Identification and analysis of behavioral phenotypes in autism spectrum disorder via unsupervised machine learning

Early developmental milestone clusters of autistic children based on electronic health records

Autistic children vary in symptoms, co-morbidities, and response to interventions. This study aimed to identify clusters of autistic children with a distinct pattern of attaining early developmental milestones (EDMs). The clustering of 5836 autistic children was based on the attainment of 43 gross motor, fine motor, language, and social developmental milestones during the first 3 years of life as recorded in baby wellness visits. K-means cluster analysis detected four EDM clusters: mild (n = 1686); moderate (n = 1691); severe (n = 2265); and global (n = 194). The most prominent cluster differences were in the language domain. The global cluster showed earlier and greater developmental delay across domains, unique early gross motor delays, and more were born preterm via cesarean section. The severe cluster had poor language development prominently in the second year of life, and later fine motor delays. Moderate cluster had mainly language delays in the third year of life. The mild cluster mostly passed milestones. EDM clusters differed demographically, with higher socioeconomic status in mild cluster and lowest in global cluster. However, the severe cluster had more immigrant and non-Jewish mothers followed by the moderate cluster. The rates of parental concerns and provider developmental referrals were significantly higher in the global, followed by the severe, moderate, and mild EDM clusters. Autistic children's language and motor delay in the first 3 years can be grouped by common magnitude and onset profiles as distinct groups that may link to specific etiologies (like prematurity or genetics) and specific intervention programs. Early autism screening should be tailored to these different developmental profiles.

Autism spectrum disorder detection using facial images: A performance comparison of pretrained convolutional neural networks

Autism spectrum disorder (ASD) is a complex psychological syndrome characterized by persistent difficulties in social interaction, restricted behaviours, speech, and nonverbal communication. The impacts of this disorder and the severity of symptoms vary from person to person. In most cases, symptoms of ASD appear at the age of 2 to 5 and continue throughout adolescence and into adulthood. While this disorder cannot be cured completely, studies have shown that early detection of this syndrome can assist in maintaining the behavioural and psychological development of children. Experts are currently studying various machine learning methods, particularly convolutional neural networks, to expedite the screening process. Convolutional neural networks are considered promising frameworks for the diagnosis of ASD. This study employs different pre-trained convolutional neural networks such as ResNet34, ResNet50, AlexNet, MobileNetV2, VGG16, and VGG19 to diagnose ASD and compared their performance. Transfer learning was applied to every model included in the study to achieve higher results than the initial models. The proposed ResNet50 model achieved the highest accuracy, 92%, compared to other transfer learning models. The proposed method also outperformed the state-of-the-art models in terms of accuracy and computational cost.

Cluster analysis of teachers report for identifying symptoms of autism spectrum and/or attention deficit hyperactivity in school population: EPINED study

An early detection of Neurodevelopmental Disorders (NDDs) is crucial for their prognosis; however, the clinical heterogeneity of some disorders, such as autism spectrum disorder (ASD) or attention-deficit hyperactivity disorder (ADHD) is an obstacle to accurate diagnoses in children. In order to facilitate the screening process, the current study aimed to identify symptom-based clusters among a community-based sample of preschool and school-aged children, using behavioral characteristics reported by teachers. A total of 6894 children were assessed on four key variables: social communication differences, restricted behavior patterns, restless-impulsiveness, and emotional lability (pre-schoolers) or inattention and hyperactivity-impulsivity (school-aged). From these behavioral profiles, four clusters were identified for each age group. A cluster of ASD + ADHD and others including children with no pathology was clearly identified, whereas two other clusters were characterized by subthreshold ASD and/or ADHD symptoms. In the school-age children, the presence of ADHD was consistently observed with ASD patterns. In pre-schoolers, teachers were more proficient at identifying children who received a diagnosis for either ASD and/or ADHD from an early stage. Considering the significance of early detection and intervention of NDDs, teachers' insights are important. Therefore, promptly identifying subthreshold symptoms in children can help to minimize consequences in social and academic functioning.

A Prediction Model of Autism Spectrum Diagnosis from Well-Baby Electronic Data Using Machine Learning

Early detection of autism spectrum disorder (ASD) is crucial for timely intervention, yet diagnosis typically occurs after age three. This study aimed to develop a machine learning model to predict ASD diagnosis using infants' electronic health records obtained through a national screening program and evaluate its accuracy. A retrospective cohort study analyzed health records of 780,610 children, including 1163 with ASD diagnoses. Data encompassed birth parameters, growth metrics, developmental milestones, and familial and post-natal variables from routine wellness visits within the first two years. Using a gradient boosting model with 3-fold cross-validation, 100 parameters predicted ASD diagnosis with an average area under the ROC curve of 0.86 (SD < 0.002). Feature importance was quantified using the Shapley Additive explanation tool. The model identified a high-risk group with a 4.3-fold higher ASD incidence (0.006) compared to the cohort (0.001). Key predictors included failing six milestones in language, social, and fine motor domains during the second year, male gender, parental developmental concerns, non-nursing, older maternal age, lower gestational age, and atypical growth percentiles. Machine learning algorithms capitalizing on preventative care electronic health records can facilitate ASD screening considering complex relations between familial and birth factors, post-natal growth, developmental parameters, and parent concern.

A meta-analysis of cognitive flexibility in autism spectrum disorder

Cognitive flexibility is a fundamental process that underlies adaptive behaviour in response to environmental change. Studies examining the profile of cognitive flexibility in autism spectrum disorder (ASD) have reported inconsistent findings. To address whether difficulties with cognitive flexibility are characteristic of autism, we conducted a random-effects meta-analysis and employed subgroup analyses and meta-regression to assess the impact of relevant moderator variables such as task, outcomes, and age. Fifty-nine studies were included and comprised of 2122 autistic individuals without intellectual disabilities and 2036 neurotypical controls, with an age range of 4 to 85 years. The results showed that autistic individuals have greater difficulties with cognitive flexibility, with an overall statistically significant small to moderate effect size. Subgroup analyses revealed a significant difference between task outcomes, with perseverative errors obtaining the largest effect size. In summary, the present meta-analysis highlights the existence of cognitive flexibility difficulties in autistic people, in the absence of learning disabilities, but also that this profile is characterised by substantial heterogeneity. Potential contributing factors are discussed.

