Exploration on Bubble Entropy

Area of Science:

• Complex systems analysis
• Information theory
• Biomedical signal processing

Background:

• Traditional entropy estimators often suffer from high computational costs and parameter sensitivity.
• Existing methods require defining parameters like embedding dimension (m) and tolerance (r), impacting reliability.
• Bubble entropy emerges as a new metric addressing these limitations.

Purpose of the Study:

• Introduce and evaluate bubble entropy as an advanced entropy estimation method.
• Compare bubble entropy's performance against established entropy estimators.
• Assess bubble entropy's utility in clinical applications and machine learning.

Main Methods:

• Signal embedding into an m-dimensional space.
• Linear time computation for bubble entropy.
• Theoretical analysis and experimental validation using patient data (congestive heart failure vs. controls).
• Machine learning-based feature ranking.

Main Results:

• Bubble entropy demonstrates minimal dependence on parameters and linear time complexity.
• Theoretical analyses show significant advantages over existing methods.
• Experimental results show superior discrimination of congestive heart failure patients compared to controls.
• Bubble entropy proves to be a valuable feature source for AI decision-support systems.

Conclusions:

• Bubble entropy offers a computationally efficient and robust alternative to traditional entropy estimators.
• It shows significant potential in biomedical signal analysis, particularly for disease discrimination.
• Its effectiveness as a feature for AI enhances its applicability in clinical decision support.