Web-based autism screening using facial images and convolutional neural network

BackgroundThe 1.6 Mb 3q29 deletion is associated with neurodevelopmental and neuropsychiatric phenotypes, including a 19-fold increased risk for autism spectrum disorder (ASD). Previous work by our team identified elevated social disability in this population via parent-report questionnaires. However, clinical features of ASD in this population have not been explored in detail.MethodsThirty-one individuals with 3q29 deletion syndrome (3q29del, 61.3% male) were evaluated using two gold-standard clinical ASD evaluations: the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2), and the Autism Diagnostic Interview, Revised (ADI-R). Four matched comparators for each subject were ascertained from the National Database for Autism Research. Item-level scores on the ADOS-2 and ADI-R were compared between subjects with 3q29del and matched comparators.ResultsSubjects with 3q29del and no ASD (3q29del-ASD) had greater evidence of social disability compared to typically developing (TD) comparison subjects across the ADOS-2. Subjects with 3q29del and ASD (3q29del + ASD) were largely indistinguishable from non-syndromic ASD (nsASD) subjects on the ADOS-2. 3q29del + ASD performed significantly better on social communication on the ADI-R than nsASD (3q29 + ASD mean = 11.36; nsASD mean = 15.70; p = 0.01), and this was driven by reduced deficits in nonverbal communication (3q29 + ASD mean = 1.73; nsASD mean = 3.63; p = 0.03). 3q29del + ASD reported significantly later age at the first two-word phrase compared to nsASD (3q29del + ASD mean = 43.89 months; nsASD mean = 37.86 months; p = 0.01). However, speech delay was not related to improved nonverbal communication in 3q29del + ASD.LimitationsThere were not enough TD comparators with ADI-R data in NDAR to include in the present analysis. Additionally, our relatively small sample size made it difficult to assess race and ethnicity effects.Conclusions3q29del is associated with significant social disability, irrespective of ASD diagnosis. 3q29del + ASD have similar levels of social disability to nsASD, while 3q29del-ASD have significantly increased social disability compared to TD individuals. However, social communication is reasonably well preserved in 3q29del + ASD relative to nsASD. It is critical that verbal ability and social disability be examined separately in this population to ensure equal access to ASD and social skills evaluations and services.

Autism Diagnostic Interview -Revised (ADI-R) and Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) are well-validated semi-structured instruments, widely used in clinical practice and research for decades.They both follow highly standardised procedures to provide a thorough assessment of individuals suspected of having autism spectrum disorder (ASD).Both instruments have been found to be useful in differentiating between children, adolescents and adults with ASD and those with non-ASD diagnoses, especially when both are used in combination.As with many measures of ASD symptoms, scores on ADI-R and ADOS-2 have been shown to be susceptible to effects of developmental variables, including age and IQ as well as behavioural difficulties of a subject.Experienced clinicians are aware that the scores can be affected by comorbidity or sex of an individual, which is the case in females with autism who have not been diagnosed early in life or those who do not have intellectual disability and/or profound communication difficulties.The article presents ADI-R and ADOS-2 in a succinct way along with basic recommendations for their use, which are the same regardless of location and country.The purpose of this article is to present guidelines and standards of use for ADI-R and ADOS-2 which are distinguished from the standards of purchase.Administration and coding of ADI-R and ADOS-2 are highly standardised, and based on these guidelines, only professionals who completed ADI-R and ADOS-2 test-specific trainings can be considered competent users to provide valid assessments.

Prevalence of Autism Spectrum Disorder in China: A Nationwide Multi-Centre Population-Based Study Among Children Aged 6 to 12 Years

Background: The accuracy of the Norwegian Social Communication Questionnaire (SCQ) in identifying autism spectrum disorder (ASD) within child psychiatry is not known. Method: We investigated the accuracy of SCQ in identifying clinically diagnosed ASD and ASD classified by the Developmental, Dimensional and Diagnostic Interview (3Di) and the Autism Diagnostic Observation Schedule (ADOS-2). Results: The SCQ cut-off of 15 showed low sensitivity (Se) for correct classification of clinically diagnosed ASD (Se = 38 %) and ADOS-2 classified ASD (Se = 30 %) with 78 % and 76 % specificity (SP), respectively. For 3Di classified ASD, Se was somewhat higher (49 %) and had good specificity (86 %), with the best positive predictive value (.73). Conclusions/Implications: The SCQ showed low accuracy in identifying a clinical ASD diagnosis in child psychiatry. However, children with SCQ scores at or above threshold had a high risk of also being categorised with ASD according to the 3Di. We recommend that future studies include more detailed information about the basis for the final diagnostic conclusions. Furthermore, SCQ’s characteristics should be evaluated in samples of children within a broad range of ASD that includes individuals with delayed development. Investigation of the suitability of other screening instruments for ASD for diagnostic work in child psychiatric clinics is also needed. Keywords: Autism Diagnostic Observation Schedule (ADOS-2), Developmental, Dimensional and Diagnostic Interview (3Di), sensitivity; specificity, Social Communication Questionnaire (SCQ)

