Agatston Score calcium quantification with arbitrary tube voltage

Agatston Score calcium quantification with arbitrary tube voltage White paper Thomas Allmendinger, PhD, HC DI CT R&D CTC SA Astrid Hamann, MSc, HC DI CT CRM CMM SIEMENS Healthineers siemens.com/healthineers White paper ∙ Agatston Score calcium quantification with arbitrary tube voltage Table of contents Introduction 3 Background: Accuracy of Agatston Score 4 Effect on HU stability – Region of Interest (ROI) phantom measurements 5 Agatston equivalent calcium scoring using Sa36f reconstruction kernel Anthropomorphic thorax phantom 8 Human heart specimens 12 Conclusion 14 References 15 2 Agatston Score calcium quantification with arbitrary tube voltage ∙ White paper Introduction Quantification of coronary artery calcium (CAC) is an established method of stratifying risk in coronary artery disease.3, 6 The standard quantification method, known as the Agatston Score, was developed by Arthur Agatston, Warren Janowitz, and colleagues.1 Initially, measurements for Agatston scoring were obtained using electron-beam computed tomography (EBCT) at 130 kV. Despite the various drawbacks described in the following, the large body of available clinical data across age, gender, and race4 and , extensive clinical risk stratification studies have helped establish the Agatston Score as the primary quantification tool in clinical practice.2, 6 When adapting the calcium score (CaSc) acquisition protocol by changing the initial kV settings, it is essential to preserve an equivalent high level of CAC detectability, and allow comparability of the resulting Agatston Score with the original results acquired at 120 kV and 130 kV respectively. The reason for two reference voltages is mainly historical in nature, as the initial EBCT study used only 130 kV, whereas for most modern CT systems only 120 kV is available.1, 5 This paper describes an Agatston-equivalent calcium scoring method using a dedicated reconstruction kernel (Sa36f) that allows use of all tube voltage levels offered by a given scanner model. Measurements are performed on both an anthropomorphic thorax phantom and two explanted hearts from human specimens. Furthermore, a calcium mass score conversion factor is derived that is also independent of tube voltage. 3 White paper ∙ Agatston Score calcium quantification with arbitrary tube voltage Background: Accuracy of Agatston Score The clinical value of the Agatston Score has been demonstrated by extensive patient data initially acquired using EBCT at 130 kV. With the introduction of multidetector computed tomography (MDCT) this database was extended to include CT examina- tions at 120 kV. Efforts were also made to develop a consensus standard for quantifi- cation of CAC.5 Recent studies, however, have highlighted the drawbacks of this approach. Willemink et al. investigated the inter-vendor variability of Agatston Scores obtained using state-of-the-art CT systems from the four major vendors.8 The resulting measurements delivered substantially different Agatston Scores among the four vendors. This variance could potentially influence the assessment of cardiovascular risk, and consequently treatment decisions. Furthermore, the influence of chest size on Agatston Score measurements performed using fixed 120 kV acquisition protocols was evaluated in a multi-vendor phantom study.7 An extension ring was used here to simulate differences in patient size. Results showed a systematic underestimation of Agatston Scores among vendors. This issue may be relevant to both larger patients and women with higher levels of thoracic fat and breast tissue. Due to the quantitative nature of such risk stratification, accurate calibration is highly recommended. An established procedure for calculation of coronary calcium mass scores has been in place for several years.5 The standardized procedure de- scribed here can easily be applied to any new scanner system or combination of tube voltage and tube filter to derive accurate conversion factors for calcium mass score calculation. Agatston scoring remains the standard method for coronary calcium quantification. 4 Agatston Score calcium quantification with arbitrary tube voltage ∙ White paper Effect on HU stability – region of interest (ROI) phantom measurements Tube voltage is critical in Agatston scoring due to the inherent dependency between materials such as calcium and their Hounsfield units (HUs) as a function of the incident spectrum.9, 10 In order to minimize the effect of this dependency, a new Sa36f kernel is introduced with sharpness properties equivalent to the Qr36f kernel, the standard CaSc kernel on SOMATOM® Force. A voltage-dependent lookup table based on raw data is used for image reconstruction of noncontrast CT acquisitions, pro- ducing images with Hounsfield unit (HU) values equivalent to 120 kV for bone and calcium (ART120). This enables Agatston scoring without changing the original Agatston weighting threshold, regardless of the original tube voltage chosen for image acquisition. To illustrate the conformity between the different voltages, HU measurements were performed for the small, medium, and large phantom setup at tube voltages of 70 kV, 80 kV, 90 kV, 100 kV, 110 kV, 120 kV, 130 kV, 140 kV, 150 kV, and Sn100 kV. An anthro- pomorphic thorax phantom was used, combined with a