in a community detection algorithm to identify subgroups within the ASD and TD groups, and revealed 3 ASD and 4 TD putative subgroups with unique behavioral profiles. We then examined differences in functional brain systems between diagnostic groups and putative subgroups using resting-state functional connectivity magnetic resonance imaging (rsfcMRI). Chi-square tests revealed a significantly greater number of between group differences (p < .05) within the cingulo-opercular, visual, and default systems as well as differences in inter-system connections in the somato-motor, dorsal attention, and subcortical systems. Many of these differences were primarily driven by specific subgroups suggesting that our method could potentially parse the variation in brain mechanisms affected by ASD.

Age and Gender Effects on Intrinsic Connectivity in Autism Using Functional Integration and Segregation

BACKGROUND

The objective of this study was to examine intrinsic whole-brain functional connectivity in autism spectrum disorder (ASD) using the framework of functional segregation and integration. Emphasis was given to potential gender and developmental effects as well as identification of specific networks that may contribute to the global results.

METHODS

We leveraged an open data resource (N = 1587) of resting-state functional magnetic resonance imaging data in the Autism Brain Imaging Data Exchange (ABIDE) initiative, combining data from more than 2100 unique cross-sectional datasets in ABIDE I and ABIDE II collected at different sites. Modularity and global efficiency were utilized to assess functional segregation and integration, respectively. A meta-analytic approach for handling site differences was used. The effects of age, gender, and diagnostic category on segregation and integration were assessed using linear regression.

RESULTS

Modularity decreased nonlinearly in the ASD group with age, as evidenced by an increase and then decrease over development. Global efficiency had an opposite relationship with age by first decreasing and then increasing in the ASD group. Both modularity and global efficiency remained largely stable in the typically developing control group during development, representing a significantly different effect than seen in the ASD group. Age effects on modularity were localized to the somatosensory network. Finally, a marginally significant interaction between age, gender, and diagnostic category was found for modularity.

CONCLUSIONS

Our results support prior work that suggested a quadratic effect of age on brain development in ASD, while providing new insights about the specific characteristics of developmental and gender effects on intrinsic connectivity in ASD.

Resting-state functional under-connectivity within and between large-scale cortical networks across three low-frequency bands in adolescents with autism

Although evidence is accumulating that autism spectrum disorder (ASD) is associated with disruption of functional connections between and within brain networks, it remains largely unknown whether these abnormalities are related to specific frequency bands. To address this question, network contingency analysis was performed on brain functional connectomes obtained from 213 adolescent participants across nine sites in the Autism Brain Imaging Data Exchange (ABIDE) multisite sample, to determine the disrupted connections between and within seven major cortical networks in adolescents with ASD at Slow-5, Slow-4 and Slow-3 frequency bands and further assess whether the aberrant intra- and inter-network connectivity varied as a function of ASD symptoms. Overall under-connectivity within and between large-scale intrinsic networks in ASD was revealed across the three frequency bands. Specifically, decreased connectivity strength within the default mode network (DMN), between DMN and visual network (VN), ventral attention network (VAN), and between dorsal attention network (DAN) and VAN was observed in the lower frequency band (slow-5, slow-4), while decreased connectivity between limbic network (LN) and frontal-parietal network (FPN) was observed in the higher frequency band (slow-3). Furthermore, weaker connectivity within and between specific networks correlated with poorer communication and social interaction skills in the slow-5 band, uniquely. These results demonstrate intrinsic under-connectivity within and between multiple brain networks within predefined frequency bands in ASD, suggesting that frequency-related properties underlie abnormal brain network organization in the disorder.

Direct modulation of aberrant brain network connectivity through real-time NeuroFeedback

The existence of abnormal connectivity patterns between resting state networks in neuropsychiatric disorders, including Autism Spectrum Disorder (ASD), has been well established. Traditional treatment methods in ASD are limited, and do not address the aberrant network structure. Using real-time fMRI neurofeedback, we directly trained three brain nodes in participants with ASD, in which the aberrant connectivity has been shown to correlate with symptom severity. Desired network connectivity patterns were reinforced in real-time, without participants’ awareness of the training taking place. This training regimen produced large, significant long-term changes in correlations at the network level, and whole brain analysis revealed that the greatest changes were focused on the areas being trained. These changes were not found in the control group. Moreover, changes in ASD resting state connectivity following the training were correlated to changes in behavior, suggesting that neurofeedback can be used to directly alter complex, clinically relevant network connectivity patterns.