Comprehensive exploration of multi-modal and multi-branch imaging markers for autism diagnosis and interpretation: insights from an advanced deep learning model

Autism spectrum disorder is a complex neurodevelopmental condition with diverse genetic and brain involvement. Despite magnetic resonance imaging advances, autism spectrum disorder diagnosis and understanding its neurogenetic factors remain challenging. We propose a dual-branch graph neural network that effectively extracts and fuses features from bimodalities, achieving 73.9% diagnostic accuracy. To explain the mechanism distinguishing autism spectrum disorder from healthy controls, we establish a perturbation model for brain imaging markers and perform a neuro-transcriptomic joint analysis using partial least squares regression and enrichment to identify potential genetic biomarkers. The perturbation model identifies brain imaging markers related to structural magnetic resonance imaging in the frontal, temporal, parietal, and occipital lobes, while functional magnetic resonance imaging markers primarily reside in the frontal, temporal, occipital lobes, and cerebellum. The neuro-transcriptomic joint analysis highlights genes associated with biological processes, such as "presynapse," "behavior," and "modulation of chemical synaptic transmission" in autism spectrum disorder's brain development. Different magnetic resonance imaging modalities offer complementary information for autism spectrum disorder diagnosis. Our dual-branch graph neural network achieves high accuracy and identifies abnormal brain regions and the neuro-transcriptomic analysis uncovers important genetic biomarkers. Overall, our study presents an effective approach for assisting in autism spectrum disorder diagnosis and identifying genetic biomarkers, showing potential for enhancing the diagnosis and treatment of this condition.

Towards an accurate autism spectrum disorder diagnosis: multiple connectome views from fMRI data

Functional connectome has revealed remarkable potential in the diagnosis of neurological disorders, e.g. autism spectrum disorder. However, existing studies have primarily focused on a single connectivity pattern, such as full correlation, partial correlation, or causality. Such an approach fails in discovering the potential complementary topology information of FCNs at different connection patterns, resulting in lower diagnostic performance. Consequently, toward an accurate autism spectrum disorder diagnosis, a straightforward ambition is to combine the multiple connectivity patterns for the diagnosis of neurological disorders. To this end, we conduct functional magnetic resonance imaging data to construct multiple brain networks with different connectivity patterns and employ kernel combination techniques to fuse information from different brain connectivity patterns for autism diagnosis. To verify the effectiveness of our approach, we assess the performance of the proposed method on the Autism Brain Imaging Data Exchange dataset for diagnosing autism spectrum disorder. The experimental findings demonstrate that our method achieves precise autism spectrum disorder diagnosis with exceptional accuracy (91.30%), sensitivity (91.48%), and specificity (91.11%).

Digital tools for direct assessment of autism risk during early childhood: A systematic review

LAY ABSTRACT

The challenge of finding autistic children, and finding them early enough to make a difference for them and their families, becomes all the greater in parts of the world where human and material resources are in short supply. Poverty of resources delays interventions, translating into a poverty of outcomes. Digital tools carry potential to lessen this delay because they can be administered by non-specialists in children's homes, schools or other everyday environments, they can measure a wide range of autistic behaviours objectively and they can automate analysis without requiring an expert in computers or statistics. This literature review aimed to identify and describe digital tools for screening children who may be at risk for autism. These tools are predominantly at the 'proof-of-concept' stage. Both portable (laptops, mobile phones, smart toys) and fixed (desktop computers, virtual-reality platforms) technologies are used to present computerised games, or to record children's behaviours or speech. Computerised analysis of children's interactions with these technologies differentiates children with and without autism, with promising results. Tasks assessing social responses and hand and body movements are the most reliable in distinguishing autistic from typically developing children. Such digital tools hold immense potential for early identification of autism spectrum disorder risk at a large scale. Next steps should be to further validate these tools and to evaluate their applicability in a variety of settings. Crucially, stakeholders from underserved communities globally must be involved in this research, lest it fail to capture the issues that these stakeholders are facing.

On effectively predicting autism spectrum disorder therapy using an ensemble of classifiers

An ensemble of classifiers combines several single classifiers to deliver a final prediction or classification decision. An increasingly provoking question is whether such an ensemble can outperform the single best classifier. If so, what form of ensemble learning system (also known as multiple classifier learning systems) yields the most significant benefits in the size or diversity of the ensemble? In this paper, the ability of ensemble learning to predict and identify factors that influence or contribute to autism spectrum disorder therapy (ASDT) for intervention purposes is investigated. Given that most interventions are typically short-term in nature, henceforth, developing a robotic system that will provide the best outcome and measurement of ASDT therapy has never been so critical. In this paper, the performance of five single classifiers against several multiple classifier learning systems in exploring and predicting ASDT is investigated using a dataset of behavioural data and robot-enhanced therapy against standard human treatment based on 3000 sessions and 300 h, recorded from 61 autistic children. Experimental results show statistically significant differences in performance among the single classifiers for ASDT prediction with decision trees as the more accurate classifier. The results further show multiple classifier learning systems (MCLS) achieving better performance for ASDT prediction (especially those ensembles with three core classifiers). Additionally, the results show bagging and boosting ensemble learning as robust when predicting ASDT with multi-stage design as the most dominant architecture. It also appears that eye contact and social interaction are the most critical contributing factors to the ASDT problem among children.

