A cross-dataset study on automatic detection of autism spectrum disorder from text data.

OBJECTIVE: The goals of this article are as follows. First, to investigate the possibility of detecting autism spectrum disorder (ASD) from text data using the latest generation of machine learning tools. Second, to compare model performance on two datasets of transcribed statements, collected using two different diagnostic tools. Third, to investigate the feasibility of knowledge transfer between models trained on both datasets and check if data augmentation can help alleviate the problem of a small number of observations. METHOD: We explore two techniques to detect ASD. The first one is based on fine-tuning HerBERT, a BERT-based, monolingual deep transformer neural network. The second one uses the newest, multipurpose text embeddings from OpenAI and a classifier. We apply the methods to two separate datasets of transcribed statements, collected using two different diagnostic tools: thought, language, and communication (TLC) and autism diagnosis observation schedule-2 (ADOS-2). We conducted several cross-dataset experiments in both a zero-shot setting and a setting where models are pretrained on one dataset and then training continues on another to test the possibility of knowledge transfer. RESULTS: Unlike previous studies, the models we tested obtained average results on ADOS-2 data but reached very good performance of the models in TLC. We did not observe any benefits from knowledge transfer between datasets. We observed relatively poor performance of models trained on augmented data and hypothesize that data augmentation by back translation obfuscates autism-specific signals. CONCLUSION: The quality of machine learning models that detect ASD from text data is improving, but model results are dependent on the type of input data or diagnostic tool.

Detecting autism from picture book narratives using deep neural utterance embeddings.

BACKGROUND: Deficits in the ability to use language in social contexts, including storytelling skills, are observed across the autism spectrum. Development in machine-learning approaches may contribute to clinical psychology and psychiatry, given its potential to support decisions concerning the diagnosis and treatment of psychiatric conditions and disorders. AIMS: To evaluate the usefulness of deep neural networks for detecting autism spectrum disorder (ASD) from textual utterances, specifically from narrations produced by individuals with ASD. METHODS & PROCEDURES: We examined two text encoders: Embeddings from Language Models (ELMo) and Universal Sentence Encoder (USE), and three classification algorithms: XGBoost, support vector machines, and dense neural network layer. We aimed to classify 25 participants with ASD and 25 participants with typical development (TD) based on their narrations produced during the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) picture book task. The results of computational approaches were compared with the results of standardized testing and classifications made by two psychiatrists (raters). The raters were asked to read utterances produced by a participant (without an examiner's statements and additional information) and assign a participant to one of the two groups: ASD or with typical development (TD). OUTCOMES & RESULTS: The computer-based models had higher sensitivity, specificity, positive predictive values and negative predictive values than the raters, and lower than the two standardized instruments: ADOS-2 and Social Communication Questionnaire (SCQ). CONCLUSIONS & IMPLICATIONS: Our findings lay the groundwork for future studies involving deep neural network-based text representation models as tools for augmenting the ASD diagnosis or screening. Both ELMo and USE text encoders provided promising specificities, sensitivities, positive predictive values and negative predictive values. Our results indicate the usefulness of page-level embeddings for utterance representation in ADOS-2 picture book task. WHAT THIS PAPER ADDS: What is already known on this subject Deficits in the use of language in social contexts, and narrative ability in particular, are observed across the autism spectrum. Most research on narrative skills has applied hand-coding methods. Hitherto, machine-learning methods were used mostly for image recognition problems and data from screening questionnaires for ASD classification. Detection of mental and developmental disorders from textual input is an emerging field for machine and deep-learning methods. What this paper adds to existing knowledge This study explored the ability of several types of deep neural network-based text representation models to detect ASD. Both ELMo and USE provided the most promising values of specificity, sensitivity, positive predictive values and negative predictive values. What are the potential or actual clinical implications of this work? Competitive accuracy, repeatability, speed and ease of operation are all advantages of computerized methods. They allow for objective and quantitative assessment of narrative ability and complex language skills. Deep neural network-based text representation models could in the future support clinicians and augment the decision-making process related to ASD diagnosis, screening and intervention planning.