Even when we are at rest, our brains are always active. For example, areas of the brain involved in vision remain active in complete darkness. Different brain regions that connect together to perform a given task often show coordinated activity at rest. Past studies have shown that these resting connections are different in people with conditions such as autism. Some brain regions are more weakly connected while others are more strongly connected in people with autism spectrum disorder compared to those without. Furthermore, people with more severe symptoms seem to have more abnormal connections.

“Neurofeedback training” is a method of changing the resting connections between different brain regions. Scientists measure a brain signal – the connection between different brain regions – from a person in real time. They then provide positive feedback to the person if this signal improves. For example, if a connection that is too weak becomes stronger, the scientists might reinforce this by providing feedback on the success. Previous work has shown that neurofeedback training may even change people’s behaviour. However, it has not yet been explored as a method of treating the abnormal connections seen in people with autism when their brains are at rest.

To address this, Ramot et al. used a technique known as “functional magnetic resonance imaging” (or fMRI for short) to measure brain activity in young men with autism. First, certain brain regions were identified as having abnormal resting connections with each other. The participants were then asked to look at a blank screen and to try to reveal a picture hidden underneath. Whenever the connections between the chosen brain regions improved, part of the picture was revealed on the screen, accompanied by an upbeat sound. The participants were unaware that it was their brain signals causing this positive feedback.

This form of neurofeedback training successfully changed the abnormal brain connections in most of the participants with autism, making their connections more similar to those seen in the wider population. These effects lasted up to a year after training. Early results also suggest that these changes were related to improvements in symptoms, although further work is needed to see if doctors could reliably use this method as a therapy. These findings show that neurofeedback training could potentially help treat not only autism spectrum disorder, but a range of other disorders that involve abnormal brain connections, including depression and schizophrenia.

Identifying Autism from Resting-State fMRI Using Long Short-Term Memory Networks

Functional magnetic resonance imaging (fMRI) has helped characterize the pathophysiology of autism spectrum disorders (ASD) and carries promise for producing objective biomarkers for ASD. Recent work has focused on deriving ASD biomarkers from resting-state functional connectivity measures. However, current efforts that have identified ASD with high accuracy were limited to homogeneous, small datasets, while classification results for heterogeneous, multi-site data have shown much lower accuracy. In this paper, we propose the use of recurrent neural networks with long short-term memory (LSTMs) for classification of individuals with ASD and typical controls directly from the resting-state fMRI time-series. We used the entire large, multi-site Autism Brain Imaging Data Exchange (ABIDE) I dataset for training and testing the LSTM models. Under a cross-validation framework, we achieved classification accuracy of 68.5%, which is 9% higher than previously reported methods that used fMRI data from the whole ABIDE cohort. Finally, we presented interpretation of the trained LSTM weights, which highlight potential functional networks and regions that are known to be implicated in ASD.

Multiple functional networks modeling for autism spectrum disorder diagnosis

Despite countless studies on autism spectrum disorder (ASD), diagnosis relies on specific behavioral criteria and neuroimaging biomarkers for the disorder are still relatively scarce and irrelevant for diagnostic workup. Many researchers have focused on functional networks of brain activities using resting-state functional magnetic resonance imaging (rsfMRI) to diagnose brain diseases, including ASD. Although some existing methods are able to reveal the abnormalities in functional networks, they are either highly dependent on prior assumptions for modeling these networks or do not focus on latent functional connectivities (FCs) by considering discriminative relations among FCs in a nonlinear way. In this article, we propose a novel framework to model multiple networks of rsfMRI with data-driven approaches. Specifically, we construct large-scale functional networks with hierarchical clustering and find discriminative connectivity patterns between ASD and normal controls (NC). We then learn features and classifiers for each cluster through discriminative restricted Boltzmann machines (DRBMs). In the testing phase, each DRBM determines whether a test sample is ASD or NC, based on which we make a final decision with a majority voting strategy. We assess the diagnostic performance of the proposed method using public datasets and describe the effectiveness of our method by comparing it to competing methods. We also rigorously analyze FCs learned by DRBMs on each cluster and discover dominant FCs that play a major role in discriminating between ASD and NC. Hum Brain Mapp 38:5804-5821, 2017. © 2017 Wiley Periodicals, Inc.

