An integrated multi-objective whale optimized support vector machine and local texture feature model for severity...

Area of Science:

• Cardiology
• Medical Imaging Analysis
• Machine Learning in Healthcare

Background:

• Left ventricular (LV) myocardium deterioration is linked to reduced ejection fraction (EF).
• Texture pattern analysis of LV myocardium is crucial for diagnosing heart muscle disease severity.
• Existing methods require improved prediction accuracy for disease severity levels.

Purpose of the Study:

• To develop a classification framework using co-occurrence of local ternary pattern (COALTP) features and a whale optimization algorithm.
• To enhance the prediction accuracy of heart muscle disease severity.
• To effectively discriminate between healthy subjects and those with reduced EF.

Main Methods:

• Segmentation of LV myocardium using optimized edge-based local Gaussian distribution energy (LGE)-based level set.
• Extraction of COALTP features at distance levels 1 and 2.
• Application of whale optimization for feature selection and a multi-classification framework using support vector machines (SVM).

Main Results:

• Optimized segmentation correlated with gold standard volume (R=0.98).
• Texture features from severe subjects at distance level 1 showed high statistical significance (p < 0.00004).
• Achieved an overall multi-class accuracy of 75% and subclass specificity of 70%.

Conclusions:

• The multi-objective whale optimized multi-class SVM framework effectively discriminates between healthy individuals and patients with reduced EF.
• This approach shows potential for supporting the heart disease treatment process.
• Texture analysis combined with optimization algorithms offers a promising avenue for cardiac disease diagnosis.