Machine learning methods for brain network classification: Application to autism diagnosis using cortical morphological networks

The diagnosis of ASD with MRI: a systematic review and meta-analysis

While diagnosing autism spectrum disorder (ASD) based on an objective test is desired, the current diagnostic practice involves observation-based criteria. This study is a systematic review and meta-analysis of studies that aim to diagnose ASD using magnetic resonance imaging (MRI). The main objective is to describe the state of the art of diagnosing ASD using MRI in terms of performance metrics and interpretation. Furthermore, subgroups, including different MRI modalities and statistical heterogeneity, are analyzed. Studies that dichotomously diagnose individuals with ASD and healthy controls by analyses progressing from magnetic resonance imaging obtained in a resting state were systematically selected by two independent reviewers. Studies were sought on Web of Science and PubMed, which were last accessed on February 24, 2023. The included studies were assessed on quality and risk of bias using the revised Quality Assessment of Diagnostic Accuracy Studies tool. A bivariate random-effects model was used for syntheses. One hundred and thirty-four studies were included comprising 159 eligible experiments. Despite the overlap in the studied samples, an estimated 4982 unique participants consisting of 2439 individuals with ASD and 2543 healthy controls were included. The pooled summary estimates of diagnostic performance are 76.0% sensitivity (95% CI 74.1-77.8), 75.7% specificity (95% CI 74.0-77.4), and an area under curve of 0.823, but uncertainty in the study assessments limits confidence. The main limitations are heterogeneity and uncertainty about the generalization of diagnostic performance. Therefore, comparisons between subgroups were considered inappropriate. Despite the current limitations, methods progressing from MRI approach the diagnostic performance needed for clinical practice. The state of the art has obstacles but shows potential for future clinical application.

Autism spectrum disorder diagnosis with EEG signals using time series maps of brain functional connectivity and a combined CNN-LSTM model

BACKGROUND AND OBJECTIVE

People with autism spectrum disorder (ASD) often have cognitive impairments. Effective connectivity between different areas of the brain is essential for normal cognition. Electroencephalography (EEG) has been widely used in the detection of neurological diseases. Previous studies on detecting ASD with EEG data have focused on frequency-related features. Most of these studies have augmented data by splitting the dataset into time slices or sliding windows. However, such approaches to data augmentation may cause the testing data to be contaminated by the training data. To solve this problem, this study developed a novel method for detecting ASD with EEG data.

METHODS

This study quantified the functional connectivity of the subject's brain from EEG signals and defined the individual to be the unit of analysis. Publicly available EEG data were gathered from 97 and 92 subjects with ASD and typical development (TD), respectively, while they were at rest or performing a task. Time-series maps of brain functional connectivity were constructed, and the data were augmented using a deep convolutional generative adversarial network. In addition, a combined network for ASD detection, based on convolutional neural network (CNN) and long short-term memory (LSTM), was designed and implemented.

RESULTS

Based on functional connectivity, the network achieved classification accuracies of 81.08% and 74.55% on resting state and task state data, respectively. In addition, we found that the functional connectivity of ASD differed from TD primarily in the short-distance functional connectivity of the parietal and occipital lobes and in the distant connections from the right temporoparietal junction region to the left posterior temporal lobe.

CONCLUSIONS

This paper provides a new perspective for better utilizing EEG to understand ASD. The method proposed in our study is expected to be a reliable tool to assist in the diagnosis of ASD.

Brain network classification based on dynamic graph attention information bottleneck

BACKGROUND AND OBJECTIVES

Graph neural networks (GNN) have demonstrated remarkable encoding capabilities in the context of brain network classification tasks. They excel at uncovering hidden static connections between brain states. However, brain network signals can be influenced by physiological traits and external variables during clinical detection, resulting in noisy brain graphs. Additionally, many existing algorithms for brain networks primarily focus on static topologies determined by threshold-based criteria, thereby overlooking the real-time variability in brain channel connectivity. These sources of noise and the persistence of static structures inevitably hinder the effective exchange of information during brain network computations.