Cross-Cultural Validation of the Polish Version of the ADI-R, Including New Algorithms for Toddlers and Young Preschoolers.

Autism Diagnostic Interview-Revised (ADI-R) is one of the most widely used standardized diagnostic instruments for autism spectrum disorder (ASD). This article presents findings from the validation of the Polish version of the Autism Diagnostic Interview-Revised (ADI-R-PL), including new algorithms for toddlers and preschoolers. The validation group consisted of 125 participants: 65 with Autism Spectrum Disorder (ASD group) and 60 in the control group, including individuals with non-ASD disorders and typical development. The normalization group consisted of 178 participants, including 118 with ASD. The ADI-R-PL was found to have good psychometric properties. Confirmatory factor analysis supported both a bifactor structure and three-factor model. The study has generated preliminary information about the psychometric properties of the new algorithms for toddlers and young preschoolers. To the best of our knowledge, this paper is the first to propose new cutoffs in three ADI-R domains for a non-English-speaking population.

Association between FKBP5 Functional Polymorphisms and Completed Suicide.

OBJECTIVES: Dysfunction of the hypothalamic-pituitary-adrenal (HPA) axis leads to impaired stress response. FK506-binding protein 51 (FKBP5), which influences HPA axis activity via glucocorticoid receptors, is supposed to play an important role in the regulation of negative feedback and glucocorticoid resistance. Since ineffective stress response mechanisms are considered as a biological background of suicide behavior, we aimed to analyze a possible association between FKBP5 functional polymorphisms and completed suicide. METHODS: The selected FKBP5 polymorphisms rs1360780 and rs3800373 were genotyped in a sample of 563 suicide victims and 475 controls. RESULTS: A significant association between the high-induction rs3800373 C allele and completed suicide was detected (OR = 1.36, p = 0.007). In this polymorphism, genotype distribution supported a codominant model of inheritance. The analyzed SNPs were in strong linkage disequilibrium (D' = 0.916 and r2 = 0.826) with the rs1360780 (T)-rs3800373 (C) haplotype apparently responsible for the observed association (OR = 1.34, p = 0.010). CONCLUSION: The results of the present study indicate that genetic alterations in FKBP5 may influence vulnerability to suicide.

[Autism spectrum disorders - epidemiology, symptoms, comorbidity and diagnosis]. / Zaburzenia ze spektrum autyzmu - epidemiologia, objawy, wspólzachorowalnosc i rozpoznawanie.

In the new classification of American Psychiatric Association - DSM-5 - a category of autistic spectrum disorders (ASD) was introduced, which replaced autistic disorder, Asperger syndrome, childhood disintegrative disorder and pervasive developmental disorder not otherwise specified. ASD are defined by two basic psychopathological dimensions: communication disturbances and stereotyped behaviors, and the diagnosis is complemented with the assessment of language development and intellectual level. In successive epidemiological studies conducted in 21 century the prevalence of ASD has been rising, and currently is estimated at 1% in general population. The lifetime psychiatric comorbidity is observed in majority of patients. The most common coexisting diagnoses comprise disorders ofanxiety-affective spectrum, and in about 1/3 of patients attention deficit/ hyperactivity disorders could be diagnosed. Prodromal symptoms of ASD may emerge before 12 months of life, however reliability of diagnosis at such an early age is poor. Several screening instruments, based on the parental and/or healthcare professional assessments may be helpful in ASD detection. However, structured interviews and observation schedules remain the gold standard of diagnosis.

Analysis of Autistic Adolescents' Essays Using Computer Techniques.

