Radon descriptor-based machine learning using CT images to predict the fat tissue on left atrium in the heart

Area of Science:

• Cardiology
• Medical Imaging
• Machine Learning

Background:

• Heart disease is a leading cause of death globally.
• Increased epicardial fat in the left atrium is linked to Atrial Fibrillation (AF), a life-threatening condition.
• Early diagnosis and treatment of AF are crucial for managing patient outcomes.

Purpose of the Study:

• To develop and evaluate a machine learning model utilizing Radon descriptors for distinguishing between epicardial fat tissue and non-fat tissue in the left atrium.
• To establish a novel approach integrating the Radon transform framework with machine learning for cardiac image analysis.

Main Methods:

• CT images from eight patients were analyzed.
• Image patches were categorized into 'epicardial fat tissue' and 'nonfat tissue' groups.
• Ten feature vectors were extracted using Radon descriptors and processed by a machine learning model (KNN).

Main Results:

• The proposed methodology effectively discriminated between fat and non-fat tissues.
• The K-Nearest Neighbors (KNN) model demonstrated superior performance with 96.77% specificity, 98.28% sensitivity, and 97.50% accuracy.
• This represents the first application of Radon transform-based machine learning for differentiating left atrial fat tissue.

Conclusions:

• The Radon descriptor-based machine learning approach offers a promising tool for accurate identification of left atrial fat tissue.
• This method has the potential for application in advanced cardiac interventions and improving the diagnosis of conditions like AF.