Use of Oculomotor Behavior to Classify Children with Autism and Typical Development: A Novel Implementation of the Machine Learning Approach

Identification of autism spectrum disorder based on electroencephalography: A systematic review

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by difficulties in social communication and repetitive and stereotyped behaviors. According to the World Health Organization, about 1 in 100 children worldwide has autism. With the global prevalence of ASD, timely and accurate diagnosis has been essential in enhancing the intervention effectiveness for ASD children. Traditional ASD diagnostic methods rely on clinical observations and behavioral assessment, with the disadvantages of time-consuming and lack of objective biological indicators. Therefore, automated diagnostic methods based on machine learning and deep learning technologies have emerged and become significant since they can achieve more objective, efficient, and accurate ASD diagnosis. Electroencephalography (EEG) is an electrophysiological monitoring method that records changes in brain spontaneous potential activity, which is of great significance for identifying ASD children. By analyzing EEG data, it is possible to detect abnormal synchronous neuronal activity of ASD children. This paper gives a comprehensive review of the EEG-based ASD identification using traditional machine learning methods and deep learning approaches, including their merits and potential pitfalls. Additionally, it highlights the challenges and the opportunities ahead in search of more effective and efficient methods to automatically diagnose autism based on EEG signals, which aims to facilitate automated ASD identification.

[Machine learning algorithms for identifying autism spectrum disorder through eye-tracking in different intention videos]

OBJECTIVES

To investigate the differences in visual perception between children with autism spectrum disorder (ASD) and typically developing (TD) children when watching different intention videos, and to explore the feasibility of machine learning algorithms in objectively distinguishing between ASD children and TD children.

METHODS

A total of 58 children with ASD and 50 TD children were enrolled and were asked to watch the videos containing joint intention and non-joint intention, and the gaze duration and frequency in different areas of interest were used as original indicators to construct classifier-based models. The models were evaluated in terms of the indicators such as accuracy, sensitivity, and specificity.

RESULTS

When using eight common classifiers, including support vector machine, linear discriminant analysis, decision tree, random forest, and K-nearest neighbors (with K values of 1, 3, 5, and 7), based on the original feature indicators, the highest classification accuracy achieved was 81.90%. A feature reconstruction approach with a decision tree classifier was used to further improve the accuracy of classification, and then the model showed the accuracy of 91.43%, the specificity of 89.80%, and the sensitivity of 92.86%, with an area under the receiver operating characteristic curve of 0.909 (P<0.001).

CONCLUSIONS

The machine learning model based on eye-tracking data can accurately distinguish ASD children from TD children, which provides a scientific basis for developing rapid and objective ASD screening tools.

Oculomotor randomness is higher in autistic children and increases with the severity of symptoms

A variety of studies have suggested that at least some children with autism spectrum disorder (ASD) view the world differently. Differences in gaze patterns as measured by eye tracking have been demonstrated during visual exploration of images and natural viewing of movies with social content. Here we analyzed the temporal randomness of saccades and blinks during natural viewing of movies, inspired by a recent measure of "randomness" applied to micro-movements of the hand and head in ASD (Torres et al., 2013; Torres & Denisova, 2016). We analyzed a large eye-tracking dataset of 189 ASD and 41 typically developing (TD) children (1-11 years old) who watched three movie clips with social content, each repeated twice. We found that oculomotor measures of randomness, obtained from gamma parameters of inter-saccade intervals (ISI) and blink duration distributions, were significantly higher in the ASD group compared with the TD group and were correlated with the ADOS comparison score, reflecting increased "randomness" in more severe cases. Moreover, these measures of randomness decreased with age, as well as with higher cognitive scores in both groups and were consistent across repeated viewing of each movie clip. Highly "random" eye movements in ASD children could be associated with high "neural variability" or noise, poor sensory-motor control, or weak engagement with the movies. These findings could contribute to the future development of oculomotor biomarkers as part of an integrative diagnostic tool for ASD.