Early Factors Predictive of Extreme High and Low Life Satisfaction 10 Years Post-Moderate to Severe Traumatic Brain Injury

OBJECTIVE

To identify demographic, injury-related, and 1-year postinjury clinical and functional predictors of high and low life satisfaction at 10 years after moderate to severe traumatic brain injury (TBI) using an extreme phenotyping approach.

SETTING

Multicenter longitudinal database study.

PARTICIPANTS

A total of 3040 people from the National Institute on Disability, Independent Living, and Rehabilitation Research TBI Model Systems database with life satisfaction data at 10 years post-TBI.

DESIGN

Multicenter, cross-sectional, observational design.

MAIN MEASURES

Satisfaction With Life Scale (outcome), Glasgow Coma Scale, Disability Rating Scale, Functional Independence Measure, Participation Assessment with Recombined Tools-Objective, Patient Health Questionnaire-9, and General Anxiety Disorder-7 (standardized predictors).

RESULTS

Greater cognitive and motor independence, more frequent community participation, and less depressive symptoms 1 year post-moderate to severe TBI predicted extreme high life satisfaction 10 years later. Non-Hispanic White and Hispanic individuals were significantly more likely than Black individuals to have extreme high life satisfaction 10 years post-TBI.

CONCLUSIONS

Extreme phenotyping analysis complements existing knowledge regarding life satisfaction post-moderate to severe TBI. From a chronic disease management perspective, future studies are needed to examine the feasibility and impact of early postinjury medical and rehabilitative interventions targeting cognitive and motor function, community participation, and mood on the maintenance/enhancement of long-term life satisfaction post-TBI.

A study of genetic heterogeneity in autism spectrum disorders based on plasma proteomic and metabolomic analysis: multiomics study of autism heterogeneity

Genetic heterogeneity poses a challenge to research and clinical translation of autism spectrum disorder (ASD). In this study, we conducted a plasma proteomic and metabolomic study of children with ASD with and without risk genes (de novo mutation) and controls to explore the impact of genetic heterogeneity on the search for biomarkers for ASD. In terms of the proteomic and metabolomic profiles, the groups of children with ASD carrying and those not carrying de novo mutation tended to cluster and overlap, and integrating them yielded differentially expressed proteins and differential metabolites that effectively distinguished ASD from controls. The mechanisms associated with them focus on several common and previously reported mechanisms. Proteomics results highlight the role of complement, inflammation and immunity, and cell adhesion. The main pathways of metabolic perturbations include amino acid, vitamin, glycerophospholipid, tryptophan, and glutamates metabolic pathways and solute carriers-related pathways. Integrating the two omics analyses revealed that L-glutamic acid and malate dehydrogenase may play key roles in the pathogenesis of ASD. These results suggest that children with ASD may have important underlying common mechanisms. They are not only potential therapeutic targets for ASD but also important contributors to the study of biomarkers for the disease.

Metabolomics: Perspectives on Clinical Employment in Autism Spectrum Disorder

Precision medicine is imminent, and metabolomics is one of the main actors on stage. We summarize and discuss the current literature on the clinical application of metabolomic techniques as a possible tool to improve early diagnosis of autism spectrum disorder (ASD), to define clinical phenotypes and to identify co-occurring medical conditions. A review of the current literature was carried out after PubMed, Medline and Google Scholar were consulted. A total of 37 articles published in the period 2010-2022 was included. Selected studies involve as a whole 2079 individuals diagnosed with ASD (1625 males, 394 females; mean age of 10, 9 years), 51 with other psychiatric comorbidities (developmental delays), 182 at-risk individuals (siblings, those with genetic conditions) and 1530 healthy controls (TD). Metabolomics, reflecting the interplay between genetics and environment, represents an innovative and promising technique to approach ASD. The metabotype may mirror the clinical heterogeneity of an autistic condition; several metabolites can be expressions of dysregulated metabolic pathways thus liable of leading to clinical profiles. However, the employment of metabolomic analyses in clinical practice is far from being introduced, which means there is a need for further studies for the full transition of metabolomics from clinical research to clinical diagnostic routine.

Autism screening: an unsupervised machine learning approach

Early screening of autism spectrum disorders (ASD) is a key area of research in healthcare. Currently artificial intelligence (AI)-driven approaches are used to improve the process of autism diagnosis using computer-aided diagnosis (CAD) systems. One of the issues related to autism diagnosis and screening data is the reliance of the predictions primarily on scores provided by medical screening methods which can be biased depending on how the scores are calculated. We attempt to reduce this bias by assessing the performance of the predictions related to the screening process using a new model that consists of a Self-Organizing Map (SOM) with classification algorithms. The SOM is employed prior to the diagnostic process to derive a new class label using clusters learnt from the independent features; these clusters are related to communication, repetitive traits, and social traits in the input dataset. Then, the new clusters are compared with existing class labels in the dataset to refine and eliminate any inconsistencies. Lastly, the refined dataset is utilised to derive classification systems for autism diagnosis. The new model was evaluated against a real-life autism screening dataset that consists of over 2000 instances of cases and controls. The results based on the refined dataset show that the proposed method achieves significantly higher accuracy, precision, and recall for the classification models derived when compared to models derived from the original dataset.

Genetic heterogeneity: Challenges, impacts, and methods through an associative lens

Genetic heterogeneity describes the occurrence of the same or similar phenotypes through different genetic mechanisms in different individuals. Robustly characterizing and accounting for genetic heterogeneity is crucial to pursuing the goals of precision medicine, for discovering novel disease biomarkers, and for identifying targets for treatments. Failure to account for genetic heterogeneity may lead to missed associations and incorrect inferences. Thus, it is critical to review the impact of genetic heterogeneity on the design and analysis of population level genetic studies, aspects that are often overlooked in the literature. In this review, we first contextualize our approach to genetic heterogeneity by proposing a high-level categorization of heterogeneity into "feature," "outcome," and "associative" heterogeneity, drawing on perspectives from epidemiology and machine learning to illustrate distinctions between them. We highlight the unique nature of genetic heterogeneity as a heterogeneous pattern of association that warrants specific methodological considerations. We then focus on the challenges that preclude effective detection and characterization of genetic heterogeneity across a variety of epidemiological contexts. Finally, we discuss systems heterogeneity as an integrated approach to using genetic and other high-dimensional multi-omic data in complex disease research.