AUTISM SPECTRUM DISORDER EVALUATION IN TODDLERS AND PRESCHOOLERS

Objectives:The objective of this study was to investigate the concordance of Diagnostic and Statistical Manual of Mental Disorders (DSM-IV and DSM-5) diagnostic criteria for autism spectrum disorder (ASD).Methods:We retrospectively reviewed the medical records of 170 subjects (age range: 3–23, 140 boys) with developmental delay or social deficit from January 2011 to July 2016 at the Department of Psychiatry of Asan Medical Center. The Autism Diagnostic Interview-Revised (ADI-R), the Autism Diagnostic Observation Schedule (ADOS), and intelligence tests were performed for each subject. Diagnosis was reviewed and confirmed for each subject with DSM-IV Pervasive Developmental Disorder (PDD) and DSM-5 ASD criteria, respectively.Results:Fifty-eight of 145 subjects (34.1%) who were previously diagnosed as having PDD in DSM-IV did not meet DSM-5 ASD criteria. Among them, 28 (48.3%) had Asperger’s disorder based on DSM-IV. Most algorithm scores on ADOS and all algorithm scores on ADI-R were highest in subjects who met both DSM-IV PDD criteria and DSM-5 ASD criteria (the Convergent group), followed by subjects with a DSM-IV PDD diagnosis who did not have a DSM-5 ASD diagnosis (the Divergent group), and subjects who did not meet either DSM-IV PDD or DSM-5 ASD criteria (the non-PDD group). Intelligence quotient was lower in the Convergent group than in the Divergent group.Conclusion:The results of our study suggest that ASD prevalence estimates could be lower under DSM-5 than DSM-IV diagnostic criteria. Further prospective study on the impact of new DSM-5 ASD diagnoses in Koreans with ASD is needed.

BackgroundMany women with eating disorders (EDs) have social impairments and difficulties with flexibility. It is unclear to what extent these are manifestations of an underlying autism spectrum disorder (ASD); or whether they are instead the consequence of starvation, anxiety, low mood or obsessive compulsive disorder, all of which are highly prevalent in EDs. The resolution of this clinically and theoretically important uncertainty will require the use of gold-standard ASD assessment measures. To date these have not been employed in ED research. This case series is the first report of a well-validated, direct-observational measure of ASD, the Autism Diagnostic Observation Schedule (ADOS), being administered to women with EDs. We aimed to learn about the feasibility of the ADOS in this population, and to contribute to debates about whether a sub-group with EDs really have ASD.MethodsTen women (mean age = 26.4 years, range = 19 to 38 years) who had a suspected ASD due to social and flexibility difficulties and were receiving treatment for ED (seven anorexia, two ED not otherwise specified, one bulimia) at a specialist service (four inpatient, six outpatient) received an ADOS Module 4 assessment.ResultsAll 10 participants completed all activities of the ADOS Module 4. Five scored in the ASD range on the ADOS diagnostic algorithm. An additional two were judged likely to have ASD, even though they scored below the ADOS’s diagnostic threshold. This was on the basis of clinical observation, participant self-report and parent report. The seven women who we estimated to have ASD all reported autistic difficulties prior to the onset of their ED. They commonly described longstanding non-autistic neurodevelopmental problems, including dyslexia, dyspraxia and epilepsy. Only one had a childhood diagnosis of ASD.ConclusionsA substantial proportion of women with EDs who present with social and flexibility difficulties may have an unrecognised ASD, indicated by a constellation of autistic difficulties that appears to predate the onset of their eating problems. The ADOS is a useful component of an ASD assessment for adult women with ED.

