A Video-Based Augmented Reality System for Human-in-the-Loop Muscle Strength Assessment of Juvenile Dermatomyositis

Insights

This study introduces an augmented reality system for assessing juvenile dermatomyositis (JDM) muscle strength. The system combines an action quality assessment algorithm with a virtual character for more accurate and efficient human evaluation.

Area of Science:

  • Biomedical Engineering
  • Computer Vision
  • Pediatric Rheumatology

Background:

  • Juvenile dermatomyositis (JDM) is a common childhood inflammatory myopathy causing muscle weakness and skin rashes.
  • Current muscle strength assessment methods like the Childhood Myositis Assessment Scale (CMAS) lack scalability and can be subjective.
  • Existing automatic action quality assessment (AQA) algorithms lack the necessary accuracy for critical biomedical applications.

Purpose of the Study:

  • To develop a scalable and objective system for assessing muscle strength in children with JDM.
  • To integrate an advanced AQA algorithm with augmented reality (AR) for human-in-the-loop verification.
  • To improve the accuracy and efficiency of JDM muscle strength assessment.

Main Methods:

  • Developed a contrastive regression AQA algorithm trained on a JDM dataset for muscle strength assessment.
  • Created a virtual character visualization using a 3D animation dataset to represent AQA results.
  • Implemented a video-based AR system that integrates computer vision for scene understanding and optimal virtual character augmentation.
  • Highlighted key visual elements within the AR system to facilitate human verification of AQA outputs.

Main Results:

  • The proposed AQA algorithm demonstrated effectiveness in assessing JDM muscle strength.
  • User studies confirmed that the AR system significantly improved the accuracy and speed of muscle strength assessment.
  • The system enables users to effectively compare real-world patient movements with virtual character representations.

Conclusions:

  • The developed video-based AR system offers a promising solution for objective and efficient muscle strength assessment in JDM.
  • Human-in-the-loop verification, facilitated by AR visualization, enhances the reliability of automated muscle strength evaluations.
  • This approach addresses the limitations of traditional assessment methods and current AQA technologies in pediatric rheumatology.