Heart motion abnormality detection via an information measure and Bayesian filtering

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Information Theory

Background:

• Functional cardiac images for heart wall motion analysis are prone to noise and segmentation errors.
• Accurate detection of heart wall motion abnormalities requires incorporating prior knowledge.
• Distinguishing normal from abnormal heart motion is challenging due to statistical similarities.

Purpose of the Study:

• To develop and evaluate an information-theoretic measure for detecting heart wall motion abnormalities.
• To improve the accuracy of heart wall motion analysis using recursive Bayesian filtering and Shannon's differential entropy (SDE).
• To explore alternative information-theoretic criteria, including Rényi entropy and Fisher information.

Main Methods:

• Utilized Kalman filtering, a recursive Bayesian filter, for estimating left ventricular (LV) cavity points from noisy and incomplete data.
• Developed a global abnormality detection measure based on Shannon's differential entropy (SDE).
• Analyzed wall motion quantitatively by constructing distributions of normalized radial distance estimates of the LV cavity.

Main Results:

• The proposed SDE criterion demonstrated significant improvement in detecting heart wall motion abnormalities.
• Performance was superior compared to conventional features like mean radial displacement and mean radial velocity.
• The method was validated using 269 segmented LV cavities from 30 subjects' cardiac MRI data.

Conclusions:

• Shannon's differential entropy provides a robust measure for heart wall motion abnormality detection.
• The proposed information-theoretic approach enhances diagnostic accuracy in cardiac imaging.
• This method offers a promising advancement for quantitative analysis of left ventricular function.