Aberrant brain dynamics and spectral power in children with ADHD and its subtypes.

Attention-deficit/hyperactivity disorder (ADHD) is a prevalent neurodevelopmental disorder in children, usually categorized as three subtypes, predominant inattention (ADHD-I), predominant hyperactivity-impulsivity (ADHD-HI), and a combined subtype (ADHD-C). Yet, common and unique abnormalities of electroencephalogram (EEG) across different subtypes remain poorly understood. Here, we leveraged microstate characteristics and power features to investigate temporal and frequency abnormalities in ADHD and its subtypes using high-density EEG on 161 participants (54 ADHD-Is and 53 ADHD-Cs and 54 healthy controls). Four EEG microstates were identified. The coverage of salience network (state C) were decreased in ADHD compared to HC (p = 1.46e-3), while the duration and contribution of frontal-parietal network (state D) were increased (p = 1.57e-3; p = 1.26e-4). Frequency power analysis also indicated that higher delta power in the fronto-central area (p = 6.75e-4) and higher power of theta/beta ratio in the bilateral fronto-temporal area (p = 3.05e-3) were observed in ADHD. By contrast, remarkable subtype differences were found primarily on the visual network (state B), of which ADHD-C have higher occurrence and coverage than ADHD-I (p = 9.35e-5; p = 1.51e-8), suggesting that children with ADHD-C might exhibit impulsivity of opening their eyes in an eye-closed experiment, leading to hyper-activated visual network. Moreover, the top discriminative features selected from support vector machine model with recursive feature elimination (SVM-RFE) well replicated the above results, which achieved an accuracy of 72.7% and 73.8% separately in classifying ADHD and two subtypes. To conclude, this study highlights EEG microstate dynamics and frequency features may serve as sensitive measurements to detect the subtle differences in ADHD and its subtypes, providing a new window for better diagnosis of ADHD.

DystoniaBoTXNet: Novel Neural Network Biomarker of Botulinum Toxin Efficacy in Isolated Dystonia.

OBJECTIVE: Isolated dystonia is characterized by abnormal, often painful, postures and repetitive movements due to sustained or intermittent involuntary muscle contractions. Botulinum toxin (BoTX) injections into the affected muscles are the first line of therapy. However, there are no objective predictive markers or standardized tests of BoTX efficacy that can be utilized for appropriate candidate selection prior to treatment initiation. METHODS: We developed a deep learning algorithm, DystoniaBoTXNet, which uses a 3D convolutional neural network architecture and raw structural brain magnetic resonance images (MRIs) to automatically discover and test a neural network biomarker of BoTX efficacy in 284 patients with 4 different forms of focal dystonia, including laryngeal dystonia, blepharospasm, cervical dystonia, and writer's cramp. RESULTS: DystoniaBoTXNet identified clusters in superior parietal lobule, inferior and middle frontal gyri, middle orbital gyrus, inferior temporal gyrus, corpus callosum, inferior fronto-occipital fasciculus, and anterior thalamic radiation as components of the treatment biomarker. These regions are known to contribute to both dystonia pathophysiology across a broad clinical spectrum of disorder and the central effects of botulinum toxin treatment. Based on its biomarker, DystoniaBoTXNet achieved an overall accuracy of 96.3%, with 100% sensitivity and 86.1% specificity, in predicting BoTX efficacy in patients with isolated dystonia. The algorithmic decision was computed in 19.2 seconds per case. INTERPRETATION: DystoniaBoTXNet and its treatment biomarker have a high translational potential as an objective, accurate, generalizable, fast, and cost-effective algorithmic platform for enhancing clinical decision making for BoTX treatment in patients with isolated dystonia. ANN NEUROL 2023;93:460-471.

Multi-site clustering and nested feature extraction for identifying autism spectrum disorder with resting-state fMRI.

