Adjustment of the underestimation of coronary artery calcification scoring and risk reclassification in low-dose coronary computed tomography angiography with Knowledge-based Iterative Model Reconstruction

Abstract

Background: Knowledge-based iterative model reconstruction (IMR) can reduce radiation exposure, but trend to underestimate coronary artery calcification score (CACS) on computed tomography. We aimed to adjust the impact of low-dose IMR on CAC scoring and risk reclassification.

Methods: From June 2016 to July 2018, two groups of patients (N = 250 and N = 346) who underwent routine-dose (120kV, 50mA) CAC scan with filtered back projection (FBP) reconstruction were enrolled as training and testing group respectively. A low-dose (120kV, 20mA) scan with IMR reconstruction was performed at the same time. Agatston scores were calculated semi-automatically on the routine-dose FBP and low-dose IMR images. In the training group, a mathematical relationship between the CAC scores obtained from FBP and IMR was modeled by weighted least square method. In the testing group, adjusted IMR (ad-IMR) scores were calculated using the equation from the training group. Differences between ad-IMR and FBP scores, and consistency rates of risk categories by IMR/ad-IMR to FBP scores were analyzed.

Results: In the training group, CAC were underestimated by 26.0% (P < 0.0001) with low-dose IMR, the adjustment equation was Y = 17.45 + 1.14X (Y: FBP, X: IMR R² = 0.96). There was no difference between ad-IMR and FBP scores in testing group. Furthermore, the consistency rate of risk categories was significantly improved by ad-IMR scores (from 74.0% to 85.3%, P < 0.001), greater improvement was observed in patients with FBP score > 10 (91.6%).

Conclusion: The underestimation of CACS by low-dose scan with IMR reconstruction could be adjusted by mathematical adjustment. The impact on risk reclassification can be improved thereby facilitating further dose reductions.