PURPOSE: Challenges associated with narrative discourse remain consistently observable across the entire spectrum of autism. We analyzed written narratives by autistic and non-autistic adolescents and aimed to investigate narrative writing using quantitative computational methods. METHODS: We employed Natural Language Processing techniques to compare 333 essays from students in the final eighth grade of primary school: 195 written by autistic and 138 by non-autistic participants. RESULTS: Autistic students used words with a positive emotional polarity statistically less frequently (p < .001), and their stories were less abstract (p < .001) than those written by peers from the non-autistic group. However, autistic adolescents wrote more complex stories in terms of readability than participants from the non-autistic group (p < .001). The writing competencies assessed by teachers did not differ significantly between the two groups. CONCLUSION: Findings suggest that written narratives by autistic individuals may exhibit characteristics similar to those detected by computational methods in spoken narratives. Collecting data from national exams and its potential usefulness in distinguishing autistic individuals could pave the way for future large-scale and cost-effective epidemiological studies on autism.

Analysis of four genes involved in the neurodevelopment shows association of rs4307059 polymorphism in the cadherin 9/10 region with completed suicide.

BACKGROUND: We hypothesized that DNA variants affecting neurodevelopment such as rs4307059 (CDH10/CDH9), rs930752 (NRXN1), rs6265 (BDNF) or rs10868235 (NTRK2) may predispose to completed suicide. METHODOLOGY: We used a case-control two-stage approach based on a discovery cohort (557 cases and ∼550 controls) and replication cohort (159 cases and 186 controls). The suicides were ascertained as consecutive cases autopsied at the Department of Forensic Medicine, Medical University of Warsaw, Poland. RESULTS: In the discovery cohort we found an association between suicide and the CC genotype in the rs4307059 polymorphism (OR 1.64, p = 0.012). The trend for an overrepresentation of the CC homozygotes among suicides was replicated in the second cohort (OR 1.97, p = 0.056). Analysis in the pooled cohorts showed that rs4307059 CC was associated with completed suicide (OR 1.71, p = 0.002) also after Bonferroni correction (p(cor.) = 0.024). In an exploratory search for genotype-phenotype correlation we found that males with the rs4307059 CC genotype committed suicide earlier than those with CT/TT genotypes (p = 0.049). CONCLUSIONS: The CC genotype of rs4307059 located in the region between CDH9 and CDH10 is associated with completed suicide in a Polish cohort.

[Polish version of the ADI-R (Autism Diagnostic Interview-Revised)]. / Polska wersja wywiadu do diagnozowania autyzmu ADI-R (Autism Diagnostic Interview-Revised).

Childhood autism belongs to pervasive developmental disorders and is characterised by qualitative abnormalities in reciprocal social interactions, communication, and by restricted, repetitive interests and behaviours. Until now there was no standardised tool for a diagnosis of autism in Poland. The paper presents the Polish version of the Autism Diagnostic Interview - Revised (ADI-R), which is the "gold standard" for the diagnosis of autism in Europe, United States and Australia. It describes the translation process and adaptation of the original version into Polish, as well as differences between the two versions. ADI-R is a complex, standardised, semi-structured investigator-based interview for parent or caregiver of person with autism, linked to ICD-10 and DSM-IV-TR criteria and appropriate for both adults and children, who have the minimum mental age of 24 months. Moreover ADI-R consists of, beside diagnostic algorithms, the current behaviour algorithms, which enable to assess and compare various levels of functioning during planning and implementation of treatment and therapy. ADI-R is also a very useful tool in the diagnosis for scientific purposes due to its standardisation.

[Polish version of the ADOS (autism diagnostic observation schedule-generic)]. / Polska wersja narz dzia obserwacyjnego do diagnozowania autyzmu ADOS.

The article presents the Polish version of the autism diagnostic observation schedule-generic (ADOS), which together with the autism diagnostic interview-revised (ADI-R) is cited as the "gold standard" for the diagnosis of autism. The ADOS is a standardised, semistructured observation protocol appropriate for children and adults of differing age and language levels. It is linked to ICD-10 and DSM-IV-TR criteria. The ADOS consists of four modules, ranging from module 1 for nonverbal individuals to module 4 for verbally fluent adults. The adequate inter-rater reliability for items has been established. The protocol has high discriminant validity and distinguishes children with pervasive developmental disorders from children, who are outside of the spectrum. Although it does not enable to distinguish individuals with pervasive developmental disorder, unspecified from individuals with childhood autism. The paper presents subsequent steps of the translation process of the original version into Polish, as well as a chosen adaptation strategy of the Polish version. The ADOS is a very useful tool both for clinical diagnosis and for the scientific purpose diagnosis. In this last case it is extremely important to use a standardised method. Until now, there was no standardised diagnostic tool for autism in Poland.

