Automatic Detection of ADHD and ASD from Expressive Behaviour in RGBD Data (1612.02374v1)

Abstract: Attention Deficit Hyperactivity Disorder (ADHD) and Autism Spectrum Disorder (ASD) are neurodevelopmental conditions which impact on a significant number of children and adults. Currently, the diagnosis of such disorders is done by experts who employ standard questionnaires and look for certain behavioural markers through manual observation. Such methods for their diagnosis are not only subjective, difficult to repeat, and costly but also extremely time consuming. In this work, we present a novel methodology to aid diagnostic predictions about the presence/absence of ADHD and ASD by automatic visual analysis of a person's behaviour. To do so, we conduct the questionnaires in a computer-mediated way while recording participants with modern RGBD (Colour+Depth) sensors. In contrast to previous automatic approaches which have focussed only detecting certain behavioural markers, our approach provides a fully automatic end-to-end system for directly predicting ADHD and ASD in adults. Using state of the art facial expression analysis based on Dynamic Deep Learning and 3D analysis of behaviour, we attain classification rates of 96% for Controls vs Condition (ADHD/ASD) group and 94% for Comorbid (ADHD+ASD) vs ASD only group. We show that our system is a potentially useful time saving contribution to the diagnostic field of ADHD and ASD.

• The paper proposes a novel method for automatically detecting ADHD and ASD by analyzing expressive behaviors (facial, head movements) captured using RGBD sensors and machine learning.
• Utilizing the KOMAA dataset, the system achieved high classification accuracy, notably 96% for distinguishing controls from those with a condition and 94% for differentiating comorbid cases from ASD only.
• This automated approach offers potential for more objective, efficient, and repeatable diagnostic procedures for ADHD and ASD, reducing human bias and clinical workload.

Automatic Detection of ADHD and ASD from Expressive Behaviour in RGBD Data: An Overview

The article proposes a novel approach to the automatic detection of Attention Deficit Hyperactivity Disorder (ADHD) and Autism Spectrum Disorder (ASD) through the analysis of expressive behaviors captured using RGBD (Color+Depth) sensors. This research addresses the existing challenges in neurodevelopmental disorder diagnostics, which are predominantly manual, subjective, costly, and time-consuming. The authors developed a method that leverages advancements in computer vision and machine learning to facilitate diagnostic procedures by automatically analyzing facial expressions and head movements.

The researchers utilized modern RGBD sensors, specifically the Kinect 2.0, to record participants during questionnaire sessions. The paper introduces the KOMAA dataset, a novel RGBD database, which comprises footage of 57 adult subjects diagnosed with ADHD, ASD, or neither. The diagnostic task was structured into two classification problems: differentiating controls from those with a condition (ADHD/ASD), and distinguishing comorbid subjects (ADHD+ASD) from those with ASD only. The system deployed achieved a remarkable classification accuracy of 96% for the former and 94% for the latter task.

The core methodology involves dynamically tracking and analyzing facial action units (AUs), head pose, and head motion speed using Dynamic Deep Learning models. High-level feature descriptors were computed and categorized into six primary sets, including Facial Action Units and Kinect Animation Units, which were then utilized to train Support Vector Machine (SVM) classifiers. The predictive models demonstrated robust capabilities in classifying individuals with significant precision, further enhanced by employing a forward feature selection strategy to combat high-dimensionality challenges.

Numerical results, notably the high classification rates, underscore the practical utility of this automatic diagnostic tool. The paper substantiates the hypothesis that expressive behavioral cues are indicative of neurodevelopmental conditions, proposing that these cues can be systematically identified without the constraints of human observation. Despite the apparent success, the dataset's limitation due to the small sample size of ADHD-only participants represents an area for future expansion.

The implications of this research are multifaceted. Practically, it promises a more efficient, objective, and repeatable diagnostic process for ADHD and ASD, potentially transforming clinical practices by minimizing human error and decision-making biases and reducing clinical workloads. Theoretically, it paves the way for enhanced understanding of behavioral markers associated with these disorders, fostering further interdisciplinary research at the intersection of computer vision, psychology, and psychiatry.

Furthermore, the developments illuminated by this paper could inspire future research in healthcare applications for machine learning. The potential to refine diagnostic algorithms, expand datasets, and continually improve accuracy through ongoing technological advances foresees a future where AI systems complement and support human expertise in mental health diagnostics.

In conclusion, the approach presented effectively marries advanced computer vision techniques with practical clinical needs, aiming to redefine the standards for diagnosing ADHD and ASD. As research continues in this vein, the expectation is for increasingly reliable, comprehensive systems capable of broader applications across diverse behavioral health scenarios.