Factors Associated With High and Low Life Satisfaction 10 Years After Traumatic Brain Injury

OBJECTIVE

To identify correlates of life satisfaction at 10 years after moderate to severe traumatic brain injury (TBI) using an extreme phenotyping approach.

DESIGN

Effect sizes were calculated in this observational cohort study to estimate relationships of 10-year postinjury extremely high, extremely low, and moderate life satisfaction with (1) pre-injury demographics, injury-related factors, and functional characteristics at inpatient rehabilitation admission and discharge; and (2) postinjury demographics and clinical and functional measures at 10 years postinjury.

SETTING

Multicenter longitudinal database study.

PARTICIPANTS

People identified from the National Institute on Disability, Independent Living, and Rehabilitation Research TBI Database with life satisfaction data at 10 years post TBI (N=4800).

INTERVENTIONS

Not applicable.

MAIN OUTCOME MEASURE

Satisfaction With Life Scale.

RESULTS

Although few pre-injury factors or clinical and functional factors shortly after injury were associated with 10-year life satisfaction groups, the following 10-year postinjury factors were associated with extremely high vs extremely low life satisfaction group membership: greater independent functioning, less disability, more frequent community participation, being employed, and having fewer depressive and anxiety symptoms. Those with extremely high life satisfaction were distinctly different from those with moderate and extremely low satisfaction. Extremely high life satisfaction was underrepresented among non-Hispanic Black persons relative to non-Hispanic White persons. Relationships between life satisfaction and independent functioning, disability, and participation were attenuated among non-Hispanic Black persons.

CONCLUSIONS

Extreme phenotyping analysis complements existing knowledge regarding life satisfaction after moderate to severe TBI and may inform acute and postacute clinical service delivery by comparing extremely high and extremely low life satisfaction subgroups. Findings suggest little association among personal, clinical, and functional characteristics early post TBI and life satisfaction 10 years later. Contemporaneous correlates of extremely high life satisfaction exist at 10 years post TBI, although the positive relationship of these variables to life satisfaction may be attenuated for non-Hispanic Black persons.

Automatic autism spectrum disorder detection using artificial intelligence methods with MRI neuroimaging: A review

Autism spectrum disorder (ASD) is a brain condition characterized by diverse signs and symptoms that appear in early childhood. ASD is also associated with communication deficits and repetitive behavior in affected individuals. Various ASD detection methods have been developed, including neuroimaging modalities and psychological tests. Among these methods, magnetic resonance imaging (MRI) imaging modalities are of paramount importance to physicians. Clinicians rely on MRI modalities to diagnose ASD accurately. The MRI modalities are non-invasive methods that include functional (fMRI) and structural (sMRI) neuroimaging methods. However, diagnosing ASD with fMRI and sMRI for specialists is often laborious and time-consuming; therefore, several computer-aided design systems (CADS) based on artificial intelligence (AI) have been developed to assist specialist physicians. Conventional machine learning (ML) and deep learning (DL) are the most popular schemes of AI used for diagnosing ASD. This study aims to review the automated detection of ASD using AI. We review several CADS that have been developed using ML techniques for the automated diagnosis of ASD using MRI modalities. There has been very limited work on the use of DL techniques to develop automated diagnostic models for ASD. A summary of the studies developed using DL is provided in the Supplementary Appendix. Then, the challenges encountered during the automated diagnosis of ASD using MRI and AI techniques are described in detail. Additionally, a graphical comparison of studies using ML and DL to diagnose ASD automatically is discussed. We suggest future approaches to detecting ASDs using AI techniques and MRI neuroimaging.

Diagnosis-Based Hybridization of Multimedical Tests and Sociodemographic Characteristics of Autism Spectrum Disorder Using Artificial Intelligence and Machine Learning Techniques: A Systematic Review

Autism spectrum disorder (ASD) is a complex neurobehavioral condition that begins in childhood and continues throughout life, affecting communication and verbal and behavioral skills. It is challenging to discover autism in the early stages of life, which prompted researchers to intensify efforts to reach the best solutions to treat this challenge by introducing artificial intelligence (AI) techniques and machine learning (ML) algorithms, which played an essential role in greatly assisting the medical and healthcare staff and trying to obtain the highest predictive results for autism spectrum disorder. This study is aimed at systematically reviewing the literature related to the criteria, including multimedical tests and sociodemographic characteristics in AI techniques and ML contributions. Accordingly, this study checked the Web of Science (WoS), Science Direct (SD), IEEE Xplore digital library, and Scopus databases. A set of 944 articles from 2017 to 2021 is collected to reveal a clear picture and better understand all the academic literature through a definitive collection of 40 articles based on our inclusion and exclusion criteria. The selected articles were divided based on similarity, objective, and aim evidence across studies. They are divided into two main categories: the first category is “diagnosis of ASD based on questionnaires and sociodemographic features” ( $n=39$ ). This category contains a subsection that consists of three categories: (a) early diagnosis of ASD towards analysis, (b) diagnosis of ASD towards prediction, and (c) diagnosis of ASD based on resampling techniques. The second category consists of “diagnosis ASD based on medical and family characteristic features” ( $n=1$ ). This multidisciplinary systematic review revealed the taxonomy, motivations, recommendations, and challenges of diagnosis ASD research in utilizing AI techniques and ML algorithms that need synergistic attention. Thus, this systematic review performs a comprehensive science mapping analysis and identifies the open issues that help accomplish the recommended solution of diagnosis ASD research. Finally, this study critically reviews the literature and attempts to address the diagnosis ASD research gaps in knowledge and highlights the available ASD datasets, AI techniques and ML algorithms, and the feature selection methods that have been collected from the final set of articles.