Relationship between the social functioning of children with autism spectrum disorders and their siblings’ competencies/problem behaviors

To estimate the diagnostic prevalence of autism spectrum disorders (ASDs) in a low birth weight (LBW) cohort. Participants belonged to a regional birth cohort of infants (N = 1105) born weighing <2000 g between October 1, 1984, and July 3, 1989, and followed up by periodic assessments to 21 years of age. At 16 years (n = 623), adolescents were screened for ASD using a wide net (previous professional diagnosis of an ASD or a score above a liberal cutoff on the Social Communication Questionnaire or the Autism Spectrum Symptoms Questionnaire). At 21 years (n = 189), 60% of screen positives and 24% of screen negatives were assessed for diagnoses of ASD by the Autism Diagnostic Observation Schedule or the Autism Diagnostic Interview-Revised. Samples retained at ages 16 and 21 years were representative of samples assessed at earlier ages except for lower levels of social risk. Of positive screens, 11 of 70 had ASD; of negative screens, 3 of 119 had ASD. The fractions of the 2 screening groups with ASD (14.3% in screen-positives and 2.5% in screen negatives) were weighted by fractions of screen-positives and screen-negatives among the adolescents (18.8% and 81.2%, respectively). This calculation produced an estimated prevalence rate of ASD in the entire cohort of 5% (31 of 623). The diagnostic prevalence of ASD in this LBW preterm cohort was higher than that reported by the Centers for Disease Control and Prevention for 8-year-olds in the general US population in 2006.

Autism Spectrum Disorder (ASD) was characterized by delay in social interactions development, repetitive behaviors and narrow interest, which usually diagnosed with standard diagnostic tools such as Autism Diagnostic Observation Schedule (ADOS) and Autism Diagnostic Interview-Revised (ADIR-R). Previous work has implemented machine-learning methods for the classification of ASD, however they used different types of dataset such as brain images for MRI and EEG, risk genes in genetic profiles and behavior evaluation based on ADOS and ADI-R. Here a trial on using Autism Spectrum Questions (AQ) to build models that have higher potential to classify ASD was developed. In this research, Chi-square and Least Absolute Shrinkage and Selection Operator (LASSO) have been selected as feature selection methods to select the most important features for 3 supervised machine learning algorithms, which are Random Forest, Logistic Regression and K-Nearest Neighbors with K-fold cross validation. The performance was evaluated in which results Logistic Regression scored the highest accuracy with 97.541% using model with 13 selected features based on Chi-square selection method.

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by difficulty in social interactions, communication, and repetitive behaviors.Protocols like ADOS (Autism Diagnostic Observation Schedule) and ADI-R (Autism Diagnostic Interview Revised) are used by experts to assess the subject's behavior which is timeconsuming.Over the decade, researchers have studied the application of various Machine Learning techniques for ASD diagnosis through facial feature analysis, eye movement tracking, questionnaire analysis, functional magnetic resonance imaging (fMRI) analysis, etc.However, these techniques are not helpful for the parent or guardian of the child to perform an initial screening.This research proposes a novel deep learning model to diagnose ASD using general videos of the subject performing some tasks with the parent/guardian.Since there is no publicly available dataset on ASD videos, a dataset is created by collecting the videos of autistic children performing some activities with parents/guardians from YouTube from different demographic locations.These videos are then converted to skeletal key points to extract the child's engagement and social interaction in a given task.The proposed CNN-LSTM model is trained on 80% of the collected videos and then tested on the remaining 20%.The experiment results on various combinations of pre-trained CNN models and LSTM/BiLSTM show that the proposed model can be used as an initial autism screening tool.Among the different combinations, the MobileNet and Bi-LSTM combo achieved the best test accuracy of 84.95% with 89% precision, recall and F1-score.

Autism spectrum disorder (ASD) represents a set of neurodevelopmental disorders with a strong genetic aetiology. Chromosomal rearrangements have been detected in 5-10% of the patients with ASD, and recent applications of array comparative genomic hybridisation (aCGH) are identifying further candidate regions and genes. In this study, we present four patients who implicate microcephalin 1 (MCPH1) in band 8p23.1 as an ASD susceptibility gene. Patient 1 was a girl with a syndromic form of autistic disorder satisfying the Autism Diagnostic Interview-Revised (ADI-R), Autism Diagnostic Observation Schedule (ADOS) and Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) criteria. Oligonucleotide aCGH (oaCGH) showed that she had a classic inv dup del(8)(qter-> p23.1::p23.1-> p21.2) containing at least three candidate genes; MCPH1 and DLGAP2 within the 6.9-Mb terminal deletion and NEF3 within the concomitant 14.1-Mb duplication. Three further patients with MCPH1 copy number changes were found using single-nucleotide polymorphism (SNP) array analysis in a cohort of 54 families with ASD patients. Our results show that ASD can be a component of the classical inv dup del(8) phenotype and identify changes in copy number of MCPH1 as a susceptibility factor for ASD in the distal short arm of chromosome 8.