METHODS

To address these challenges, we propose a novel framework called the dynamic graph attention information bottleneck (DGAIB). This framework is designed to dynamically enhance the input raw brain graph structure from the perspective of information theory and graph theory. First, we employ the Spearman function to construct a raw graph. Then, we use a graph information bottleneck (GIB) to optimize the internal graph connections by selectively masking redundant feature embeddings. Finally, we enhance the feature aggregation of each brain state by utilizing a graph attention network (GAT), which promotes improved information exchange among distinct brain regions within the model. These processed representations serve as input for subsequent classification tasks.

EXPERIMENT AND RESULTS

We systematically evaluated the robustness and generalizability of our proposed framework through a series of experiments. This evaluation included patient-specific experiments using the electroencephalography (EEG)-based CHB-MIT dataset and cross-patient experiments leveraging the functional magnetic resonance imaging (fMRI)-based ABIDE-I dataset from multiple perspectives.

Systematic bibliometric and visualized analysis of research hotspots and trends in artificial intelligence in autism spectrum disorder

Background

Artificial intelligence (AI) has been the subject of studies in autism spectrum disorder (ASD) and may affect its identification, diagnosis, intervention, and other medical practices in the future. Although previous studies have used bibliometric techniques to analyze and investigate AI, there has been little research on the adoption of AI in ASD. This study aimed to explore the broad applications and research frontiers of AI used in ASD.

Methods

Citation data were retrieved from the Web of Science Core Collection (WoSCC) database to assess the extent to which AI is used in ASD. CiteSpace.5.8. R3 and VOSviewer, two online tools for literature metrology analysis, were used to analyze the data.

Results

A total of 776 publications from 291 countries and regions were analyzed; of these, 256 publications were from the United States and 173 publications were from China, and England had the largest centrality of 0.33; Stanford University had the highest H-index of 17; and the largest cluster label of co-cited references was machine learning. In addition, keywords with a high number of occurrences in this field were autism spectrum disorder (295), children (255), classification (156) and diagnosis (77). The burst keywords from 2021 to 2023 were infants and feature selection, and from 2022 to 2023, the burst keyword was corpus callosum.

Conclusion

This research provides a systematic analysis of the literature concerning AI used in ASD, presenting an overall demonstration in this field. In this area, the United States and China have the largest number of publications, England has the greatest influence, and Stanford University is the most influential. In addition, the research on AI used in ASD mostly focuses on classification and diagnosis, and "infants, feature selection, and corpus callosum are at the forefront, providing directions for future research. However, the use of AI technologies to identify ASD will require further research.

Multi-view graph network learning framework for identification of major depressive disorder

Functional connectivity (FC) derived from resting-state functional magnetic resonance imaging (rs-fMRI) exhibits non-Euclidean topological structures, which have pathological foundations and serve as ideal objective data for intelligent diagnosis of major depressive disorder (MDD) patients. Additionally, the fully connected FC demonstrates uniform spatial structures. To learn and integrate information from these two structural forms for a more comprehensive identification of MDD patients, we propose a novel hierarchical learning structure called Multi-View Graph Neural Network (MV-GNN). In MV-GNN, the collaborative FC of subjects is filtered and reconstructed from topological view to obtain the reconstructed FC, incorporating various threshold values to calculate the topological attributes of brain regions. ROC analysis is performed on the average scores of these attributes for MDD and healthy control (HC) groups to determine an efficient threshold. Group differences analysis is conducted on the efficient topological attributes of brain regions, followed by their selection. These efficient attributes, along with the reconstructed FC, are combined to construct a graph view using self-attention graph pooling and graph convolutional neural networks, enabling efficient embedding. To extract efficient FC pattern difference information from spatial view, a dual leave-one-out cross-feature selection method is proposed. It selects and extracts relevant information from uniformly sized FC structures' high-dimensional spatial features, constructing a relationship view between brain regions. This approach incorporates both the whole graph topological view and spatial relationship view in a multi-layered structure, fusing them using gating mechanisms. By incorporating multiple views, it enhances the inference of whether subjects suffer from MDD and reveals differential information between MDD and HC groups across different perspectives. The proposed model structure is evaluated through leave-one-site cross-validation and achieves an average accuracy of 65.61% in identifying MDD patients at a single-center site, surpassing state-of-the-art methods in MDD recognition. The model provides valuable discriminatory information for objective diagnosis of MDD and serves as a reference for pathological foundations.