Brain functional connectivity (FC) derived from resting-state functional magnetic resonance imaging (rs-fMRI) has been widely employed to study neuropsychiatric disorders such as autism spectrum disorder (ASD). Existing studies usually suffer from (1) significant data heterogeneity caused by different scanners or studied populations in multiple sites, (2) curse of dimensionality caused by millions of voxels in each fMRI scan and a very limited number (tens or hundreds) of training samples, and (3) poor interpretability, which hinders the identification of reproducible disease biomarkers. To this end, we propose a Multi-site Clustering and Nested Feature Extraction (MC-NFE) method for fMRI-based ASD detection. Specifically, we first divide multi-site training data into ASD and healthy control (HC) groups. To model inter-site heterogeneity within each category, we use a similarity-driven multiview linear reconstruction model to learn latent representations and perform subject clustering within each group. We then design a nested singular value decomposition (SVD) method to mitigate inter-site heterogeneity and extract FC features by learning both local cluster-shared features across sites within each category and global category-shared features across ASD and HC groups, followed by a linear support vector machine (SVM) for ASD detection. Experimental results on 609 subjects with rs-fMRI from the ABIDE database with 21 imaging sites suggest that the proposed MC-NFE outperforms several state-of-the-art methods in ASD detection. The most discriminative FCs identified by the MC-NFE are mainly located in default mode network, salience network, and cerebellum region, which could be used as potential biomarkers for fMRI-based ASD analysis.

Investigating ADHD subtypes in children using temporal dynamics of the electroencephalogram (EEG) microstates.

Attention-deficit/hyperactivity disorder (ADHD) is a prevalent neurodevelopmental disorder in children, usually categorized as three predominant subtypes, persistent inattention (ADHD-I), hyperactivity-impulsivity (ADHD-HI) and a combination of both (ADHD-C). Identifying reliable features to distinguish different subtypes is significant for clinical individualized treatment. In this work, we conducted a two-stage electroencephalogram (EEG) microstate analysis on 54 healthy controls and 107 ADHD children, including 54 ADHD-Is and 53 ADHD-Cs, aiming to examine the dynamic temporal alterations in ADHDs compared to healthy controls (HCs), as well as different EEG signatures between ADHD subtypes. Results demonstrated that the dynamics of resting-state EEG microstates, particularly centering on salience (state C) and frontal-parietal network (state D), were significantly aberrant in ADHDs. Specifically, the occurrence and coverage of state C were decreased in ADHDs (p=0.002; p=0.0015), while the duration and contribution of state D were observably increased (p=0.0016; p=0.0001) compared to HCs. Moreover, the transition probability between state A and C was significantly decreased (p=9.85e-7; p=2.33e-7) in ADHDs, but otherwise increased between state B and D (p=1.02e-7; p=1.07e-6). By contrast, remarkable subtype differences were found primarily on the visual network (state B) between ADHD-Is and ADHD-Cs. Specifically, ADHD-Cs have higher occurrence and coverage of state B than ADHD-Is (p=9.35e-5; p=1.51e-8), suggesting these patients more impulsively aimed to open their eyes when asked to keep eyes closed during the data collection. In summary, this work carefully leveraged EEG temporal dynamics to investigate the aberrant microstate features in ADHDs and provided a new window to look into the subtle differences between ADHD subtypes, which may help to assist precision diagnosis in future.Clinical Relevance- This work established the use of EEG microstate features to investigate ADHD dysfunction and its subtypes, providing a new window for better diagnosis of ADHD.

BNCPL: Brain-Network-based Convolutional Prototype Learning for Discriminating Depressive Disorders.

Deep learning has shown great potential to adaptively learn hidden patterns from high dimensional neuroimaging data, so as to extract subtle group differences. Motivated by the convolutional neural networks and prototype learning, we developed a brain-network-based convolutional prototype learning model (BNCPL), which can learn representations that simultaneously maximize inter-class separation while minimize within-class distance. When applying BNCPL to distinguish 208 depressive disorders from 210 healthy controls using resting-state functional connectivity (FC), we achieved an accuracy of 71.0% in multi-site pooling classification (3 sites), with 2.4-7.2% accuracy increase compared to 3 traditional classifiers and 2 alternative deep neural networks. Saliency map was also used to examine the most discriminative FCs learned by the model; the prefrontal-subcortical circuits were identified, which were also correlated with disease severity and cognitive ability. In summary, by integrating convolutional prototype learning and saliency map, we improved both the model interpretability and classification performance, and found that the dysregulation of the functional prefrontal-subcortical circuit may play a pivotal role in discriminating depressive disorders from healthy controls.

