Training an Artificial Intelligence Model for Aortic Dissection Detection Using Non-Contrast Computed Tomography...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiovascular Disease

Background:

• Aortic dissection (AD) is a life-threatening condition requiring prompt diagnosis.
• Current diagnostic methods can be time-consuming, delaying critical treatment.
• There is a clinical need for faster, more accessible AD identification tools.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI)-based model for identifying aortic dissection (AD).
• To utilize non-contrast computed tomography (CT) for AD detection.
• To create an accessible tool for preliminary AD screening.

Main Methods:

• Collected chest CT and aortic CT angiography datasets from AD and non-AD patients.
• Manually segmented and annotated vascular structures on axial images using LabelMe software.
• Developed and validated an AI model using an 8:1:1 training, test, and validation data split.

Main Results:

• Successfully developed an AI model with robust detection performance for AD.
• Established an online platform for effective visualization and presentation of results.
• Demonstrated the model's capability for rapid, preliminary screening of AD.

Conclusions:

• The developed AI model offers a powerful and intelligent solution for AD identification.
• This approach addresses the unmet clinical need for accessible early detection of aortic dissection.
• The AI tool can aid in rapid preliminary screening across various clinical settings.