Diagnosing Autism Spectrum Disorder from Brain Resting-State Functional Connectivity Patterns Using a Deep Neural Network with a Novel Feature Selection Method

The whole-brain functional connectivity (FC) pattern obtained from resting-state functional magnetic resonance imaging data are commonly applied to study neuropsychiatric conditions such as autism spectrum disorder (ASD) by using different machine learning models. Recent studies indicate that both hyper- and hypo- aberrant ASD-associated FCs were widely distributed throughout the entire brain rather than only in some specific brain regions. Deep neural networks (DNN) with multiple hidden layers have shown the ability to systematically extract lower-to-higher level information from high dimensional data across a series of neural hidden layers, significantly improving classification accuracy for such data. In this study, a DNN with a novel feature selection method (DNN-FS) is developed for the high dimensional whole-brain resting-state FC pattern classification of ASD patients vs. typical development (TD) controls. The feature selection method is able to help the DNN generate low dimensional high-quality representations of the whole-brain FC patterns by selecting features with high discriminating power from multiple trained sparse auto-encoders. For the comparison, a DNN without the feature selection method (DNN-woFS) is developed, and both of them are tested with different architectures (i.e., with different numbers of hidden layers/nodes). Results show that the best classification accuracy of 86.36% is generated by the DNN-FS approach with 3 hidden layers and 150 hidden nodes (3/150). Remarkably, DNN-FS outperforms DNN-woFS for all architectures studied. The most significant accuracy improvement was 9.09% with the 3/150 architecture. The method also outperforms other feature selection methods, e.g., two sample t-test and elastic net. In addition to improving the classification accuracy, a Fisher's score-based biomarker identification method based on the DNN is also developed, and used to identify 32 FCs related to ASD. These FCs come from or cross different pre-defined brain networks including the default-mode, cingulo-opercular, frontal-parietal, and cerebellum. Thirteen of them are statically significant between ASD and TD groups (two sample t-test p < 0.05) while 19 of them are not. The relationship between the statically significant FCs and the corresponding ASD behavior symptoms is discussed based on the literature and clinician's expert knowledge. Meanwhile, the potential reason of obtaining 19 FCs which are not statistically significant is also provided.

Progress and roadblocks in the search for brain-based biomarkers of autism and attention-deficit/hyperactivity disorder

Children with neurodevelopmental disorders benefit most from early interventions and treatments. The development and validation of brain-based biomarkers to aid in objective diagnosis can facilitate this important clinical aim. The objective of this review is to provide an overview of current progress in the use of neuroimaging to identify brain-based biomarkers for autism spectrum disorder (ASD) and attention-deficit/hyperactivity disorder (ADHD), two prevalent neurodevelopmental disorders. We summarize empirical work that has laid the foundation for using neuroimaging to objectively quantify brain structure and function in ways that are beginning to be used in biomarker development, noting limitations of the data currently available. The most successful machine learning methods that have been developed and applied to date are discussed. Overall, there is increasing evidence that specific features (for example, functional connectivity, gray matter volume) of brain regions comprising the salience and default mode networks can be used to discriminate ASD from typical development. Brain regions contributing to successful discrimination of ADHD from typical development appear to be more widespread, however there is initial evidence that features derived from frontal and cerebellar regions are most informative for classification. The identification of brain-based biomarkers for ASD and ADHD could potentially assist in objective diagnosis, monitoring of treatment response and prediction of outcomes for children with these neurodevelopmental disorders. At present, however, the field has yet to identify reliable and reproducible biomarkers for these disorders, and must address issues related to clinical heterogeneity, methodological standardization and cross-site validation before further progress can be achieved.

Computational neuroscience approach to biomarkers and treatments for mental disorders

Psychiatry research has long experienced a stagnation stemming from a lack of understanding of the neurobiological underpinnings of phenomenologically defined mental disorders. Recently, the application of computational neuroscience to psychiatry research has shown great promise in establishing a link between phenomenological and pathophysiological aspects of mental disorders, thereby recasting current nosology in more biologically meaningful dimensions. In this review, we highlight recent investigations into computational neuroscience that have undertaken either theory- or data-driven approaches to quantitatively delineate the mechanisms of mental disorders. The theory-driven approach, including reinforcement learning models, plays an integrative role in this process by enabling correspondence between behavior and disorder-specific alterations at multiple levels of brain organization, ranging from molecules to cells to circuits. Previous studies have explicated a plethora of defining symptoms of mental disorders, including anhedonia, inattention, and poor executive function. The data-driven approach, on the other hand, is an emerging field in computational neuroscience seeking to identify disorder-specific features among high-dimensional big data. Remarkably, various machine-learning techniques have been applied to neuroimaging data, and the extracted disorder-specific features have been used for automatic case-control classification. For many disorders, the reported accuracies have reached 90% or more. However, we note that rigorous tests on independent cohorts are critically required to translate this research into clinical applications. Finally, we discuss the utility of the disorder-specific features found by the data-driven approach to psychiatric therapies, including neurofeedback. Such developments will allow simultaneous diagnosis and treatment of mental disorders using neuroimaging, thereby establishing 'theranostics' for the first time in clinical psychiatry.