A Mutual Multi-Scale Triplet Graph Convolutional Network for Classification of Brain Disorders Using Functional or Structural Connectivity.

Brain connectivity alterations associated with mental disorders have been widely reported in both functional MRI (fMRI) and diffusion MRI (dMRI). However, extracting useful information from the vast amount of information afforded by brain networks remains a great challenge. Capturing network topology, graph convolutional networks (GCNs) have demonstrated to be superior in learning network representations tailored for identifying specific brain disorders. Existing graph construction techniques generally rely on a specific brain parcellation to define regions-of-interest (ROIs) to construct networks, often limiting the analysis into a single spatial scale. In addition, most methods focus on the pairwise relationships between the ROIs and ignore high-order associations between subjects. In this letter, we propose a mutual multi-scale triplet graph convolutional network (MMTGCN) to analyze functional and structural connectivity for brain disorder diagnosis. We first employ several templates with different scales of ROI parcellation to construct coarse-to-fine brain connectivity networks for each subject. Then, a triplet GCN (TGCN) module is developed to learn functional/structural representations of brain connectivity networks at each scale, with the triplet relationship among subjects explicitly incorporated into the learning process. Finally, we propose a template mutual learning strategy to train different scale TGCNs collaboratively for disease classification. Experimental results on 1,160 subjects from three datasets with fMRI or dMRI data demonstrate that our MMTGCN outperforms several state-of-the-art methods in identifying three types of brain disorders.

Deep Bayesian Hashing With Center Prior for Multi-Modal Neuroimage Retrieval.

Multi-modal neuroimage retrieval has greatly facilitated the efficiency and accuracy of decision making in clinical practice by providing physicians with previous cases (with visually similar neuroimages) and corresponding treatment records. However, existing methods for image retrieval usually fail when applied directly to multi-modal neuroimage databases, since neuroimages generally have smaller inter-class variation and larger inter-modal discrepancy compared to natural images. To this end, we propose a deep Bayesian hash learning framework, called CenterHash, which can map multi-modal data into a shared Hamming space and learn discriminative hash codes from imbalanced multi-modal neuroimages. The key idea to tackle the small inter-class variation and large inter-modal discrepancy is to learn a common center representation for similar neuroimages from different modalities and encourage hash codes to be explicitly close to their corresponding center representations. Specifically, we measure the similarity between hash codes and their corresponding center representations and treat it as a center prior in the proposed Bayesian learning framework. A weighted contrastive likelihood loss function is also developed to facilitate hash learning from imbalanced neuroimage pairs. Comprehensive empirical evidence shows that our method can generate effective hash codes and yield state-of-the-art performance in cross-modal retrieval on three multi-modal neuroimage datasets.

Shared and distinct resting functional connectivity in children and adults with attention-deficit/hyperactivity disorder.

