top of page

Research

The development of social, cognitive, motor, and language abilities in children is remarkably heterogeneous. Children exhibit different developmental trajectories with some advancing very quickly and others achieving developmental milestones slowly and in unexpected order. When development is altered, such as in autism, different children can exhibit a wide variety of distinct behavioral symptoms and challenges. Heterogeneity is also apparent in the genetics and neurophysiology of autism, which are highly diverse.

 

This heterogeneity creates a huge challenge for research, because it makes it difficult to identify developmental conditions (such as autism) and their underlying biological mechanisms. Moreover, existing techniques for measuring behavioral skills and difficulties in children rely on indirect observations made by care-givers or clinicians. These measures are subjective, limited in their sensitivity, and biased in different ways. For example, there is often little agreement across measures collected from different reporters (e.g., parent questionnaires and clinical assessments).

 

Hence, advancing research requires the development of direct, objective, and reliable techniques for measuring relevant behaviors, skills, and abilities in standardized ways. These techniques need to be scalable for use in longitudinal studies with large cohorts that will enable the field to establish population, age dependent norms.

​​​

With this in mind, we are developing multiple AI-enabled digital phenotyping techniques to measure the development of social skills, language abilities, motor abilities, anxiety, physical activity, and sleep. We test these techniques within the context of autism research where we compare behaviors and abilities across typically and atypically developing children in multiple age groups.

​

Specific research domains in the lab include:​

Movement tracking using Kinect
Movement tracking using OpenPose

Movement analysis

Many children with autism move differently. This research focuses on three types of movement analyses that utilize computer vision techniques to analyze video recordings from ADOS assessments:

  1. Social exploration - in this project we study how the child moves in the assessment room using measures such as the distance between child and clinician, the amount of approaches towards the clinician, the amount of time facing the clinician, and more (see relevant paper).

  2. Stereotypical movements - in this project we are training deep learning algorithms to identify video segments where children exhibit stereotypical movements, thereby enabling us to quantify their severity in individual children (see relevant paper).

  3. Facial expressions - social communication is heavily reliant on proper use of facial expressions, a social skill that develops early in life. In this project we are measuring movements of facial muscle groups (action units) and analyzing their kinematics to identify difficulties in the expression of emotions.

Diagnosis and heterogeneous developmental trajectories

Children with autism are diagnosed at different ages and develop differently over time. Some demonstrate remarkable improvements while others deteriorate. Our research attempts to identify biological, environmental, and intervention factors that impact this heterogeneity of outcomes. We examine development in specific behavioral domains (e.g., social skills or language) and identify factors that are associated with successful progress including:​

  1. Age of diagnosis - We are examining how  differences in the age of diagnosis are associated with short and long term outcomes of autistic children (relevant paper).

  2. Educational settings - Availability of structured early intervention programs differs dramatically across special education and mainstream preschool settings. We are examining the impact of these settings on short and long term outcomes of autistic children (relevant paper).​

  3. Relationship across behavioral domains - Behavioral abilities and difficulties in different domains are likely to be associated, thereby explaining some of the heterogeneity in autism profiles. We are studying, for example, how sleep disturbances are associated with sensory sensitivities throughout early development (relevant paper) and how language regression is associated with early motor development (relevant paper).​

Clinical assessment
Autism education
EEG sleep lab

Sleep studies

At least 50% of autistic children have insomnia (i.e., severe sleep disturbances), which is highly debilitating for both the children and their families. We are characterizing sleep behaviors and physiology to better understand these disturbances, reveal their potential causes, and quantify their impact on development of social, language, and motor skills:

  1. Polysomnography - we are performing classical sleep lab studies and have demonstrated that some children with autism have reduced sleep pressure, which may explain their difficulty to fall asleep (relevant paper).

  2. Sensors - we are using actigraphy and consumer devices such as Fitbit to study sleep at home over extended periods of time.

  3. Impact - we are studying how sleep disturbances generate aberrant behaviors including irritability, tantrums, hyperactivity, and other symptoms (relevant paper).

  4. Behavioral intervention - we are testing the effectiveness of parent training in sleep hygiene and intervention techniques that reduce the dependency of children's sleep on parental presence. 

Eye tracking

The development of social skills requires attention to social information from an early age. However, some children with autism observe the world differently. This is apparent, for example, in the tendency of some children to avoid eye contact with others. Over the last several years we have acquired >700 eye tracking sessions from children with and without autism as they observe different movies of social interactions. We are studying:

  1. How social attention differs in children with autism? (relevant paper).

  2. Are early gaze preference differences in autism due to differences in basic oculomotor control? (relevant paper) or regulation of pupil diameter (relevant paper)? 

  3. Can we predict the social skills of an individual child from their eye gaze patterns? 

Eye tracking lab
White matter DTI
Anatomical MRI

Structural and functional MRI
We have an extensive track record of studying brain structure and function in autism using a variety of imaging techniques. Studies include:

  1. Assessment ​of differences in brain volume, cortical thickness, and cortical surface area across individuals with and without autism (relevant paper) as well as comparisons of early head circumference development (relevant paper).

  2. Assessment of mirror system fMRI responses in adults with autism as they observe and perform movements including an analysis of movement-selectivity (relevant paper).

  3. Demonstration that some adults with autism have unreliable brain responses to sensory stimuli across multiple modalities (relevant paper).

  4. Demonstration that inter-hemispheric functional connectivity (i.e., neural synchronization) is weaker in some children with autism (relevant paper). The same children also exhibit DTI differences in corpus callosum white matter fibers indicative of poor inter-hemispheric structural connectivity (relevant paper).

  5. In newer projects we have revealed that some children with autism exhibit excessive extra-axial CSF also known as benign external hydrocephalus during early periods of development, before the age of 2-years-old (relevant paper).

Azrieli National Centre for Autism and Neurodevelopment
Dinstein lab, Psychology department, BGU
Ben Gurion University of the Negev
bottom of page