Automation of the Timed-Up-and-Go Test Using a Conventional Video Camera.

Area of Science:

• Biomedical Engineering
• Computer Vision
• Rehabilitation Technology

Background:

• The Timed-Up-and-Go (TUG) test is a standard clinical mobility assessment.
• Existing automated TUG systems often require sensors or specialized hardware like RGBD cameras.
• There is a need for accessible, accurate TUG automation methods.

Purpose of the Study:

• To develop and validate a novel video-based system for automating the TUG test.
• To assess the feasibility of using a regular RGB camera for TUG analysis.
• To eliminate the need for specialized hardware in TUG automation.

Main Methods:

• Utilized Mask Regional Convolutional Neural Network (R-CNN) and Deep Multitask Architecture for Human Sensing (DMHS) for 3D pose extraction from RGB video.
• Recorded 30 healthy participants performing TUG tests with both Kinect V2 and standard video cameras.
• Extracted transition timings between TUG sub-phases using heuristic features from 3D pose time series.

Main Results:

• The video-based vTUG system achieved comparable accuracy to Kinect-based systems for TUG transition points.
• The system demonstrated average errors of less than 0.15 seconds compared to hand-labeled ground truth.
• Successful extraction of global 3D poses and key movement phase transitions was achieved.

Conclusions:

• A novel, cost-effective method for automating the TUG test using a single standard camera has been demonstrated.
• This video-based approach removes the barrier of specialized equipment for TUG automation.
• The vTUG system facilitates enhanced clinical mobility assessment and data extraction.