Attention-deficit/hyperactivity disorder (ADHD) often persists into adulthood, with a shift of symptoms including less hyperactivity/impulsivity and more co-morbidity of affective disorders in ADHDadult. Many studies have questioned the stability in diagnosing of ADHD from childhood to adulthood, and the shared and distinct aberrant functional connectivities (FCs) between ADHDchild and ADHDadult remain unidentified. We aim to explore shared and distinct FC patterns in ADHDchild and ADHDadult, and further investigated the cross-cohort predictability using the identified FCs. After investigating the ADHD-discriminative FCs from healthy controls (HCs) in both child (34 ADHDchild, 28 HCs) and adult (112 ADHDadult,77 HCs) cohorts, we identified both shared and distinct aberrant FC patterns between cohorts and their association with clinical symptoms. Moreover, the cross-cohort predictability using the identified FCs were tested. The ADHD-HC classification accuracies were 84.4% and 81.0% for children and male adults, respectively. The ADHD-discriminative FCs shared in children and adults lie in the intra-network within default mode network (DMN) and the inter-network between DMN and ventral attention network, positively correlated with total scores of ADHD symptoms. Particularly, inter-network FC between somatomotor network and dorsal attention network was uniquely impaired in ADHDchild, positively correlated with hyperactivity index; whereas the aberrant inter-network FC between DMN and limbic network exhibited more adult-specific ADHD dysfunction. And their cross-cohort predictions were 70.4% and 75.6% between each other. This work provided imaging evidence for symptomatic changes and pathophysiological continuity in ADHD from childhood to adulthood, suggesting that FCs may serve as potential biomarkers for ADHD diagnosis.

Spatial-Temporal Dependency Modeling and Network Hub Detection for Functional MRI Analysis via Convolutional-Recurrent Network.

Early identification of dementia at the stage of mild cognitive impairment (MCI) is crucial for timely diagnosis and intervention of Alzheimer's disease (AD). Although several pioneering studies have been devoted to automated AD diagnosis based on resting-state functional magnetic resonance imaging (rs-fMRI), their performance is somewhat limited due to non-effective mining of spatial-temporal dependency. Besides, few of these existing approaches consider the explicit detection and modeling of discriminative brain regions (i.e., network hubs) that are sensitive to AD progression. In this paper, we propose a unique Spatial-Temporal convolutional-recurrent neural Network (STNet) for automated prediction of AD progression and network hub detection from rs-fMRI time series. Our STNet incorporates the spatial-temporal information mining and AD-related hub detection into an end-to-end deep learning model. Specifically, we first partition rs-fMRI time series into a sequence of overlapping sliding windows. A sequence of convolutional components are then designed to capture the local-to-global spatially-dependent patterns within each sliding window, based on which we are able to identify discriminative hubs and characterize their unique contributions to disease diagnosis. A recurrent component with long short-term memory (LSTM) units is further employed to model the whole-brain temporal dependency from the spatially-dependent pattern sequences, thus capturing the temporal dynamics along time. We evaluate the proposed method on 174 subjects with 563 rs-fMRI scans from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, with results suggesting the effectiveness of our method in both tasks of disease progression prediction and AD-related hub detection.

Deep Disentangled Hashing with Momentum Triplets for Neuroimage Search.

Neuroimaging has been widely used in computer-aided clinical diagnosis and treatment, and the rapid increase of neuroimage repositories introduces great challenges for efficient neuroimage search. Existing image search methods often use triplet loss to capture high-order relationships between samples. However, we find that the traditional triplet loss is difficult to pull positive and negative sample pairs to make their Hamming distance discrepancies larger than a small fixed value. This may reduce the discriminative ability of learned hash code and degrade the performance of image search. To address this issue, in this work, we propose a deep disentangled momentum hashing (DDMH) framework for neuroimage search. Specifically, we first investigate the original triplet loss and find that this loss function can be determined by the inner product of hash code pairs. Accordingly, we disentangle hash code norms and hash code directions and analyze the role of each part. By decoupling the loss function from the hash code norm, we propose a unique disentangled triplet loss, which can effectively push positive and negative sample pairs by desired Hamming distance discrepancies for hash codes with different lengths. We further develop a momentum triplet strategy to address the problem of insufficient triplet samples caused by small batch-size for 3D neuroimages. With the proposed disentangled triplet loss and the momentum triplet strategy, we design an end-to-end trainable deep hashing framework for neuroimage search. Comprehensive empirical evidence on three neuroimage datasets shows that DDMH has better performance in neuroimage search compared to several state-of-the-art methods.

Temporal-Adaptive Graph Convolutional Network for Automated Identification of Major Depressive Disorder Using Resting-State fMRI.

