Deep learning-based motion correction: cardiac motion artifact and image quality improvements on chest CT.
Authors
Affiliations (8)
Affiliations (8)
- Department of Diagnostic Radiology, Fujita Health University School of Medicine, Toyoake, Aichi, Japan.
- Department of Radiology, Fujita Health University Okazaki Medical Center, Okazaki, Aichi, Japan.
- Joint Research Laboratory of Advanced Medical Imaging and Artificial Intelligence, Fujita Health University School of Medicine, Toyoake, Aichi, Japan.
- Canon Medical Systems Corporation, Otawara, Tochigi, Japan.
- Department of Radiology, Fujita Health University Bantane Hospital, Nagoya, Aichi, Japan.
- Department of Diagnostic Radiology, Shizuoka Cancer Center, Sunto-Gun, Shizuoka, Japan.
- Department of Diagnostic Radiology, Fujita Health University School of Medicine, Toyoake, Aichi, Japan. [email protected].
- Joint Research Laboratory of Advanced Medical Imaging and Artificial Intelligence, Fujita Health University School of Medicine, Toyoake, Aichi, Japan. [email protected].
Abstract
Since the clinical application of computed tomography (CT), cardiac and respiratory motion artifacts have caused decreased image quality and reduced detection or quantitative or qualitative evaluation of lung parenchymal or vascular abnormalities on chest CT with lung window settings in patients with pulmonary diseases. Recently, a deep learning (DL)-based motion correction algorithm (CLEAR Motion) has been developed and clinically used for chest CT. We hypothesized that CLEAR Motion can significantly reduce motion artifacts on chest CT examinations relative to conventional chest CT images reconstructed without CLEAR Motion. The purpose of this study was to determine the utility of CLEAR Motion for image quality improvement in chest CT with lung window settings in patients with various pulmonary diseases. Fifty-six consecutive patients with various thoracic diseases underwent non-electrocardiogram-gated chest helical CT examination using a 320-detector row CT and underwent reconstruction using the conventional reconstruction method and CLEAR Motion. To compare the quantitative image quality, the cardio-pulmonary edge distance (CPED) and slope (CPES) were measured on each CT scan in the axial plane. Comparing cardiac motion reduction capability, overall image quality, cardiac motion artifact, and region conspicuity were visually assessed in the lung window setting on the axial, coronal, and sagittal planes. The paired t-test and Wilcoxon signed-rank test were then performed. The CPEDs and CPESs of the entire lung and left lung on CT with CLEAR Motion were significantly superior to those of CT without CLEAR Motion (p < 0.001). The overall image quality, cardiac motion artifact, and region conspicuity on CT with CLEAR Motion were significantly higher than those without CLEAR Motion on each plane (p < 0.001). The DL-based motion correction algorithm named as 'CLEAR Motion' has a potential to improve image quality on chest CT with lung window setting in patients with pulmonary diseases.