CAP-Net: Carotid Artery Plaque Segmentation System Based on Computed Tomography Angiography

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiovascular Diagnostics

Background:

• Carotid plaque diagnosis from CT angiography (CTA) is crucial for stroke risk assessment but is often time-consuming and labor-intensive.
• Current diagnostic methods limit the scope of research and can lead to suboptimal patient outcomes.
• Automated analysis of CTA scans is needed to improve efficiency and accuracy.

Purpose of the Study:

• To develop and evaluate a deep learning-based model for the automated detection and segmentation of carotid plaques using CTA images.
• To enhance the diagnostic workflow for carotid artery disease.

Main Methods:

• A workflow utilizing three modified deep learning networks (U-Net, Attention U-Net, ConvNeXt-based U-Net) was developed for artery and plaque segmentation.
• The model was trained on CTA data from 1061 patients and validated using a five-fold cross-validation approach.
• Post-processing techniques were employed to refine predictions and minimize false positives.

Main Results:

• The automated system achieved high accuracy in artery segmentation (DSC of 0.91±0.04) and plaque segmentation (DSC of 0.75±0.14 per artery, 0.67±0.15 per patient).
• The model demonstrated strong plaque detection capabilities, identifying 95.5% of all plaques, including various types (calcified, mixed, soft).
• The system exhibited a low false positive rate, with an average of 0.63±0.93 false positives per patient.

Conclusions:

• A deep learning-based model, CAP-Net, was successfully developed for automatic carotid plaque detection and segmentation from CTA.
• The proposed model shows significant promise for improving the diagnosis and management of carotid artery disease.
• This AI-driven approach offers a more efficient and potentially more accurate method for analyzing carotid plaques.