Linking sequence restoration capability of shuffled coronary angiography to coronary artery disease diagnosis

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Medical Diagnostics

Background:

• The diagnostic potential of sequential data within Coronary Angiography (CA) frames for Coronary Artery Disease (CAD) remains underexplored.
• Current diagnostic methods may not fully leverage the dynamic information present in CA sequences.

Purpose of the Study:

• To investigate the "Sequence Value" within CA frames for CAD diagnosis.
• To develop and validate a method for assessing the Sequence Restoration Capability (SRC) for CAD indication.
• To establish the robustness and efficiency of the proposed diagnostic approach.

Main Methods:

• A survey using Amazon Mechanical Turk was conducted to evaluate SRC's effectiveness in indicating CAD.
• A self-supervised deep learning model was developed for automated SRC assessment.
• Statistical analysis was performed using varied selections of coronary angiographies/modules to ensure result robustness.

Main Results:

• The self-supervised deep learning model achieved an average Area Under the Curve (AUC) of 80.1% across five-fold validation.
• The model demonstrated robustness against static data noise.
• The diagnostic assessment was completed efficiently, with calculations within 30 seconds.

Conclusions:

• The study highlights the significant diagnostic potential of sequence value in coronary angiography for CAD.
• Methodologies for harnessing this potential were successfully illustrated.
• This research contributes valuable insights and tools to the field of CAD diagnosis.