Machine learning approaches to identify Parkinson's disease using voice signal features

Parkinson's Disease (PD) is the second most common age-related neurological disorder that leads to a range of motor and cognitive symptoms. A PD diagnosis is difficult since its symptoms are quite similar to those of other disorders, such as normal aging and essential tremor. When people reach 50, visible symptoms such as difficulties walking and communicating begin to emerge. Even though there is no cure for PD, certain medications can relieve some of the symptoms. Patients can maintain their lifestyles by controlling the complications caused by the disease. At this point, it is essential to detect this disease and prevent it from progressing. The diagnosis of the disease has been the subject of much research. In our project, we aim to detect PD using different types of Machine Learning (ML), and Deep Learning (DL) models such as Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), K-Nearest Neighbor (KNN), and Multi-Layer Perceptron (MLP) to differentiate between healthy and PD patients by voice signal features. The dataset taken from the University of California at Irvine (UCI) machine learning repository consisted of 195 voice recordings of examinations carried out on 31 patients. Moreover, our models were trained using different techniques such as Synthetic Minority Over-sampling Technique (SMOTE), Feature Selection, and hyperparameter tuning (GridSearchCV) to enhance their performance. At the end, we found that MLP and SVM with a ratio of 70:30 train/test split using GridSearchCV with SMOTE gave the best results for our project. MLP performed with an overall accuracy of 98.31%, an overall recall of 98%, an overall precision of 100%, and f1-score of 99%. In addition, SVM performed with an overall accuracy of 95%, an overall recall of 96%, an overall precision of 98%, and f1-score of 97%. The experimental results of this research imply that the proposed method can be used to reliably predict PD and can be easily incorporated into healthcare for diagnosis purposes.

Template-based graph registration network for boosting the diagnosis of brain connectivity disorders

Brain graphs are powerful representations to explore the biological roadmaps of the human brain in its healthy and disordered states. Recently, a few graph neural networks (GNNs) have been designed for brain connectivity synthesis and diagnosis. However, such non-Euclidean deep learning architectures might fail to capture the neural interactions between different brain regions as they are trained without guidance from any prior biological template-i.e., template-free learning. Here we assume that using a population-driven brain connectional template (CBT) that captures well the connectivity patterns fingerprinting a given brain state (e.g., healthy) can better guide the GNN training in its downstream learning task such as classification or regression. To this aim we design a plug-in graph registration network (GRN) that can be coupled with any conventional graph neural network (GNN) so as to boost its learning accuracy and generalizability to unseen samples. Our GRN is a graph generative adversarial network (gGAN), which registers brain graphs to a prior CBT. Next, the registered brain graphs are used to train typical GNN models. Our GRN can be integrated into any GNN working in an end-to-end fashion to boost its prediction accuracy. Our experiments showed that GRN remarkably boosted the prediction accuracy of four conventional GNN models across four neurological datasets.

Identifying autism spectrum disorder in resting-state fNIRS signals based on multiscale entropy and a two-branch deep learning network

BACKGROUND

The demand for early and precise identification of autism spectrum disorder (ASD) presented a challenge to the prediction of ASD with a non-invasive neuroimaging method.

