Assessment of Image Quality of Coronary Computed Tomography Angiography in Obese Patients by Comparing Deep Learning Image Reconstruction With Adaptive Statistical Iterative Reconstruction Veo.

Abstract

The aim of the study was to evaluate the image quality of coronary computed tomography (CT) angiography (CCTA) in obese patients by using deep learning image reconstruction (DLIR) in comparison with adaptive statistical iterative reconstruction Veo (ASiR-V). We prospectively evaluated 60 obese patients (body mass index [BMI] ≥ 30 kg/m2) who underwent coronary CT angiography in a single center. All CT scans were performed with GE Revolution 256-row CT at 120 kV (group A; 20 men, 10 women; mean age = 54.3 years; mean BMI = 33.4 kg/m2) or 100 kV (group B; 18 men; 12 women; mean age = 56.8 years; mean BMI = 32.9 kg/m2). Images in group A were reconstructed using ASiR-V, whereas images in group B were reconstructed using ASiR-V, DLIR-medium (DLIR-M), and DLIR-high (DLIR-H). Three blinded independent readers assessed the subjective image quality and measured the objective image quality. Radiation dose estimates were calculated and compared between patients by using 0.014 and 0.026 mSv·mGy-1 cm-1 corresponding to chest and heart conversion coefficients, respectively. The subjective score was significantly higher for images reconstructed using 120-kV ASiR-V (3.8), DLIR-M (3.9), and DLIR-H (4.0) compared with those reconstructed using 100-kV ASiR-V (3.5). Image noise was significantly lower in images reconstructed using DLIR-H compared with those reconstructed using other reconstruction algorithm (P < 0.001, respectively). The contrast-to-noise ratio was significantly higher in the DLIR-H group than in the groups using other reconstruction algorithm (P < 0.001). The effective radiation dose was significantly lower in group B than in group A (P < 0.001). Compared with ASiR-V, DLIR improved image quality in obese individuals without comprising image quality or increasing the radiation dose.