A Cascaded Multi-Task Generative Framework for Detecting Aortic Dissection on 3-D Non-Contrast-Enhanced Computed...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiovascular Diseases

Background:

• Contrast-enhanced computed tomography (CE-CT) is standard for aortic dissection (AD) but poses risks like allergic reactions and renal failure.
• Non-contrast-enhanced CT (NCE-CT) has limited sensitivity for AD diagnosis, leading to potential misdiagnosis.
• Developing safer and more sensitive AD detection methods is crucial for patient care.

Purpose of the Study:

• To propose a novel cascaded multi-task generative framework for aortic dissection detection using non-contrast-enhanced CT (NCE-CT) volumes.
• To improve the sensitivity and accuracy of AD diagnosis by leveraging NCE-CT data.
• To provide a potential alternative to CE-CT, mitigating risks associated with contrast agents.

Main Methods:

• A cascaded framework combining a 3D nnU-Net for aorta segmentation and a 3D multi-task generative architecture (3D MTGA).
• The 3D MTGA simultaneously synthesizes CE-CT volumes, segments true and false lumens, and classifies patients for AD.
• Theoretical formulation using Jensen-Shannon Divergence (JSD) to enhance AD detection performance.

Main Results:

• The proposed framework achieved an average accuracy of 0.831, sensitivity of 0.938, and F1-score of 0.847.
• Outperformed seven state-of-the-art classification models across radiologists with varying experience levels.
• Demonstrated superior performance compared to existing methods in aortic dissection detection.

Conclusions:

• The novel cascaded multi-task generative framework shows significant potential for accurate AD detection using NCE-CT.
• This approach can reduce misdiagnosis rates and offers a safer alternative to CE-CT.
• The framework's superior performance highlights its clinical applicability in diagnosing aortic dissection.