Cardiac CT motion artifact grading via semi-automatic labeling and vessel tracking using synthetic image-augmented...

Area of Science:

• Medical Imaging
• Cardiovascular Disease Evaluation
• Artificial Intelligence in Healthcare

Background:

• Cardiac CT is crucial for diagnosing cardiovascular diseases.
• Patient and organ motion during scanning causes artifacts, reducing image quality and diagnostic value.
• Objective grading of motion artifacts is essential for reliable cardiac CT analysis.

Purpose of the Study:

• To develop and validate effective methods for grading motion artifacts in cardiac CT angiography (CCTA).
• To implement semi-automatic labeling and vessel tracking for image quality assessment.
• To train a neural network for fully-automatic motion artifact grading, enhanced by synthetic data.

Main Methods:

• Semi-automatic CCTA image quality grading using vessel tracking algorithms.
• Development of a neural network model for fully-automatic motion artifact grading.
• Utilizing XCAT simulation tools to generate synthetic CT data for training.

Main Results:

• Semi-automatic grading scores closely align with expert readers (within 0.85 points on a 5-point scale).
• Synthetic data supplementation significantly improved the neural network's scoring performance.
• Mean square error for right coronary artery motion grading reduced by 36% with synthetic data.

Conclusions:

• Proposed semi-automatic and automatic methods effectively grade motion artifacts in cardiac CT.
• Synthetic data augmentation enhances the performance of automated grading systems.
• This approach can potentially reduce the volume of clinical data required for training AI models.