Social language in autism spectrum disorder: A computational analysis of sentiment and linguistic abstraction.

Individuals with autism spectrum disorder (ASD) demonstrate impairments with pragmatic (social) language, including narrative skills and conversational abilities. We aimed to quantitatively characterize narrative performance in ASD using natural language processing techniques: sentiment and language abstraction analyses based on the Linguistic Category Model. Individuals with ASD and with typical development matched for age, gender, ethnicity, and verbal and nonverbal intelligence quotients produced language samples during two standardized tasks from the Autism Diagnostic Observation Schedule, Second Edition assessment: Telling a Story from a Book and Description of a Picture. Only the narratives produced during the Book Task differed between ASD and control groups in terms of emotional polarity and language abstraction. Participants with typical development used words with positive sentiment more often in comparison to individuals with ASD. In the case of words with negative sentiment, the differences were marginally significant (participants with typical development used words with negative sentiment more often). The Book Task narratives of individuals with ASD were also characterized by a lower level of language abstraction than narratives of peers with typical development. Linguistic abstraction was strongly positively correlated with a higher number of words with emotional polarity. Neither linguistic abstraction nor emotional polarity correlated with participants' age or verbal and nonverbal IQ. The results support the promise of sentiment and language abstraction analyses as a useful tool for the quantitative, fully automated assessment of narrative abilities among individuals with ASD.

Adaptation and psychometric properties of the Polish version of the Short Sensory Profile 2.

The article presents findings from the validation of the Polish version of the Short Sensory Profile, 2nd edition (SSP-2-PL).A total of 1230 participants were recruited: 310 diagnosed with autism spectrum disorder (ASD), 264 with nonspectrum neurodevelopmental disorders, and 656 typically developing (TD). The reliability and validity of the questionnaire were estimated using several methods, including internal consistency, test-retest, and factor analysis.Exploratory factor analysis identified a unidimensional solution in both the TD and ASD groups. The structure of SSP-2 seems to be homogeneous; therefore, the findings support the validity of calculating the SSP-2 overall score. Cronbach alphas and intraclass correlation coefficients exceeded 0.90 for overall total in all study groups. The Social Communication Questionnaire total score correlated moderately with SSP-2 scores. A 1-way analysis of variance yielded statistically significant differences at P < .001 between groups on all scales/quadrants and the overall score. Our results indicate greater severity of sensory processing problems among children with ASD and non-ASD disorders than among TD peers. Among children with ASD, 85% experienced problems with sensory processing. Scores in SSP-2-PL were not affected by the children's age, gender, informant, and informant's level of education.To the best of our knowledge, this is the 1st study on non-English participants using a revised version of the SSP-2. The results confirm the prevalence of sensory processing problems among children with neurodevelopmental disorders, especially with ASD. SSP-2-PL has high reliability in terms of both internal consistency and stability of scores. The results suggest that SSP-2 overall score could be used for screening purposes, namely to identify sensory processing and behavioral problems combined into one factor. Further analyzes of the SSP-2 factor structure are needed to confirm the findings of the present study.

Adaptation and Validation of the ADOS-2, Polish Version.

