Automated Tracking and Quantification of Autistic Behavioral Symptoms Using Microsoft Kinect

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Autism spectrum disorder (ASD) prevalence is increasing, with stereotypical motor movements being a key symptom.
• Current methods for quantifying these movements are insufficient, hindering intervention and drug study assessments.
• Objective, automated quantification of motor symptoms in ASD is needed.

Purpose of the Study:

• To introduce a novel approach for automatically detecting and quantifying autistic body movements.
• To evaluate the effectiveness of the Microsoft Kinect v.2 and machine learning for this task.
• To provide a reliable tool for assessing ASD intervention outcomes.

Main Methods:

• Utilized Microsoft Kinect v.2 to capture skeletal data of actors performing stereotypical movements.
• Employed Visual Gesture Builder (VGB) with a machine learning approach for movement analysis.
• Developed complementary movement detection algorithms in Matlab.
• Validated VGB and Matlab results through manual grading.

Main Results:

• Both VGB and Matlab methods successfully detected stereotypical movements with high probability.
• The machine learning approach using VGB demonstrated the highest detection rates.
• The system proved effective in objectively quantifying complex autistic behaviors.

Conclusions:

• The Kinect v.2 combined with machine learning offers a viable solution for automated, objective quantification of autistic motor behaviors.
• This technology can significantly improve the assessment of ASD interventions and drug efficacy.
• Further development can enhance the accuracy and applicability of automated movement analysis in clinical settings.