Deep Learning for Head and Neck CT Angiography: Stenosis and Plaque Classification.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiovascular Imaging

Background:

• Head and neck CT angiography interpretation for stenosis is time-consuming and labor-intensive.
• Accurate stenosis detection and plaque classification are crucial for patient management.

Purpose of the Study:

• To develop and validate a deep learning (DL) algorithm for automated stenosis detection and plaque classification in head and neck CT angiography.
• To compare the DL algorithm's performance against experienced radiologists.

Main Methods:

• A convolutional neural network-based DL algorithm was trained on retrospective head and neck CT angiography data (7:2:1 train/validation/test split).
• Prospective data was used for independent testing.
• Algorithm performance was evaluated against radiologist consensus for stenosis grade and plaque classification.

Main Results:

• The DL algorithm demonstrated high consistency with radiologists in plaque classification (85.6% per-vessel).
• The AI model improved diagnostic confidence and significantly reduced radiologist diagnosis and report writing time (28.8 min to 12.4 min, P < .001).

Conclusions:

• The developed deep learning algorithm achieves accurate vessel stenosis detection and plaque classification in head and neck CT angiography.
• The AI tool offers diagnostic performance equivalent to experienced radiologists, enhancing efficiency.