Autism Diagnostic Observation Schedule (ADOS) is one of the most popular instruments used world-widely in the diagnosis of autism spectrum disorders (ASD). Unfortunately, there are only a few studies of the psychometric properties of non-English language versions of this instrument and none of the adaptation of its second edition (ADOS-2). The objective of this study was to verify the psychometric properties of the Polish version of the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2-PL). The authors recruited 401 participants: 193 with ASDs (ASD group) and 78 with non-spectrum disorders, plus 130 typically developing participants (control group). ADOS-2-PL was found to have high interrater reliability, internal consistency and test-retest reliability. Confirmatory factor analysis confirmed a good fit of the Polish data to the two-factor model of ADOS-2. As no significant differences were found between participants with childhood autism and other ASDs, only one cut-off was established for Modules 1-4. The sensitivity, specificity and positive predictive value of ADOS-2-PL are high: sensitivity was over 90% (only for the "Older with some words" algorithm in the Toddler Module the sensitivity was 71% and "Aged 5 years or older" algorithm in Module 2 sensitivity was 84%), specificity was above 80% (with the exception of the Module 4 and Module 2 "Aged 5 years or older" algorithm where it was above 70%). The results support the use of ADOS-2-PL in clinical practice and scientific research. To the best of our knowledge, there have been no reports to date about adaptations of ADOS-2 and the psychometric properties of non-English language versions. As such, this constitutes the first attempt at adapting ADOS-2, and its results could be of interest for researchers outside of Poland.

The role of genetic factors and pre- and perinatal influences in the etiology of autism spectrum disorders - indications for genetic referral. / Znaczenie czynników genetycznych oraz przed- i okoloporodowych w etiologii zaburzen ze spektrum autyzmu - wskazania do konsultacji genetycznej.

Autism spectrum disorders (ASD) are caused by disruptions in early stages of central nervous system development and are usually diagnosed in first years of life. Despite common features such as impairment of socio-communicative development and stereotypical behaviours, ASD are characterised by heterogeneous course and clinical picture. The most important aetiological factors comprise genetic and environmental influences acting at prenatal, perinatal and neonatal period. The role of rare variants with large effect i.e. copy number variants in genes regulating synapse formation and intrasynaptic connections is emphasised. Common variants with small effect may also be involved, i.e. polymorphisms in genes encoding prosocial peptides system - oxytocin and vasopressin. The environmental factors may include harmful effects acting during pregnancy and labour, however their specificity until now is not confirmed, and in some of them a primary genetic origin cannot be excluded. In several instances, especially with comorbid disorders - intellectual disability, epilepsy and dysmorphias - a detailed molecular diagnostics is warranted, which currently may elucidate the genetic background of disorder in about 20% of cases.

Neural correlates of own name and own face detection in autism spectrum disorder.

Autism spectrum disorder (ASD) is a heterogeneous neurodevelopmental condition clinically characterized by social interaction and communication difficulties. To date, the majority of research efforts have focused on brain mechanisms underlying the deficits in interpersonal social cognition associated with ASD. Recent empirical and theoretical work has begun to reveal evidence for a reduced or even absent self-preference effect in patients with ASD. One may hypothesize that this is related to the impaired attentional processing of self-referential stimuli. The aim of our study was to test this hypothesis. We investigated the neural correlates of face and name detection in ASD. Four categories of face/name stimuli were used: own, close-other, famous, and unknown. Event-related potentials were recorded from 62 electrodes in 23 subjects with ASD and 23 matched control subjects. P100, N170, and P300 components were analyzed. The control group clearly showed a significant self-preference effect: higher P300 amplitude to the presentation of own face and own name than to the close-other, famous, and unknown categories, indicating preferential attentional engagement in processing of self-related information. In contrast, detection of both own and close-other's face and name in the ASD group was associated with enhanced P300, suggesting similar attention allocation for self and close-other related information. These findings suggest that attention allocation in the ASD group is modulated by the personal significance factor, and that the self-preference effect is absent if self is compared to close-other. These effects are similar for physical and non-physical aspects of the autistic self. In addition, lateralization of face and name processing is attenuated in ASD, suggesting atypical brain organization.

Inverse association between obesity predisposing FTO genotype and completed suicide.

