Adaptive Fruitfly Based Modified Region Growing Algorithm for Cardiac Fat Segmentation Using Optimal Neural Network

Area of Science:

• Medical Imaging
• Computational Biology
• Cardiovascular Disease Research

Background:

• Epicardial adipose tissue (EAT) is strongly linked to coronary heart disease (CHD), posing diagnostic challenges due to its overlap with pericardial fat and influence from individual factors.
• Accurate segmentation and classification of cardiac fats are crucial for precise risk assessment and treatment strategies.
• Computed Tomography (CT) is a primary diagnostic tool, necessitating improved algorithmic approaches for fat analysis.

Purpose of the Study:

• To develop and validate a robust algorithm for accurate segmentation and classification of epicardial, pericardial, and mediastinal fats using CT images.
• To enhance the diagnostic capabilities for conditions associated with visceral fat accumulation around the heart.
• To improve the efficiency and accuracy of cardiac fat quantification in clinical practice.

Main Methods:

• Implementation of a modified region growing algorithm optimized with the Fruitfly Algorithm for precise fat segmentation in CT images.
• Extraction of Gray-Level Co-occurrence Matrix (GLCM) features from CT images for fat characterization.
• Development of a Grey Wolf Optimizer (GWO) based neural network for the classification of cardiac fats.

Main Results:

• The proposed methodology successfully segments epicardial, pericardial, and mediastinal fats with high accuracy, sensitivity, and specificity.
• Quantitative analysis demonstrated a clear distinction between different types of cardiac fats, outperforming existing methods.
• Performance metrics including accuracy, sensitivity, specificity, False Positive Rate (FPR), and False Negative Rate (FNR) were systematically evaluated and compared.

Conclusions:

• The Fruitfly Algorithm-based modified region growing approach offers a significant advancement in the accurate segmentation and classification of cardiac fats from CT scans.
• This computational technique holds promise for improving the early detection and management of cardiovascular diseases linked to adipose tissue distribution.
• The study advocates for the integration of advanced computational methods to enhance efficiency and precision in healthcare diagnostics.