Exploring and Identifying Prognostic Phenotypes of Patients with Heart Failure Guided by Explainable Machine Learning

Identifying patient prognostic phenotypes facilitates precision medicine. This study aimed to explore phenotypes of patients with heart failure (HF) corresponding to prognostic condition (risk of mortality) and identify the phenotype of new patients by machine learning (ML). A unsupervised ML was applied to explore phenotypes of patients in a derivation dataset (n = 562) based on their medical records. Thereafter, supervised ML models were trained on the derivation dataset to classify these identified phenotypes. Then, the trained classifiers were further validated on an independent validation dataset (n = 168). Finally, Shapley additive explanations were used to interpret decision making of phenotype classification. Three patient phenotypes corresponding to stratified mortality risk (high, low, and intermediate) were identified. Kaplan−Meier survival curves among the three phenotypes had significant difference (pairwise comparison p < 0.05). Hazard ratio of all-cause mortality between patients in phenotype 1 (n = 91; high risk) and phenotype 3 (n = 329; intermediate risk) was 2.08 (95%CI 1.29−3.37, p = 0.003), and 0.26 (95%CI 0.11−0.61, p = 0.002) between phenotype 2 (n = 142; low risk) and phenotype 3. For phenotypes classification by random forest, AUCs of phenotypes 1, 2, and 3 were 0.736 ± 0.038, 0.815 ± 0.035, and 0.721 ± 0.03, respectively, slightly better than the decision tree. Then, the classifier effectively identified the phenotypes for new patients in the validation dataset with significant difference on survival curves and hazard ratios. Finally, age and creatinine clearance rate were identified as the top two most important predictors. ML could effectively identify patient prognostic phenotypes, facilitating reasonable management and treatment considering prognostic condition.

Annual Research Review: The transdiagnostic revolution in neurodevelopmental disorders

Practitioners frequently use diagnostic criteria to identify children with neurodevelopmental disorders and to guide intervention decisions. These criteria also provide the organising framework for much of the research focussing on these disorders. Study design, recruitment, analysis and theory are largely built on the assumption that diagnostic criteria reflect an underlying reality. However, there is growing concern that this assumption may not be a valid and that an alternative transdiagnostic approach may better serve our understanding of this large heterogeneous population of young people. This review draws on important developments over the past decade that have set the stage for much-needed breakthroughs in understanding neurodevelopmental disorders. We evaluate contemporary approaches to study design and recruitment, review the use of data-driven methods to characterise cognition, behaviour and neurobiology, and consider what alternative transdiagnostic models could mean for children and families. This review concludes that an overreliance on ill-fitting diagnostic criteria is impeding progress towards identifying the barriers that children encounter, understanding underpinning mechanisms and finding the best route to supporting them.

Annual Research Review: Translational machine learning for child and adolescent psychiatry

Children and adolescents could benefit from the use of predictive tools that facilitate personalized diagnoses, prognoses, and treatment selection. Such tools have not yet been deployed using traditional statistical methods, potentially due to the limitations of the paradigm and the need to leverage large amounts of digital data. This review will suggest that a machine learning approach could address these challenges and is designed to introduce new readers to the background, methods, and results in the field. A rationale is first introduced followed by an outline of fundamental elements of machine learning approaches. To provide an overview of the use of the techniques in child and adolescent literature, a scoping review of broad trends is then presented. Selected studies are also highlighted in order to draw attention to research areas that are closest to translation and studies that exhibit a high degree of experimental innovation. Limitations to the research, and machine learning approaches generally, are outlined in the penultimate section highlighting issues related to sample sizes, validation, clinical utility, and ethical challenges. Finally, future directions are discussed that could enhance the possibility of clinical implementation and address specific questions relevant to the child and adolescent psychiatry. The review gives a broad overview of the machine learning paradigm in order to highlight the benefits of a shift in perspective towards practically oriented statistical solutions that aim to improve clinical care of children and adolescents.

Applications of Unsupervised Machine Learning in Autism Spectrum Disorder Research: a Review

Large amounts of autism spectrum disorder (ASD) data is created through hospitals, therapy centers, and mobile applications; however, much of this rich data does not have pre-existing classes or labels. Large amounts of data—both genetic and behavioral—that are collected as part of scientific studies or a part of treatment can provide a deeper, more nuanced insight into both diagnosis and treatment of ASD. This paper reviews 43 papers using unsupervised machine learning in ASD, including k-means clustering, hierarchical clustering, model-based clustering, and self-organizing maps. The aim of this review is to provide a survey of the current uses of unsupervised machine learning in ASD research and provide insight into the types of questions being answered with these methods.

Application and research progress of machine learning in the diagnosis and treatment of neurodevelopmental disorders in children

The prevalence of neurodevelopment disorders (NDDs) among children has been on the rise. This has affected the health and social life of children. This condition has also imposed a huge economic burden on families and health care systems. Currently, it is difficult to perform early diagnosis of NDDs, which results in delayed intervention. For this reason, patients with NDDs have a prognosis. In recent years, machine learning (ML) technology, which integrates artificial intelligence technology and medicine, has been applied in the early detection and prediction of diseases based on data mining. This paper reviews the progress made in the application of ML in the diagnosis and treatment of NDDs in children based on supervised and unsupervised learning tools. The data reviewed here provide new perspectives on early diagnosis and treatment of NDDs.

