Segmentation of Coronary Angiograms Using Gabor Filters and Boltzmann Univariate Marginal Distribution Algorithm

Area of Science:

• Medical Imaging
• Image Processing
• Biomedical Engineering

Background:

• Single-scale Gabor filters (SSG) are crucial for analyzing X-ray angiograms, particularly for detecting coronary arteries.
• Optimizing the three parameters of SSG is essential for enhancing detection performance and reducing computational load.
• Accurate vessel segmentation in angiograms is vital for diagnosing cardiovascular conditions.

Purpose of the Study:

• To present a novel method for improving the training of single-scale Gabor filters (SSG) using the Boltzmann univariate marginal distribution algorithm (BUMDA).
• To optimize SSG parameters for enhanced coronary artery detection and reduced computational time in X-ray angiograms.
• To achieve accurate vessel segmentation in X-ray angiograms.

Main Methods:

• The Boltzmann univariate marginal distribution algorithm (BUMDA) was employed to optimize the training of single-scale Gabor filters (SSG).
• The area under the receiver operating characteristic curve (A_ROC) was utilized as the fitness function for parameter selection.
• Interclass variance thresholding was adopted for classifying vessel and non-vessel pixels based on Gabor filter responses.

Main Results:

• The proposed method achieved a high detection rate with A_ROC = 0.9502 on a training set of 40 images.
• The method demonstrated excellent performance on a test set of 40 images, yielding A_ROC = 0.9583.
• Experimental results for vessel segmentation showed an accuracy of 0.944 on the test set of angiograms.

Conclusions:

• The BUMDA-based optimization method effectively improves the training of SSG for X-ray angiograms.
• The optimized SSG parameters lead to superior coronary artery detection rates and accurate vessel segmentation.
• This novel approach offers a promising tool for enhancing the analysis of cardiovascular images.