Bidimensional ensemble entropy: Concepts and application to emphysema lung computerized tomography scans.

Area of Science:

• Image analysis
• Signal processing
• Biomedical engineering

Background:

• Traditional 1D and 2D entropy algorithms quantify image textures but require parameter tuning, affecting results.
• Ensemble entropy techniques enhance signal analysis stability and reduce bias but haven't been extended to 2D.

Purpose of the Study:

• To develop and evaluate novel 2D ensemble entropy algorithms for image texture analysis.
• To assess the performance of these algorithms on synthetic and biomedical data.

Main Methods:

• Proposed six 2D ensemble entropy algorithms: ensemble sample entropy, permutation entropy, dispersion entropy, distribution entropy, and two fuzzy entropy versions.
• Tested algorithms on synthetic images and a biomedical dataset for pulmonary emphysema classification.

Main Results:

• Ensemble techniques effectively detect image dynamics and randomness levels.
• Achieved high accuracy (92.7%) and sensitivity (88.4%) in classifying pulmonary emphysema using k-nearest neighbors.
• Demonstrated more stable entropy values with lower coefficients of variation on synthetic data.

Conclusions:

• 2D ensemble entropy algorithms represent a significant advancement for clinical deployment in medical imaging.
• These algorithms offer improved stability and reduced bias in image pattern analysis compared to existing methods.
• Potential applications span various imaging fields requiring robust texture quantification.