Comparison of deep learning reconstruction method in coronary CT angiography with other reconstructi̇on methods

AIM: With the developing technology in coronary CTA, it is aimed to obtain higher image quality with low radiation dose and studies are being carried out on this. However, further reduction in radiation dose causes an increase in noise in the obtained CTA images and therefore a decrease in image quality. Therefore, it may compromise the diagnostic ability of CT images.We aimed to compare deep learning reconstruction with other reconstruction methods in terms of image quality and sharpness and ease of coronary artery evaluation. MATERIALS AND METHODS: We enrolled 75 patients (38 men, 37 women) who underwent coronary CTA on a 640-slice CT scanner. The images were reconstructed with model-based IR, hybrid IR, FBP, and DLR. The image noise, the signal-to-noise ratio (SNR) and the contrast-to-noise ratio (CNR) in the ascending aorta and interventricular septum was measured on all images . We also generated CT attenuation profiles across the proximal coronary arteries (RCA, LMCA, LAD and LCX) and measured the width of the edge rise distance (ERD) and the edge rise slope (ERS) in addition to CNR and SNR. Two observers visually scored the subjective image quality of coronary arteries using a 4-point scale (1 = poor, 4 = excellent) in terms of eligibility for assessment. RESULTS: MBIR provided higher Hounsfield unit (HU) values in the aorta and coronary arteries (p < 0.01) and similar attenuation in the fat and muscle compared with DLR, FBP and hybrid IR. The image noise in all tissues was significantly lower in DLR than in MBIR, hybrid IR and FBP (p < 0.01). Additionally, The SNR and CNR were significantly higher in the all DLR groups than in the MBIR, FBP and hybrid IR groups (p < 0.01). The mean ERD on each coronary artery was significantly shorter on DLR than MBIR, hybrid IR and FBP images (p < 0.01). whereas the mean ERS was steeper on DLR than MBIR, hybrid IR and FBP images (p < 0.01). The subjective visual scores were higher in the DLR than in the images reconstructed with MBIR, FBP, and hybrid IR. However, in terms of diagnostic sufficiency, no significant differences were observed between DLR and MBIR. CONCLUSION: Application of DLR increases the spatial resolution and noise reduction performance and show significant improvement in sharpness of the coronary artery walls and the evaluation of the lumen on CCTA images.