Artificial Intelligence Based Clustering Algorithm for Pulse Diagnosis.

Area of Science:

• Integrative Medicine
• Biomedical Engineering
• Artificial Intelligence

Background:

• Traditional pulse palpation, a subjective diagnostic technique, faces challenges in reliability and objectivity.
• Artificial intelligence (AI) offers potential for objective analysis of bio-signals.
• Developing AI-driven tools can enhance the scientific rigor of traditional diagnostic methods.

Purpose of the Study:

• To develop and validate an AI-based algorithm for clustering pulse waveform signals.
• To objectively assess diagnostic patterns identified by traditional medicine practitioners.
• To explore the potential of AI in standardizing traditional pulse diagnosis.

Main Methods:

• Collected pulse signals from both wrists of healthy individuals.
• Employed unsupervised clustering techniques on pulse waveform data.
• Utilized Dynamic Time Warping (DTW) for pulse similarity and Multidimensional Scaling (MDS) for dimensionality reduction.

Main Results:

• The AI algorithm demonstrated high alignment with expert diagnoses, clustering data-driven patterns effectively.
• The
• Floating-Sinking
• pulse pattern showed the highest Cosine similarity (0.83).
• Pulse signals from the left wrist exhibited slightly better alignment (0.56 ± 0.13) compared to the right (0.54 ± 0.15).

Conclusions:

• AI-driven pattern identification shows significant alignment with expert diagnoses in traditional pulse diagnosis.
• The developed algorithm offers a pathway for standardizing and quantifying subjective diagnostic techniques.
• Further research with diverse populations is recommended to refine AI diagnostic tools for broader clinical application.