Low tube voltage with deep learning and freezing for coronary CTA in obese patients with suspected CAD

Abstract

Objective to investigate the application of low tube voltage combined with deep learning reconstruction algorithms and freeze technology in obese patients undergoing coronary computed tomography angiography (CCTA). A total of 100 patients with a BMI of 28 kg/m2 or higher who underwent CCTA were prospectively enrolled and randomly divided into two groups, A and B, each consisting of 50 patients, constituting a prospective randomized study. Group A used 120kVp, while Group B used 100kVp. Both groups breathed freely, with a trigger threshold of 150HU and a scan initiated 3.1 s after triggering. Both groups utilized Smart-mA technology and non-ionic iodine contrast agent iopamidol (370mgI/mL), with an injection volume of BMI * 1.5 mL and an injection time of 12 s. The original images were reconstructed using three algorithms: ASiR-V50%, DLIR-M, and DLIR-H. The CT number and noise levels of the aortic root (Ao), right coronary artery (RCA), left main artery (LMA), left anterior descending artery (LAD), and proximal left circumflex artery (LCX) were measured for both groups. The objective image quality indicators, including the contrast-to-noise ratio (CNR) and signal-to-noise ratio (SNR), were compared, and the subjective image quality of the coronary arteries was evaluated using a 5-point scale. There were no significant differences between the two groups in general patient data. The CT number of Ao, pLAD, pLCX, and pRCA for the two groups (DLIR-H, DLIR-M, and ASiR-V) showed no statistically significant differences (P > 0.05). DLIR-H reconstruction demonstrated significantly lower image noise, higher SNR and CNR, and better subjective image quality compared to ASIR-V and DLIR-M. Under low tube voltage (100kVp), Group B showed higher CT number compared to Group A, with a statistically significant difference (P < 0.001). However, there were no statistically significant differences in image noise(SD), SNR, and CNR between the two groups of DLIR-H (P > 0.05). The subjective image quality scores for both groups of DLIR-H were significantly higher than those for ASIR-V and DLIR-M, with statistically significant differences (P < 0.05), and Group B scored higher than Group A. Additionally, the effective doses for both groups were (4.2 ± 1.6) mSv and (3.0 ± 1.4) mSv, respectively, with Group B showing a reduction of about 20% compared to Group A. Using low tube voltage 100kVp, deep learning reconstruction algorithms, and freeze technology for coronary artery CTA examinations in obese patients, the distal vessel enhancement is significantly improved, and the vascular branches are more clearly defined. As the denoising level of the deep learning reconstruction algorithm increases, it can further reduce image noise, achieving image quality that meets clinical diagnostic needs.