Atherosclerotic risk stratification strategy for carotid arteries using texture-based features

Area of Science:

• Medical Imaging
• Cardiovascular Diagnostics
• Artificial Intelligence in Medicine

Background:

• Carotid artery plaques causing stenosis are a primary cause of stroke.
• Accurate patient selection for stenosis treatment is crucial due to associated risks.
• Distinguishing symptomatic from asymptomatic plaques is vital for stroke risk stratification.

Purpose of the Study:

• To develop and validate a computer-aided diagnostic (CAD) system for classifying carotid plaques.
• To differentiate between symptomatic and asymptomatic carotid plaques using ultrasound imaging.
• To assess the efficacy of texture-based features for plaque characterization.

Main Methods:

• Utilized ultrasound imaging of carotid plaque and intima-media thickness (IMT) regions.
• Employed semi-automatic and automatic segmentation techniques for data acquisition.
• Extracted texture features using Local Binary Patterns (LBP) and Law's Texture Energy (LTE).
• Applied machine learning classifiers including Support Vector Machine (SVM), K-Nearest Neighbor (KNN), and Radial Basis Probabilistic Neural Network (RBPNN).

Main Results:

• The Atheromatic system achieved 83% accuracy in classifying plaque regions using SVM.
• An accuracy of 89.5% was recorded for IMT regions using SVM, KNN, or RBPNN.
• Proposed AtheromaticPi and AtheromaticWi indices demonstrated significant separation between symptomatic and asymptomatic groups.

Conclusions:

• LBP/LTE-based texture analysis provides a noninvasive, fast, objective, and cost-effective method for carotid plaque characterization.
• The developed CAD system enhances existing carotid plaque diagnostics protocols.
• The proposed indices support the texture hypothesis for classifying plaque-related stroke risk.