EYE-C: Eye-Contact Robust Detection and Analysis during Unconstrained Child-Therapist Interactions in the Clinical Setting of Autism Spectrum Disorders

The high level of heterogeneity in Autism Spectrum Disorder (ASD) and the lack of systematic measurements complicate predicting outcomes of early intervention and the identification of better-tailored treatment programs. Computational phenotyping may assist therapists in monitoring child behavior through quantitative measures and personalizing the intervention based on individual characteristics; still, real-world behavioral analysis is an ongoing challenge. For this purpose, we designed EYE-C, a system based on OpenPose and Gaze360 for fine-grained analysis of eye-contact episodes in unconstrained therapist-child interactions via a single video camera. The model was validated on video data varying in resolution and setting, achieving promising performance. We further tested EYE-C on a clinical sample of 62 preschoolers with ASD for spectrum stratification based on eye-contact features and age. By unsupervised clustering, three distinct sub-groups were identified, differentiated by eye-contact dynamics and a specific clinical phenotype. Overall, this study highlights the potential of Artificial Intelligence in categorizing atypical behavior and providing translational solutions that might assist clinical practice.

Deep learning for neuroimaging-based diagnosis and rehabilitation of Autism Spectrum Disorder: A review

Accurate diagnosis of Autism Spectrum Disorder (ASD) followed by effective rehabilitation is essential for the management of this disorder. Artificial intelligence (AI) techniques can aid physicians to apply automatic diagnosis and rehabilitation procedures. AI techniques comprise traditional machine learning (ML) approaches and deep learning (DL) techniques. Conventional ML methods employ various feature extraction and classification techniques, but in DL, the process of feature extraction and classification is accomplished intelligently and integrally. DL methods for diagnosis of ASD have been focused on neuroimaging-based approaches. Neuroimaging techniques are non-invasive disease markers potentially useful for ASD diagnosis. Structural and functional neuroimaging techniques provide physicians substantial information about the structure (anatomy and structural connectivity) and function (activity and functional connectivity) of the brain. Due to the intricate structure and function of the brain, proposing optimum procedures for ASD diagnosis with neuroimaging data without exploiting powerful AI techniques like DL may be challenging. In this paper, studies conducted with the aid of DL networks to distinguish ASD are investigated. Rehabilitation tools provided for supporting ASD patients utilizing DL networks are also assessed. Finally, we will present important challenges in the automated detection and rehabilitation of ASD and propose some future works.

High spatial frequency filtered primes hastens happy faces categorization in autistic adults

Coarse information of a visual stimulus is conveyed by Low Spatial Frequencies (LSF) and is thought to be rapidly extracted to generate predictions. This may guide fast recognition with the subsequent integration of fine information, conveyed by High Spatial Frequencies (HSF). In autism, emotional face recognition is challenging, and might be related to alterations in LSF predictive processes. We analyzed the data of 27 autistic and 34 non autistic (NA) adults on an emotional Stroop task (i.e., emotional face with congruent or incongruent emotional word) with spatially filtered primes (HSF vs.LSF). We hypothesized that LSF primes would generate predictions leading to faster categorization of the target face compared to HSF primes, in the NA group but not in autism. Surprisingly, HSF primes led to faster categorization than LSF primes in both groups. Moreover, the advantage of HSF vs.LSF primes was stronger for angry than happy faces in NA, but was stronger for happy than angry faces in autistic participants. Drift diffusion modelling confirmed HSF advantage and showed a longer non-decision time (e.g., encoding) in autism. Despite LSF predictive impairments in autism was not corroborated, our analyses suggest low level processing specificities in autism.

Machine Learning for Subtyping Concussion Using a Clustering Approach

Background: Concussion subtypes are typically organized into commonly affected symptom areas or a combination of affected systems, an approach that may be flawed by bias in conceptualization or the inherent limitations of interdisciplinary expertise.

Objective: The purpose of this study was to determine whether a bottom-up, unsupervised, machine learning approach, could more accurately support concussion subtyping.

Methods: Initial patient intake data as well as objective outcome measures including, the Patient-Reported Outcomes Measurement Information System (PROMIS), Dizziness Handicap Inventory (DHI), Pain Catastrophizing Scale (PCS), and Immediate Post-Concussion Assessment and Cognitive Testing Tool (ImPACT) were retrospectively extracted from the Advance Concussion Clinic's database. A correlation matrix and principal component analysis (PCA) were used to reduce the dimensionality of the dataset. Sklearn's agglomerative clustering algorithm was then applied, and the optimal number of clusters within the patient database were generated. Between-group comparisons among the formed clusters were performed using a Mann-Whitney U test.

Results: Two hundred seventy-five patients within the clinics database were analyzed. Five distinct clusters emerged from the data when maximizing the Silhouette score (0.36) and minimizing the Davies-Bouldin score (0.83). Concussion subtypes derived demonstrated clinically distinct profiles, with statistically significant differences (p < 0.05) between all five clusters.

Conclusion: This machine learning approach enabled the identification and characterization of five distinct concussion subtypes, which were best understood according to levels of complexity, ranging from Extremely Complex to Minimally Complex. Understanding concussion in terms of Complexity with the utilization of artificial intelligence, could provide a more accurate concussion classification or subtype approach; one that better reflects the true heterogeneity and complex system disruptions associated with mild traumatic brain injury.

Imputing cognitive impairment in SPARK, a large autism cohort

Diverse large cohorts are necessary for dissecting subtypes of autism, and intellectual disability is one of the most robust endophenotypes for analysis. However, current cognitive assessment methods are not feasible at scale. We developed five commonly used machine learning models to predict cognitive impairment (FSIQ<80 and FSIQ<70) and FSIQ scores among 521 children with autism using parent-reported online surveys in SPARK, and evaluated them in an independent set (n = 1346) with a missing data rate up to 70%. We assessed accuracy, sensitivity, and specificity by comparing predicted cognitive levels against clinical IQ data. The elastic-net model has good performance (AUC = 0.876, sensitivity = 0.772, specificity = 0.803) using 129 predictive features to impute cognitive impairment (FSIQ<80). Top-ranked predictive features included parent-reported language and cognitive levels, age at autism diagnosis, and history of services. Prediction of FSIQ<70 and FSIQ scores also showed good performance. We show cognitive levels can be imputed with high accuracy for children with autism, using commonly collected parent-reported data and standardized surveys. The current model offers a method for large-scale autism studies seeking estimates of cognitive ability when standardized psychometric testing is not feasible. LAY SUMMARY: Children with autism who have more severe learning challenges or cognitive impairment have different needs that are important to consider in research studies. When children in our study were missing standardized cognitive testing scores, we were able to use machine learning with other information to correctly "guess" when they have cognitive impairment about 80% of the time. We can use this information in research in the future to develop more appropriate treatments for children with autism and cognitive impairment.

