An acupuncture manipulation classification system based on three-axis attitude sensor and computer vision.

Area of Science:

• Integrative and Complementary Medicine
• Biomedical Engineering
• Acupuncture Research

Background:

• Acupuncture manipulation techniques are traditionally learned through observation and practice.
• Objective quantification of these techniques is challenging, hindering standardization and knowledge transfer.

Purpose of the Study:

• To develop a system for accurate identification and classification of acupuncture manipulations.
• To combine visual and sensor data for quantitative analysis of acupuncture techniques.

Main Methods:

• Utilized a three-axis attitude sensor to capture needle manipulation parameters (acceleration, velocity, angle).
• Employed computer vision and a hybrid 3D Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) model for dynamic gesture recognition.
• Integrated time-domain physical parameters with spatio-temporal features for classification.

Main Results:

• Achieved high classification accuracy for lifting-thrusting (95.56% reinforcing, 93.33% reducing) and twisting (95.56% reinforcing, 91.11% reducing) techniques.
• Demonstrated significant improvement in recognition accuracy compared to sensor-only methods.
• Identified distinct physical parameter patterns for different manipulation types.

Conclusions:

• The developed system enables quantitative analysis of acupuncture manipulation parameters.
• It provides a foundation for objective classification and digital inheritance of acupuncture techniques.
• This approach enhances the identification and classification accuracy of acupuncture manipulations.