Calibration of coronary calcium scores determined using iterative image reconstruction (AIDR 3D) at 120, 100, and 80 kVp.

PURPOSE: Computed tomography (CT) radiation dose reduction is frequently achieved by applying lower tube voltages and using iterative reconstruction (IR). For calcium scoring, the reference protocol at 120 kVp with filtered back projection (FBP) is still used, because kVp and IR may influence the Agatston score (AS) and volume score (VS). The authors present a two-step method to optimize dose: first, to determine the lowest feasible exposure and highest noise thresholds; second, to define a calibration method that ensures that the AS and VS are similar to the reference protocol. METHODS: AS and VS were measured for an anthropomorphic thoracic phantom that includes a calcium calibration module. The phantom was scanned on a 320-row CT scanner, at tube voltages of 120 kVp using FBP, and 120, 100, and 80 kVp using adaptive iterative dose reduction (AIDR 3D) reconstruction. The minimum CTDIs were determined based on three objective quality criteria. Calibration was performed to estimate adjusted CT number thresholds for the lower kVp acquisitions. Finally, the accuracies of the total and individual insert scores at dose level close to the minimum CTDI level were investigated and compared to low (FBPLD - 120) and high (FBPHD - 120) dose reference protocols (based on ten repeated acquisitions for each group). RESULTS: IR allows the exposure to be reduced by 69% at 120 kVp, with no significant effect on the total scores when averaged over all included dose steps and compared to FBP-120 (AS: 693 vs 699, p = 0.182; VS: 588 vs 587 mm(3), p = 0.569). Also when averaged over ten repeated scans and compared to FBPHD - 120 (AS: 709 vs 704, p = 0.435; VS: 604 vs 601 mm(3), p = 0.479), there is no statistical significant effect. Reducing the peak tube voltage allows even greater dose reductions: 73% at 100 kVp and 76% at 80 kVp. The calibrated CT number thresholds for analysis at 120, 100, and 80 kVp were, respectively, 130, 133, and 160 HU for the Agatston score, and 130, 132, and 140 HU for the volume score. Following the calibration, the mean scores of the four groups with dose variation were not significantly different from the reference protocol, at 100 kVp (AS: 698 vs 699, p = 0.818; VS: 584 vs 587 mm(3), p = 0.365) or at 80 kVp (AS: 698 vs 699, p = 0.996; VS: 586 vs 587 mm(3), p = 0.827). Similarly, there was no significant score difference with FBPLD - 120 during repeated scanning: 100 kVp (AS: 690 vs 694, p = 0.394; VS: 579 vs 585 mm(3), p = 0.168) and 80 kVp (AS: 703 vs 694, p = 0.115; VS: 588 vs 585 mm(3), p = 0.613). Compared to FBPHD - 120 group, the relative score deviation for the accuracy of the 400 and 800 mg/cm(3) HA inserts with 3 and 5 mm diameter is less than 7%. However, the relative deviation of the smaller 1 mm inserts is poorer (up to 41% deviations for scores <3). CONCLUSIONS: With iterative reconstruction using AIDR 3D, deviations of the total Agatston and volume scores remain within 4% of the reference protocol. The 1 mm inserts were detected as calcification, but scores less than ten tend to be underestimated. Following the calibration process, the application of IR in combination with reduced tube voltages allows up to 76% lower radiation dose.

Determining the radiation dose reduction potential for coronary calcium scanning with computed tomography: an anthropomorphic phantom study comparing filtered backprojection and the adaptive iterative dose reduction algorithm for image reconstruction.

PURPOSE: This study describes a method to determine the lowest possible thresholds for volume computed tomographic dose index (CTDI(min)) and maximum tolerable pixel noise (SD(max)) values for coronary calcium scanning while maintaining accurate Agatston score values. The method was applied to a comparison between the iterative reconstruction (IR) and filtered backprojection (FBP) image reconstruction algorithms in a phantom study. MATERIALS AND METHODS: An anthropomorphic thoracic phantom with a calibration insert for the quantification of coronary calcium, containing 200, 400, and 800 mg HA/cm of calcium mass spheres of 1, 3, and 5 mm diameter (QRM GmbH, Moehrendorf, Germany), was scanned without (G1) and with (G2) an additional 2 cm-thick wrap of muscle-equivalent material. Electrocardiographically simulated volume scans were performed on a 320-row computed tomographic scanner (Aquilion ONE, Toshiba Medical Systems, Otawara, Japan) set to 120 kilovolt peak [kVp] and 10 to 580 mA variations in 21 steps. For the IR, the Adaptive Iterative Dose Reduction 3-dimensional algorithm (AIDR 3D) was used. Agatston scores were calculated semiautomatically on the computed tomographic console. Inclusion tests to assess the accuracy of the Agatston scores were performed to determine the CTDI(min) thresholds and the associated maximum pixel noise SD(max) for FBP and IR from identical raw data. The inclusion tests were as follows: (1) the semiautomatic identification of the 1 mm sphere with 800 mg HA/cm, (2) the exclusion of false-positive lesions, and (3) a statistical outlier test. Statistical differences between the Agatston score means from both image reconstruction algorithms were evaluated using the paired t test. RESULTS: All Agatston scores using both reconstruction methods were normally distributed (P > 0.49). For FBP and IR, the mean ± 1σ of Agatston score, CTDI(min), and SD(max), respectively, were determined as follows: 697.8 ± 7.7, 7.5 mGy, and 24.4 Hounsfield unit (HU) (G1-FBP); 678.8 ± 14.3, 1.5 mGy, and 20.1 HU (G1-IR); 677.0 ± 11.6, 14.5 mGy, and 27.3 HU (G2-FBP); and 643.9 ± 13.4, 2.6 mGy, and 20.0 HU (G2-IR). The mean Agatston scores obtained using IR (both with and without the additional 2 cm muscle shell) were slightly (approximately 5%) but significantly lower (P ≤ 0.001) than those obtained using FBP reconstruction. CONCLUSIONS: The Adaptive Iterative Dose Reduction algorithm AIDR 3D shows potential to reduce dose exposure by approximately 80% in comparison with the dose currently applied with FBP image processing. On the basis of phantom evaluation, a target noise of 20 HU for the application of this method in coronary calcium scanning is recommended to avoid loss in accuracy.