NEW METHOD

A deep learning model was proposed to identify children with ASD using the resting-state functional near-infrared spectroscopy (fNIRS) signals. In this model, the input was the pattern of brain complexity represented by multiscale entropy of fNIRS time-series signals, with the purpose to solve the problem of deep learning analysis when the raw signals were limited by length and the number of subjects. The model consisted of a two-branch deep learning network, where one branch was a convolution neural network and the other was a long short-term memory neural network based on an attention mechanism.

RESULTS

Our model could achieve an identification accuracy of 94%. Further analysis used the SHapley Additive exPlanations (SHAP) method to balance the accuracy and the number of optical channels, thus reducing the complexity of fNIRS experiment. COMPARISON WITH PREVIOUSLY USED METHOD(S): in identification accuracy, our model was about 14% higher than previously used deep learning models with the same input and 4% higher than the same model but directly using fNIRS signals as input. We could obtain a discriminative accuracy of 90% with nearly half of the measurement channels by the SHAP method.

CONCLUSIONS

Using the pattern of brain complexity as input was effective in the deep learning model when the fNIRS signals were insufficient. With the SHAP method, it was possible to reduce the number of optical channels, while maintaining high accuracy in ASD identification.

Choice of intraoperative ultrasound adjuncts for brain tumor surgery

Background

Gliomas are among the most typical brain tumors tackled by neurosurgeons. During navigation for surgery of glioma brain tumors, preoperatively acquired static images may not be accurate due to shifts. Surgeons use intraoperative imaging technologies (2-Dimensional and navigated 3-Dimensional ultrasound) to assess and guide resections. This paper aims to precisely capture the importance of preoperative parameters to decide which type of ultrasound to be used for a particular surgery.

Methods

This paper proposes two bagging algorithms considering base classifier logistic regression and random forest. These algorithms are trained on different subsets of the original data set. The goodness of fit of Logistic regression-based bagging algorithms is established using hypothesis testing. Furthermore, the performance measures for random-forest-based bagging algorithms used are AUC under ROC and AUC under the precision-recall curve. We also present a composite model without compromising the explainability of the models.

Results

These models were trained on the data of 350 patients who have undergone brain surgery from 2015 to 2020. The hypothesis test shows that a single parameter is sufficient instead of all three dimensions related to the tumor ($$p < 0.05$$ p < 0.05 ). We observed that the choice of intraoperative ultrasound depends on the surgeon making a choice, and years of experience of the surgeon could be a surrogate for this dependence.

Conclusion

This study suggests that neurosurgeons may not need to focus on a large set of preoperative parameters in order to decide on ultrasound. Moreover, it personalizes the use of a particular ultrasound option in surgery. This approach could potentially lead to better resource management and help healthcare institutions improve their decisions to make the surgery more effective.

Machine learning for autism spectrum disorder diagnosis using structural magnetic resonance imaging: Promising but challenging

Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder that affects approximately 1% of the population and causes significant burdens. ASD's pathogenesis remains elusive; hence, diagnosis is based on a constellation of behaviors. Structural magnetic resonance imaging (sMRI) studies have shown several abnormalities in volumetric and geometric features of the autistic brain. However, inconsistent findings prevented most contributions from being translated into clinical practice. Establishing reliable biomarkers for ASD using sMRI is crucial for the correct diagnosis and treatment. In recent years, machine learning (ML) and specifically deep learning (DL) have quickly extended to almost every sector, notably in disease diagnosis. Thus, this has led to a shift and improvement in ASD diagnostic methods, fulfilling most clinical diagnostic requirements. However, ASD discovery remains difficult. This review examines the ML-based ASD diagnosis literature over the past 5 years. A literature-based taxonomy of the research landscape has been mapped, and the major aspects of this topic have been covered. First, we provide an overview of ML's general classification pipeline and the features of sMRI. Next, representative studies are highlighted and discussed in detail with respect to methods, and biomarkers. Finally, we highlight many common challenges and make recommendations for future directions. In short, the limited sample size was the main obstacle; Thus, comprehensive data sets and rigorous methods are necessary to check the generalizability of the results. ML technologies are expected to advance significantly in the coming years, contributing to the diagnosis of ASD and helping clinicians soon.