The A allele of rs9939609 in the FTO gene predisposes to increased body mass index (BMI) and obesity. Recently we showed an inverse association between the obesity related A allele of rs9939609 and alcohol dependence which was replicated by others. Since this finding raises a possibility that FTO may be associated with other psychiatric phenotypes, we aimed to examine association of rs9939609 with completed suicide. We genotyped rs9939609 in 912 suicide victims and 733 controls using TaqMan approach. We observed an inverse association between suicide and the rs9939609 A allele (OR = 0.80, P = 0.002, Pcor = 0.006) with genotype distribution suggesting a co-dominant effect. Given the link between alcoholism and suicide under influence of alcohol reported in Polish population, confounding by alcohol addiction was unlikely due to apparently similar effect size among cases who were under influence of ethanol at the time of death (OR = 0.76, P = 0.003, N = 361) and those who were not (OR = 0.80, P = 0.007, N = 469). The search for genotype-phenotype correlations did not show significant results. In conclusion, our study proves that there is an inverse association between rs9939609 polymorphism in FTO gene and completed suicide which is independent from association between FTO and alcohol addiction.

Possible association between suicide committed under influence of ethanol and a variant in the AUTS2 gene.

BACKGROUND: rs6943555 in AUTS2 has been shown to modulate ethanol consumption. We hypothesized that rs6943555 might be associated with completed suicide. METHODS: We genotyped rs6943555 in 625 completed suicides and 3861 controls using real-time TaqMan Allelic Discrimination Assay. All individuals were Polish Caucasians. RESULTS: We detected an association between suicide and rs6943555 A allele (OR = 1.17, P = 0.018 for allelic comparison, OR = 1.24, P = 0.013 for dominant, and OR = 1.18, P = 0.020 for co-dominant model of inheritance). The association remained significant after adjusting for age and gender (co-dominant: P = 0.002 and dominant model: P = 0.001). After stratifying suicides according to blood ethanol concentration (BAC≤ 20 mg/dl and BAC > 20 mg/dl) the association remained significant only for cases who committed suicide under influence of alcohol (co-dominant: OR  =  1.37, P = 0.004 and dominant model: OR = 1.45, P = 0.006). To validate this finding we genotyped another cohort of 132 cases. We reproduced the association between rs6943555 A allele and suicide under influence of ethanol (allelic comparison: OR = 1.55, P = 0.023; co-dominant : OR = 1.54, P = 0.031; dominant model: OR = 1.84, P = 0.015). Analyzing pooled suicides with BAC >20 mg/dl (N = 300) we found the association of rs6943555 A allele not only vs. controls (allelic OR = 1.41, P = 0.00029) but also vs. cases with BAC ≤ 20 mg/dl (N = 449, allelic OR = 1.33, P = 0.019). CONCLUSIONS: In our study rs6943555 A allele is associated with suicide committed after drinking ethanol shortly before death. The rs6943555 A allele may be linked to adverse emotional reaction to ethanol, which could explain the association with lower consumption in general population as well as the predisposition to suicide under influence of ethanol.

No association between MTHFR C677T polymorphism and completed suicide.

MTHFR C677T polymorphism (rs1801133) was associated with numerous psychiatric conditions but no prior study investigated whether it predisposes to completed suicide. We typed rs1801133 in 692 suicide victims and 3257 controls representative of a Polish adult population (the WOBASZ cohort). Although we had a power of 0.8 to detect (at alpha 0.05) an allelic OR=1.19, we did not find significant difference among suicides vs. controls in the prevalence of the MTHFR 677T allele (OR=1.02, p=0.759) or the TT genotype (OR=1.01, p=0.926). Since among controls we found an association between TT and depression defined by Beck Depression Inventory (BDI, OR=1.61, p=0.049) we also compared suicides with controls without signs of depression (BDI ≤ 11) but found no association (OR=1.0, p=0.976). Analyses within suicides showed trends (not significant after Bonferroni correction) for correlations between the dose of the T allele and age at death among males and blood ethanol concentration among females, who committed suicide under the influence of alcohol. We conclude that MTHFR C677T polymorphism is not a risk factor for completed suicide. The sex-specific trends for correlations between rs1801133 and age at death, and blood ethanol concentration should be studied further.