Exploratory data analysis of image texture and statistical features on myocardium and infarction areas in cardiac...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Patients with implanted cardioverter-defibrillators (ICDs) are stratified by arrhythmia risk.
• Cardiac magnetic resonance (CMR) imaging visualizes myocardial scars, a potential substrate for arrhythmias.

Purpose of the Study:

• To explore quantitative imaging features from CMR for predicting arrhythmia risk in ICD patients.
• To identify imaging biomarkers that differentiate low-risk from high-risk arrhythmia patients.

Main Methods:

• Manually segmented CMR images of myocardial scars were analyzed.
• Several hundred quantitative features (size, statistics, texture) were extracted.
• Maximum likelihood estimation-based Bayes classifiers were employed for risk stratification.

Main Results:

• Exploratory analysis identified potential discriminative features for arrhythmia risk.
• A hypothesis emerged regarding the significance of textural information in scarred and non-scarred regions.
• The study provides a basis for future hypothesis testing on specific imaging biomarkers.

Conclusions:

• Quantitative analysis of CMR images, particularly textural features, may predict arrhythmia risk in ICD patients.
• Further research is warranted to validate these imaging biomarkers for clinical risk stratification.
• Textural characteristics of myocardial scars could be crucial for understanding arrhythmia substrate.