Exploring syndrome differentiation using non-negative matrix factorization and cluster analysis in patients with...

Area of Science:

• Informatics
• Traditional Medicine
• Dermatology

Background:

• Syndrome differentiation (SD) is crucial for Traditional Medicine (TM) diagnosis, relying on symptom clusters.
• Standardizing SD is essential for consistent TM treatment, particularly for conditions like atopic dermatitis (AD).

Purpose of the Study:

• To explore syndrome differentiation (SD) in atopic dermatitis (AD) patients using data analysis.
• To identify and standardize symptom clusters for AD using non-negative matrix factorization and k-means clustering.

Main Methods:

• Utilized non-negative matrix factorization and k-means clustering on a dataset of 73 AD patients.
• Collected 15 dermatologic and 18 systemic symptoms/signs for analysis.
• Employed permutation tests to identify significant cluster-specific symptoms.

Main Results:

• Identified five distinct patient clusters with a silhouette score of 0.484.
• Clustering revealed variations in symptom presentation and severity among patients.
• Found significant cluster-specific symptoms/signs, differentiating patient groups.

Conclusions:

• Informatics-driven SD can standardize diagnosis for atopic dermatitis (AD).
• Objective SD through data analysis supports consistent and reliable Traditional Medicine (TM) treatment.
• This approach moves SD from subjective observation to objective, data-driven conclusions.