Informing Developmental Milestone Achievement for Children With Autism: Machine Learning Approach

Background

Care for children with autism spectrum disorder (ASD) can be challenging for families and medical care systems. This is especially true in low- and- middle-income countries such as Bangladesh. To improve family–practitioner communication and developmental monitoring of children with ASD, mCARE (Mobile-Based Care for Children with Autism Spectrum Disorder Using Remote Experience Sampling Method) was developed. Within this study, mCARE was used to track child milestone achievement and family sociodemographic assets to inform mCARE feasibility/scalability and family asset–informed practitioner recommendations.

Objective

The objectives of this paper are threefold. First, it documents how mCARE can be used to monitor child milestone achievement. Second, it demonstrates how advanced machine learning models can inform our understanding of milestone achievement in children with ASD. Third, it describes family/child sociodemographic factors that are associated with earlier milestone achievement in children with ASD (across 5 machine learning models).

Methods

Using mCARE-collected data, this study assessed milestone achievement in 300 children with ASD from Bangladesh. In this study, we used 4 supervised machine learning algorithms (decision tree, logistic regression, K-nearest neighbor [KNN], and artificial neural network [ANN]) and 1 unsupervised machine learning algorithm (K-means clustering) to build models of milestone achievement based on family/child sociodemographic details. For analyses, the sample was randomly divided in half to train the machine learning models and then their accuracy was estimated based on the other half of the sample. Each model was specified for the following milestones: Brushes teeth, Asks to use the toilet, Urinates in the toilet or potty, and Buttons large buttons.

Results

This study aimed to find a suitable machine learning algorithm for milestone prediction/achievement for children with ASD using family/child sociodemographic characteristics. For Brushes teeth, the 3 supervised machine learning models met or exceeded an accuracy of 95% with logistic regression, KNN, and ANN as the most robust sociodemographic predictors. For Asks to use toilet, 84.00% accuracy was achieved with the KNN and ANN models. For these models, the family sociodemographic predictors of “family expenditure” and “parents’ age” accounted for most of the model variability. The last 2 parameters, Urinates in toilet or potty and Buttons large buttons, had an accuracy of 91.00% and 76.00%, respectively, in ANN. Overall, the ANN had a higher accuracy (above ~80% on average) among the other algorithms for all the parameters. Across the models and milestones, “family expenditure,” “family size/type,” “living places,” and “parent’s age and occupation” were the most influential family/child sociodemographic factors.

Conclusions

mCARE was successfully deployed in a low- and middle-income country (ie, Bangladesh), providing parents and care practitioners a mechanism to share detailed information on child milestones achievement. Using advanced modeling techniques this study demonstrates how family/child sociodemographic elements can inform child milestone achievement. Specifically, families with fewer sociodemographic resources reported later milestone attainment. Developmental science theories highlight how family/systems can directly influence child development and this study provides a clear link between family resources and child developmental progress. Clinical implications for this work could include supporting the larger family system to improve child milestone achievement.

Unsupervised Machine Learning for Identifying Challenging Behavior Profiles to Explore Cluster-Based Treatment Efficacy in Children With Autism Spectrum Disorder: Retrospective Data Analysis Study

Background

Challenging behaviors are prevalent among individuals with autism spectrum disorder; however, research exploring the impact of challenging behaviors on treatment response is lacking.

Objective

The purpose of this study was to identify types of autism spectrum disorder based on engagement in different challenging behaviors and evaluate differences in treatment response between groups.

Methods

Retrospective data on challenging behaviors and treatment progress for 854 children with autism spectrum disorder were analyzed. Participants were clustered based on 8 observed challenging behaviors using k means, and multiple linear regression was performed to test interactions between skill mastery and treatment hours, cluster assignment, and gender.

Results

Seven clusters were identified, which demonstrated a single dominant challenging behavior. For some clusters, significant differences in treatment response were found. Specifically, a cluster characterized by low levels of stereotypy was found to have significantly higher levels of skill mastery than clusters characterized by self-injurious behavior and aggression (P<.003).

Conclusions

These findings have implications on the treatment of individuals with autism spectrum disorder. Self-injurious behavior and aggression were prevalent among participants with the worst treatment response, thus interventions targeting these challenging behaviors may be worth prioritizing. Furthermore, the use of unsupervised machine learning models to identify types of autism spectrum disorder shows promise.

Validation strategies for subtypes in psychiatry: A systematic review of research on autism spectrum disorder

Heterogeneity within autism spectrum disorder (ASD) is recognized as a challenge to both biological and psychological research, as well as clinical practice. To reduce unexplained heterogeneity, subtyping techniques are often used to establish more homogeneous subtypes based on metrics of similarity and dissimilarity between people. We review the ASD literature to create a systematic overview of the subtyping procedures and subtype validation techniques that are used in this field. We conducted a systematic review of 156 articles (2001-June 2020) that subtyped participants (range N of studies = 17-20,658), of which some or all had an ASD diagnosis. We found a large diversity in (parametric and non-parametric) methods and (biological, psychological, demographic) variables used to establish subtypes. The majority of studies validated their subtype results using variables that were measured concurrently, but were not included in the subtyping procedure. Other investigations into subtypes' validity were rarer. In order to advance clinical research and the theoretical and clinical usefulness of identified subtypes, we propose a structured approach and present the SUbtyping VAlidation Checklist (SUVAC), a checklist for validating subtyping results.

