CT texture analysis of vulnerable plaques on optical coherence tomography

Area of Science:

• Cardiovascular Imaging
• Radiology
• Artificial Intelligence in Medicine

Background:

• Thin-cap fibroatheroma (TCFA) represents a vulnerable plaque morphology associated with increased risk of cardiovascular events.
• Accurate identification of TCFA is crucial for risk stratification and guiding treatment strategies.
• Current imaging modalities have limitations in reliably detecting TCFA non-invasively.

Purpose of the Study:

• To evaluate the efficacy of CT texture analysis, specifically radiomics, in identifying TCFA.
• To compare the performance of CT texture analysis against conventional CT-derived plaque features and fat attenuation index (FAI) for TCFA detection.

Main Methods:

• Retrospective analysis of 33 patients with 43 lesions who underwent both CCTA and OCT.
• Extraction of 12 conventional CT plaque features, FAI, and 1691 radiomics features.
• Utilized Minimum Redundancy Maximum Relevance (mRMR) for feature selection and constructed a logistic radiomics model.
• Compared the diagnostic performance of the radiomics model with conventional HRP and FAI models.

Main Results:

• CT texture analysis identified 35 significantly different features between TCFA and non-TCFA lesions.
• The radiomics model achieved a significantly higher Area Under the Curve (AUC) of 0.952 for TCFA detection compared to HRP (0.621) and FAI (0.52) models.
• Low attenuation plaque (LAP) was more frequent in TCFA, but FAI showed no significant difference between groups.

Conclusions:

• CT texture analysis, leveraging radiomics, demonstrates superior performance in identifying TCFA compared to conventional CT parameters and FAI.
• Texture analysis offers a promising non-invasive tool for evaluating vulnerable atherosclerotic plaque.
• This technique could enhance the non-invasive assessment of plaque vulnerability in clinical practice.