Impact of Deep Learning-Based Image Conversion on Fully Automated Coronary Artery Calcium Scoring Using Thin-Slice, Sharp-Kernel, Non-Gated, Low-Dose Chest CT Scans: A Multi-Center Study.
Authors
Affiliations (14)
Affiliations (14)
- Department of Radiology, Korea University Ansan Hospital, Ansan, Republic of Korea.
- Coreline Soft Co., Ltd, Seoul, Republic of Korea.
- Medical Science Research Center, Korea University College of Medicine, Seoul, Republic of Korea.
- Department of Radiology, Severance Hospital, Research Institute of Radiological Science, Yonsei University College of Medicine, Seoul, Republic of Korea.
- Department of Radiology, Dongsan Hospital, Keimyung University College of Medicine, Daegu, Republic of Korea.
- Department of Radiology, Seoul St. Mary's Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea.
- Department of Radiology, Soonchunhyang University Cheonan Hospital, Cheonan, Republic of Korea.
- Department of Radiology, Hanyang University Guri Hospital, Hanyang University College of Medicine, Guri, Republic of Korea.
- Department of Radiology and Research Institute of Radiology, Cardiac Imaging Center, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea.
- Department of Radiology, Korea University Guro Hospital, Korea University College of Medicine, Seoul, Republic of Korea.
- Department of Radiology and Imaging Sciences, Emory University School of Medicine, Atlanta, GA, USA.
- Translational Laboratory for Cardiothoracic Imaging and Artificial Intelligence, Emory University School of Medicine, Atlanta, GA, USA.
- Division of Cardiothoracic Imaging, Department of Radiology, Emory University, Atlanta, GA, USA.
- Department of Radiology, Severance Hospital, Research Institute of Radiological Science, Yonsei University College of Medicine, Seoul, Republic of Korea. [email protected].
Abstract
To evaluate the impact of deep learning-based image conversion on the accuracy of automated coronary artery calcium quantification using thin-slice, sharp-kernel, non-gated, low-dose chest computed tomography (LDCT) images collected from multiple institutions. A total of 225 pairs of LDCT and calcium scoring CT (CSCT) images scanned at 120 kVp and acquired from the same patient within a 6-month interval were retrospectively collected from four institutions. Image conversion was performed for LDCT images using proprietary software programs to simulate conventional CSCT. This process included 1) deep learning-based kernel conversion of low-dose, high-frequency, sharp kernels to simulate standard-dose, low-frequency kernels, and 2) thickness conversion using the raysum method to convert 1-mm or 1.25-mm thickness images to 3-mm thickness. Automated Agaston scoring was conducted on the LDCT scans before (LDCT-Org<sub>auto</sub>) and after the image conversion (LDCT-CONV<sub>auto</sub>). Manual scoring was performed on the CSCT images (CSCT<sub>manual</sub>) and used as a reference standard. The accuracy of automated Agaston scores and risk severity categorization based on the automated scoring on LDCT scans was analyzed compared to the reference standard, using the Bland-Altman analysis, concordance correlation coefficient (CCC), and weighted kappa (κ) statistic. LDCT-CONV<sub>auto</sub> demonstrated a reduced bias for Agaston score, compared with CSCT<sub>manual</sub>, than LDCT-Org<sub>auto</sub> did (-3.45 vs. 206.7). LDCT-CONV<sub>auto</sub> showed a higher CCC than LDCT-Org<sub>auto</sub> did (0.881 [95% confidence interval {CI}, 0.750-0.960] vs. 0.269 [95% CI, 0.129-0.430]). In terms of risk category assignment, LDCT-Org<sub>auto</sub> exhibited poor agreement with CSCT<sub>manual</sub> (weighted κ = 0.115 [95% CI, 0.082-0.154]), whereas LDCT-CONV<sub>auto</sub> achieved good agreement (weighted κ = 0.792 [95% CI, 0.731-0.847]). Deep learning-based conversion of LDCT images originally obtained with thin slices and a sharp kernel can enhance the accuracy of automated coronary artery calcium score measurement using the images.