BackgroundA small percentage of people with autism spectrum disorders (ASD) have alterations in chromosome 15q11.2-q3, the critical region for Prader-Willi syndrome (PWS). Data are limited, however, on the rates and characteristics of ASD in PWS. Previous estimates of ASD in PWS (25 to 41%) are questionable as they are based solely on autism screeners given to parents. Inaccurate diagnoses of ASD in PWS can mislead intervention and future research.MethodsOne hundred forty-six children and youth with PWS aged 4 to 21 years (M = 11) were assessed with the Autism Diagnostic Observation Schedule-2 (ADOS-2). An expert clinical team-made best-estimate ASD diagnoses based on ADOS-2 videotapes, calibrated severity scores, and children’s developmental histories and indices of current functioning. Children were also administered the Kaufman Brief Intelligence Test-2, and parents completed the Repetitive Behavior Scale-Revised and Vineland Adaptive Behavior Scales. Scores were compared across children with PWS + ASD versus PWS only. The performance of an ASD screener, the Social Communication Questionnaire (SCQ) and the ADOS-2 were evaluated in relation to best-estimate diagnoses.ResultsBest-estimate diagnoses of ASD were made in 18 children, or 12.3% of the sample, and the majority of them had the maternal uniparental disomy (mUPD) PWS genetic subtype. Compared to the PWS-only group, children with PWS + ASD had lower verbal and composite IQ’s and adaptive daily living and socialization skills, as well as elevated stereotypies and restricted interests. Regardless of ASD status, compulsivity and insistence on sameness in routines or events were seen in 76–100% of children and were robustly correlated with lower adaptive functioning. The SCQ yielded a 29–49% chance that screen-positive cases will indeed have ASD. The ADOS-2 had higher sensitivity, specificity and predictive values. Communication problems were seen in children who were ADOS-2 positive but deemed not to have ASD by the clinical team.ConclusionsAutism screeners should not be the sole index of probable ASD in PWS; children need to be directly observed and evaluated. Compulsivity and insistence on sameness are salient in PWS and likely impede adaptive functioning. Most children with PWS only evidenced sub-threshold problems in social interactions that could signal risks for other psychopathologies.

The administration requirements of the Autism Diagnostic Observation Schedule and the Autism Diagnostic Interview-Revised, widely used in high-income countries, make them less feasible for diagnosis of autism spectrum disorder in low- and middle-income countries. The flexible administration requirements of the Childhood Autism Rating Scale have resulted in its use in both high-income countries and low- and middle-income countries. This study examines the agreement between assessments using the Childhood Autism Rating Scale with those using the Autism Diagnostic Observation Schedule or Autism Diagnostic Observation Schedule, Second Edition and Autism Diagnostic Interview-Revised in Jamaica. Children aged 2-8 years (n = 149) diagnosed with autism by an experienced clinician using the Childhood Autism Rating Scale were re-evaluated using the Autism Diagnostic Observation Schedule and Autism Diagnostic Interview-Revised. The proportion diagnosed with autism spectrum disorder using the Autism Diagnostic Observation Schedule, Autism Diagnostic Observation Schedule, Second Edition, and Autism Diagnostic Interview-Revised was determined and mean domain scores compared using analysis of variance (ANOVA). The mean age was 64.4 (standard deviation = 21.6) months; the male:female ratio was 6:1. The diagnostic agreement of the Childhood Autism Rating Scale with the Autism Diagnostic Observation Schedule and Autism Diagnostic Observation Schedule, Second Edition was 100.0% and 98.0%, respectively. Agreement with the Autism Diagnostic Interview-Revised was 94.6%. Domain scores were highest for children with more severe symptoms (p < 0.01). Despite a high level of agreement of the Childhood Autism Rating Scale with the Autism Diagnostic Observation Schedule, Autism Diagnostic Observation Schedule, Second Edition, and Autism Diagnostic Interview-Revised, the Childhood Autism Rating Scale should be evaluated further with a broader range of autism spectrum disorder symptomatology, and by clinicians with varying experience before recommendation for use in low- and middle-income countries.

Status and Potential of Community-Engaged Research to Investigate Physical Activity Interventions for Children with Autism Spectrum Disorder in Chinese-American Communities