Identifying Subgroups of Patients With Autism by Gene Expression Profiles Using Machine Learning Algorithms

Objectives: The identification of subgroups of autism spectrum disorder (ASD) may partially remedy the problems of clinical heterogeneity to facilitate the improvement of clinical management. The current study aims to use machine learning algorithms to analyze microarray data to identify clusters with relatively homogeneous clinical features. Methods: The whole-genome gene expression microarray data were used to predict communication quotient (SCQ) scores against all probes to select differential expression regions (DERs). Gene set enrichment analysis was performed for DERs with a fold-change >2 to identify hub pathways that play a role in the severity of social communication deficits inherent to ASD. We then used two machine learning methods, random forest classification (RF) and support vector machine (SVM), to identify two clusters using DERs. Finally, we evaluated how accurately the clusters predicted language impairment. Results: A total of 191 DERs were initially identified, and 54 of them with a fold-change >2 were selected for the pathway analysis. Cholesterol biosynthesis and metabolisms pathways appear to act as hubs that connect other trait-associated pathways to influence the severity of social communication deficits inherent to ASD. Both RF and SVM algorithms can yield a classification accuracy level >90% when all 191 DERs were analyzed. The ASD subtypes defined by the presence of language impairment, a strong indicator for prognosis, can be predicted by transcriptomic profiles associated with social communication deficits and cholesterol biosynthesis and metabolism. Conclusion: The results suggest that both RF and SVM are acceptable options for machine learning algorithms to identify AD subgroups characterized by clinical homogeneity related to prognosis.

A Review of Machine Learning Methods of Feature Selection and Classification for Autism Spectrum Disorder

Autism Spectrum Disorder (ASD), according to DSM-5 in the American Psychiatric Association, is a neurodevelopmental disorder that includes deficits of social communication and social interaction with the presence of restricted and repetitive behaviors. Children with ASD have difficulties in joint attention and social reciprocity, using non-verbal and verbal behavior for communication. Due to these deficits, children with autism are often socially isolated. Researchers have emphasized the importance of early identification and early intervention to improve the level of functioning in language, communication, and well-being of children with autism. However, due to limited local assessment tools to diagnose these children, limited speech-language therapy services in rural areas, etc., these children do not get the rehabilitation they need until they get into compulsory schooling at the age of seven years old. Hence, efficient approaches towards early identification and intervention through speedy diagnostic procedures for ASD are required. In recent years, advanced technologies like machine learning have been used to analyze and investigate ASD to improve diagnostic accuracy, time, and quality without complexity. These machine learning methods include artificial neural networks, support vector machines, a priori algorithms, and decision trees, most of which have been applied to datasets connected with autism to construct predictive models. Meanwhile, the selection of features remains an essential task before developing a predictive model for ASD classification. This review mainly investigates and analyzes up-to-date studies on machine learning methods for feature selection and classification of ASD. We recommend methods to enhance machine learning's speedy execution for processing complex data for conceptualization and implementation in ASD diagnostic research. This study can significantly benefit future research in autism using a machine learning approach for feature selection, classification, and processing imbalanced data.

Exploring the Eating Disorder Examination Questionnaire, Clinical Impairment Assessment, and Autism Quotient to Identify Eating Disorder Vulnerability: A Cluster Analysis

Eating disorders are very complicated and many factors play a role in their manifestation. Furthermore, due to the variability in diagnosis and symptoms, treatment for an eating disorder is unique to the individual. As a result, there are numerous assessment tools available, which range from brief survey questionnaires to in-depth interviews conducted by a professional. One of the many benefits to using machine learning is that it offers new insight into datasets that researchers may not previously have, particularly when compared to traditional statistical methods. The aim of this paper was to employ k-means clustering to explore the Eating Disorder Examination Questionnaire, Clinical Impairment Assessment, and Autism Quotient scores. The goal is to identify prevalent cluster topologies in the data, using the truth data as a means to validate identified groupings. Our results show that a model with k = 2 performs the best and clustered the dataset in the most appropriate way. This matches our truth data group labels, and we calculated our model’s accuracy at 78.125%, so we know that our model is working well. We see that the Eating Disorder Examination Questionnaire (EDE-Q) and Clinical Impairment Assessment (CIA) scores are, in fact, important discriminators of eating disorder behavior.

Deep representation learning of electronic health records to unlock patient stratification at scale

Deriving disease subtypes from electronic health records (EHRs) can guide next-generation personalized medicine. However, challenges in summarizing and representing patient data prevent widespread practice of scalable EHR-based stratification analysis. Here we present an unsupervised framework based on deep learning to process heterogeneous EHRs and derive patient representations that can efficiently and effectively enable patient stratification at scale. We considered EHRs of 1,608,741 patients from a diverse hospital cohort comprising a total of 57,464 clinical concepts. We introduce a representation learning model based on word embeddings, convolutional neural networks, and autoencoders (i.e., ConvAE) to transform patient trajectories into low-dimensional latent vectors. We evaluated these representations as broadly enabling patient stratification by applying hierarchical clustering to different multi-disease and disease-specific patient cohorts. ConvAE significantly outperformed several baselines in a clustering task to identify patients with different complex conditions, with 2.61 entropy and 0.31 purity average scores. When applied to stratify patients within a certain condition, ConvAE led to various clinically relevant subtypes for different disorders, including type 2 diabetes, Parkinson's disease, and Alzheimer's disease, largely related to comorbidities, disease progression, and symptom severity. With these results, we demonstrate that ConvAE can generate patient representations that lead to clinically meaningful insights. This scalable framework can help better understand varying etiologies in heterogeneous sub-populations and unlock patterns for EHR-based research in the realm of personalized medicine.