Bringing machine learning to research on intellectual and developmental disabilities: taking inspiration from neurological diseases

Intellectual and Developmental Disabilities (IDDs), such as Down syndrome, Fragile X syndrome, Rett syndrome, and autism spectrum disorder, usually manifest at birth or early childhood. IDDs are characterized by significant impairment in intellectual and adaptive functioning, and both genetic and environmental factors underpin IDD biology. Molecular and genetic stratification of IDDs remain challenging mainly due to overlapping factors and comorbidity. Advances in high throughput sequencing, imaging, and tools to record behavioral data at scale have greatly enhanced our understanding of the molecular, cellular, structural, and environmental basis of some IDDs. Fueled by the "big data" revolution, artificial intelligence (AI) and machine learning (ML) technologies have brought a whole new paradigm shift in computational biology. Evidently, the ML-driven approach to clinical diagnoses has the potential to augment classical methods that use symptoms and external observations, hoping to push the personalized treatment plan forward. Therefore, integrative analyses and applications of ML technology have a direct bearing on discoveries in IDDs. The application of ML to IDDs can potentially improve screening and early diagnosis, advance our understanding of the complexity of comorbidity, and accelerate the identification of biomarkers for clinical research and drug development. For more than five decades, the IDDRC network has supported a nexus of investigators at centers across the USA, all striving to understand the interplay between various factors underlying IDDs. In this review, we introduced fast-increasing multi-modal data types, highlighted example studies that employed ML technologies to illuminate factors and biological mechanisms underlying IDDs, as well as recent advances in ML technologies and their applications to IDDs and other neurological diseases. We discussed various molecular, clinical, and environmental data collection modes, including genetic, imaging, phenotypical, and behavioral data types, along with multiple repositories that store and share such data. Furthermore, we outlined some fundamental concepts of machine learning algorithms and presented our opinion on specific gaps that will need to be filled to accomplish, for example, reliable implementation of ML-based diagnosis technology in IDD clinics. We anticipate that this review will guide researchers to formulate AI and ML-based approaches to investigate IDDs and related conditions.

Identifying and Predicting Autism Spectrum Disorder Based on Multi-Site Structural MRI With Machine Learning

Although emerging evidence has implicated structural/functional abnormalities of patients with Autism Spectrum Disorder(ASD), definitive neuroimaging markers remain obscured due to inconsistent or incompatible findings, especially for structural imaging. Furthermore, brain differences defined by statistical analysis are difficult to implement individual prediction. The present study has employed the machine learning techniques under the unified framework in neuroimaging to identify the neuroimaging markers of patients with ASD and distinguish them from typically developing controls(TDC). To enhance the interpretability of the machine learning model, the study has processed three levels of assessments including model-level assessment, feature-level assessment, and biology-level assessment. According to these three levels assessment, the study has identified neuroimaging markers of ASD including the opercular part of bilateral inferior frontal gyrus, the orbital part of right inferior frontal gyrus, right rolandic operculum, right olfactory cortex, right gyrus rectus, right insula, left inferior parietal gyrus, bilateral supramarginal gyrus, bilateral angular gyrus, bilateral superior temporal gyrus, bilateral middle temporal gyrus, and left inferior temporal gyrus. In addition, negative correlations between the communication skill score in the Autism Diagnostic Observation Schedule (ADOS_G) and regional gray matter (GM) volume in the gyrus rectus, left middle temporal gyrus, and inferior temporal gyrus have been detected. A significant negative correlation has been found between the communication skill score in ADOS_G and the orbital part of the left inferior frontal gyrus. A negative correlation between verbal skill score and right angular gyrus and a significant negative correlation between non-verbal communication skill and right angular gyrus have been found. These findings in the study have suggested the GM alteration of ASD and correlated with the clinical severity of ASD disease symptoms. The interpretable machine learning framework gives sight to the pathophysiological mechanism of ASD but can also be extended to other diseases.