Extensive studies focus on analyzing human brain functional connectivity from a network perspective, in which each network contains complex graph structures. Based on resting-state functional MRI (rs-fMRI) data, graph convolutional networks (GCNs) enable comprehensive mapping of brain functional connectivity (FC) patterns to depict brain activities. However, existing studies usually characterize static properties of the FC patterns, ignoring the time-varying dynamic information. In addition, previous GCN methods generally use fixed group-level (e.g., patients or controls) representation of FC networks, and thus, cannot capture subject-level FC specificity. To this end, we propose a Temporal-Adaptive GCN (TAGCN) framework that can not only take advantage of both spatial and temporal information using resting-state FC patterns and time-series but also explicitly characterize subject-level specificity of FC patterns. Specifically, we first segment each ROI-based time-series into multiple overlapping windows, then employ an adaptive GCN to mine topological information. We further model the temporal patterns for each ROI along time to learn the periodic brain status changes. Experimental results on 533 major depressive disorder (MDD) and health control (HC) subjects demonstrate that the proposed TAGCN outperforms several state-of-the-art methods in MDD vs. HC classification, and also can be used to capture dynamic FC alterations and learn valid graph representations.

Discriminating ADHD From Healthy Controls Using a Novel Feature Selection Method Based on Relative Importance and Ensemble Learning.

Attention-deficit/hyperactivity disorder (ADHD) is a childhood-onset neurodevelopmental disorder that often persists into adulthood, resulting in adverse effects on work performance and social function. The current diagnosis of ADHD primarily depends on the judgment of clinical symptoms, which highlights the need for objective imaging biomarkers. In this study, we aim to classify ADHD (both children and adults [34/112]) from age-matched healthy controls (HCs [28/77]) with functional connectivity (FCs) pattern derived from resting-state functional magnetic resonance imaging (rs-fMRI) data. However, the neuroimaging classification of brain disorders often meets a situation of high dimensional features were presented with limited sample size. Thus an efficient method that is able to reduce original feature dimension into a much more refined subspace is highly desired. Here we proposed a novel Feature Selection method based on Relative Importance and Ensemble Learning (FS_RIEL). Compared with traditional feature selection methods, FS_RIEL algorithm improved the ADHD classification by about 15% in both child and adult ADHD classification, achieving 80-86% accuracy. Moreover, we found the most frequently selected FCs were mainly involved in frontoparietal network, default network, salience network, basal ganglia network and cerebellum network in both child and adult ADHD cohorts, which indicates that ADHD is characterized by a widely-impaired brain connectivity profile that may serve as potential biomarkers for its early diagnosis.

An ensemble learning system for a 4-way classification of Alzheimer's disease and mild cognitive impairment.

Discriminating Alzheimer's disease (AD) from its prodromal form, mild cognitive impairment (MCI), is a significant clinical problem that may facilitate early diagnosis and intervention, in which a more challenging issue is to classify MCI subtypes, i.e., those who eventually convert to AD (cMCI) versus those who do not (MCI). To solve this difficult 4-way classification problem (AD, MCI, cMCI and healthy controls), a competition was hosted by Kaggle to invite the scientific community to apply their machine learning approaches on pre-processed sets of T1-weighted magnetic resonance images (MRI) data and the demographic information from the international Alzheimer's disease neuroimaging initiative (ADNI) database. This paper summarizes our competition results. We first proposed a hierarchical process by turning the 4-way classification into five binary classification problems. A new feature selection technology based on relative importance was also proposed, aiming to identify a more informative and concise subset from 426 sMRI morphometric and 3 demographic features, to ensure each binary classifier to achieve its highest accuracy. As a result, about 2% of the original features were selected to build a new feature space, which can achieve the final four-way classification with a 54.38% accuracy on testing data through hierarchical grouping, higher than several alternative methods in comparison. More importantly, the selected discriminative features such as hippocampal volume, parahippocampal surface area, and medial orbitofrontal thickness, etc. as well as the MMSE score, are reasonable and consistent with those reported in AD/MCI deficits. In summary, the proposed method provides a new framework for multi-way classification using hierarchical grouping and precise feature selection.