Deep learning-based motion correction: cardiac motion artifact and image quality improvements on chest CT

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Cardiac and respiratory motion artifacts degrade chest CT image quality.
• This degradation hinders the detection and evaluation of lung abnormalities.
• Deep learning (DL) algorithms offer potential solutions for motion artifact reduction.

Purpose of the Study:

• To evaluate the effectiveness of the CLEAR Motion DL algorithm in reducing motion artifacts on chest CT.
• To determine if CLEAR Motion improves image quality in patients with pulmonary diseases using lung window settings.

Main Methods:

• Fifty-six patients with thoracic diseases underwent chest CT scans.
• Scans were reconstructed using conventional methods and the CLEAR Motion algorithm.
• Quantitative (CPED, CPES) and qualitative (visual assessment) metrics were used for comparison.

Main Results:

• CLEAR Motion significantly improved cardio-pulmonary edge distance (CPED) and slope (CPES) (p < 0.001).
• Overall image quality, cardiac motion artifact reduction, and region conspicuity were significantly enhanced with CLEAR Motion (p < 0.001).

Conclusions:

• The CLEAR Motion DL algorithm demonstrates significant potential for improving chest CT image quality.
• CLEAR Motion effectively reduces motion artifacts, enhancing diagnostic capabilities for pulmonary diseases.