A comparative study of machine learning methods for predicting the evolution of brain connectivity from a baseline timepoint

BACKGROUND

Predicting the evolution of the brain network, also called connectome, by foreseeing changes in the connectivity weights linking pairs of anatomical regions makes it possible to spot connectivity-related neurological disorders in earlier stages and detect the development of potential connectomic anomalies. Remarkably, such a challenging prediction problem remains least explored in the predictive connectomics literature. It is a known fact that machine learning (ML) methods have proven their predictive abilities in a wide variety of computer vision problems. However, ML techniques specifically tailored for the prediction of brain connectivity evolution trajectory from a single timepoint are almost absent.

NEW METHOD

To fill this gap, we organized a Kaggle competition where 20 competing teams designed advanced machine learning pipelines for predicting the brain connectivity evolution from a single timepoint. The teams developed their ML pipelines with combination of data pre-processing, dimensionality reduction and learning methods. Each ML framework inputs a baseline brain connectivity matrix observed at baseline timepoint t₀ and outputs the brain connectivity map at a follow-up timepoint t₁. The longitudinal OASIS-2 dataset was used for model training and evaluation. Both random data split and 5-fold cross-validation strategies were used for ranking and evaluating the generalizability and scalability of each competing ML pipeline.

RESULTS

Utilizing an inclusive approach, we ranked the methods based on two complementary evaluation metrics (mean absolute error (MAE) and Pearson Correlation Coefficient (PCC)) and their performances using different training and testing data perturbation strategies (single random split and cross-validation). The final rank was calculated using the rank product for each competing team across all evaluation measures and validation strategies. Furthermore, we added statistical significance values to each proposed pipeline.

CONCLUSION

In support of open science, the developed 20 ML pipelines along with the connectomic dataset are made available on GitHub (https://github.com/basiralab/Kaggle-BrainNetPrediction-Toolbox). The outcomes of this competition are anticipated to lead the further development of predictive models that can foresee the evolution of the brain connectivity over time, as well as other types of networks (e.g., genetic networks).

ASD-SAENet: A Sparse Autoencoder, and Deep-Neural Network Model for Detecting Autism Spectrum Disorder (ASD) Using fMRI Data

Autism spectrum disorder (ASD) is a heterogenous neurodevelopmental disorder which is characterized by impaired communication, and limited social interactions. The shortcomings of current clinical approaches which are based exclusively on behavioral observation of symptomology, and poor understanding of the neurological mechanisms underlying ASD necessitates the identification of new biomarkers that can aid in study of brain development, and functioning, and can lead to accurate and early detection of ASD. In this paper, we developed a deep-learning model called ASD-SAENet for classifying patients with ASD from typical control subjects using fMRI data. We designed and implemented a sparse autoencoder (SAE) which results in optimized extraction of features that can be used for classification. These features are then fed into a deep neural network (DNN) which results in superior classification of fMRI brain scans more prone to ASD. Our proposed model is trained to optimize the classifier while improving extracted features based on both reconstructed data error and the classifier error. We evaluated our proposed deep-learning model using publicly available Autism Brain Imaging Data Exchange (ABIDE) dataset collected from 17 different research centers, and include more than 1,035 subjects. Our extensive experimentation demonstrate that ASD-SAENet exhibits comparable accuracy (70.8%), and superior specificity (79.1%) for the whole dataset as compared to other methods. Further, our experiments demonstrate superior results as compared to other state-of-the-art methods on 12 out of the 17 imaging centers exhibiting superior generalizability across different data acquisition sites and protocols. The implemented code is available on GitHub portal of our lab at: https://github.com/pcdslab/ASD-SAENet.