[3D Pulse Image Detection and Pulse Pattern Recognition Based on Subtle Motion Magnification Technology].

Area of Science:

• Biomedical engineering
• Medical imaging
• Traditional Chinese Medicine (TCM)

Context:

• Existing 3D pulse signal detection methods in TCM suffer from reconstruction errors due to small out-of-plane membrane displacement.
• Accurate 3D pulse signal acquisition is crucial for objective TCM diagnosis.

Purpose:

• To propose and validate a novel 3D pulse image detection method incorporating subtle motion magnification technology.
• To improve the accuracy and reliability of 3D pulse signal reconstruction and subsequent pulse pattern recognition.

Summary:

• A binocular vision system captures pulse image signals, which are then processed using phase motion video magnification to amplify subtle movements.
• The amplified signals undergo 3D reconstruction, followed by feature extraction and selection using the Kolmogorov-Smirnov test.
• Machine learning algorithms (decision trees, random forests) are employed to classify five pulse patterns: deep, intermittent, flooding, slippery, and rapid.

Impact:

• The proposed method significantly enhances waveform clarity, amplitude stability, and periodic regularity of 3D pulse signals.
• Achieves an average accuracy of 96.29%±0.26% in pulse pattern recognition, outperforming traditional methods.
• Offers a more precise and objective approach to TCM pulse diagnosis, potentially improving diagnostic outcomes.