Stenosis-DetNet: Sequence consistency-based stenosis detection for X-ray coronary angiography

Area of Science:

• Medical Imaging
• Cardiovascular Diagnostics
• Artificial Intelligence in Medicine

Background:

• Accurate detection of coronary artery stenosis is crucial for diagnosing coronary heart disease.
• Conventional methods struggle with motion artifacts and subtle vascular features in single-frame X-ray images.

Purpose of the Study:

• To develop an automated method for detecting coronary artery stenosis using X-ray image sequences.
• To enhance the accuracy and reliability of stenosis detection by leveraging temporal information.

Main Methods:

• Proposes Stenosis-DetNet, an object detection network incorporating a sequence feature fusion module and a sequence consistency alignment module.
• These modules fuse candidate box features and utilize temporal information from adjacent images to optimize detection.
• The method aims to maximize temporal information for improved accuracy.

Main Results:

• Stenosis-DetNet achieved a precision of 94.87% and a sensitivity of 82.22% on 166 X-ray image sequences.
• These results surpassed those of three existing stenosis detection methods.
• The method effectively suppressed false positive and false negative detection results.

Conclusions:

• The proposed Stenosis-DetNet method demonstrates superior performance compared to state-of-the-art techniques.
• It effectively addresses challenges posed by motion and weak vascular features in angiography images.
• This approach offers a more reliable tool for coronary artery stenosis detection.