Automatically detecting pain in video through facial action units

Area of Science:

• Biomedical Engineering
• Computer Vision
• Pain Medicine

Background:

• Clinical pain assessment relies on subjective patient self-reports or observer evaluations, which lack objectivity and precise timing.
• Objective pain measurement is crucial for accurate diagnosis and treatment monitoring in clinical settings.

Purpose of the Study:

• To develop and evaluate an automated system for objective pain detection using facial action units (AUs).
• To address the challenges of spontaneous emotion detection, including facial expressions and head movements, in pain assessment.

Main Methods:

• Utilized video data from patients with shoulder injuries.
• Developed an active appearance model (AAM)-based system for automatic pain detection on a frame-by-frame basis.
• Compared AAM performance against state-of-the-art methods using similarity-normalized appearance features.

Main Results:

• The AAM-based system successfully detected frames indicating patient pain.
• Demonstrated the AAM's capability to handle significant head and facial movements associated with pain.
• Achieved significant improvements in both AU and pain detection performance compared to existing methods.

Conclusions:

• Automated pain detection using AAMs offers an objective, frame-by-frame alternative to subjective pain reporting.
• The developed system effectively overcomes challenges posed by spontaneous expressions and patient movements.
• This approach shows promise for enhancing pain assessment accuracy in clinical environments.