calcium calibration insert consisting of nine small cylindrical calcifications of varying size and calcium hydroxy- apatite (HA) density (Fig. 1). A complete, detailed description of the phantom is provided by McCollough et al.5 All HU measurements in this section were performed on the 5 mm cylindrical calcification inserts with HA density levels of 200, 400 and 800 mg/cm³ respectively. HUs were measured by setting ROIs (3 mm diameter, blue dot) automatically centered on the respective insert. Measurements were repeated for both the Qr36f and the new Sa36f kernel. In addition, an ROI in the center of the phantom (1 cm diameter, black circle) taken from the central slice was used to measure the plastic base material of the phantom, intended to be equivalent to soft tissue. Figure 1 illustrates the reference phantom with the respective ROIs positioned for both HU measurements and the subsequent Agatston-based CaSc measurements discussed in the following chapter. All phantom images were acquired in sequential cardiac quick scan mode following the default clinical CaSc protocol of SOMATOM Force at maximum tube current so as to minimize statistical uncertainties. The effective mean water-equivalent diameter was 20 cm, 28 cm and 35 cm for the small, medium and large phantom respectively. Fig. 1: A photo of the reference calcium scoring phantom used in this study with a schematic overlay showing the ROIs used for HU measurements of the large inserts (blue dots) and Agatston-based calcium scoring (black circles). The 1-cm diameter ROI (black circle) for soft tissue equivalence measurements can be seen in the center of the phantom. 5 White paper ∙ Agatston Score calcium quantification with arbitrary tube voltage Figure 2 shows the results of these HU measurements, with the uncorrected Qr36f reconstructions (top) and the corrected Sa36f reconstructions (bottom). In the top graph, the inherent dependency between calcium and its HUs as a function of the incident spectrum is clearly apparent in the increase in calcium HU values for lower tube voltages. Conversely, the Sa36f graph below clearly shows HU stability for all insert densities regardless of the source tube voltage. 1750 Small phantom Medium phantom Large phantom 500 1500 250 1250 200 Houndsfield Units 1000 150 750 100 500 50 250 0 .......... 0 -50 70 kV 70 kV 70 kV 80 kV 80 kV 80 kV 90 kV 90 kV 90 kV 110 kV 110 kV 110 kV 120 kV 120 kV 120 kV 150 kV 150 kV 150 kV 130 kV 130 kV 130 kV 140 kV 140 kV 140 kV 100 kV 100 kV 100 kV Sn100 kV Sn100 kV Sn100 kV HU Center HU HA200 HU HA400 HU HA800 Tube voltage 1750 Small phantom Medium phantom Large phantom 500 1500 250 1250 200 Houndsfield Units Fig. 2: HU values derived from 1000 150 measurements of the calcium scoring reference phantom using the uncorrected 750 100 Qr36f reconstructions (top) and corrected Sa36f reconstructions (bottom). The y-axis on the left measures the HU values 500 50 for the insert with different HA density ........... values as a function of tube voltage for the different phantom sizes. The increase 250 0 in the calcium HU values for lower tube voltages in the top plot is expected from 0 -50 the material, whereas HU stability can be appreciated in the bottom Sa36f plot. The 70 kV 70 kV 70 kV 80 kV 80 kV 80 kV 90 kV 90 kV 90 kV y-axis on the right measures the HU 110 kV 110 kV 110 kV 120 kV 120 kV 120 kV 150 kV 150 kV 150 kV 130 kV 130 kV 130 kV 140 kV 140 kV 140 kV 100 kV 100 kV 100 kV values for the central base material ROI, Sn100 kV Sn100 kV Sn100 kV shown as a dotted blue line in the plot. A .... HU Center HU HA200 HU HA400 HU HA800 relatively strong “soft tissue” dependency of the phantom’s plastic base material is apparent in both plots. Tube voltage 6 Agatston Score calcium quantification with arbitrary tube voltage ∙ White paper In addition, both figures also show the mean HU value of the central base material ROI (dotted line) with the y-axis scale shown on the right. In both cases the base material displays a relatively strong tube voltage dependency between 70 kV and 150 kV with a delta of up to 50 HU. However, experience of clinical patient scans shows that this drop-off behavior towards lower tube voltages is not realistic for actual human soft tissue. In order to estimate the extent of this effect in cardiac tissue, two hearts explanted from human cadavers and embedded in 27 cm water cylinders were scanned using SOMATOM Force with the CaSc protocol setup described above. Manual ROI measurements were performed in multiple slices in each scan and an average HU value was extracted for each tube voltage and individual heart. Figure 3 shows the results of these measurements for each heart, clearly illustrating very little dependency between the HU value and the different tube voltages. In light of these results, it can be concluded that a scoring method based on thresholds in HU units such as the Agatston Score should yield similar scoring results for tube voltages between 70 kV and 150 kV if the Sa36f (ART120) reconstruction method described above is applied to the data. Moreover, the intrinsic beam hardening correction properties of this method suggest that greater independence of the Agatston Score from patient diameter can be expected. This is a result of the increased Hounsfield unit stability with respect to the object size. The dependency of the score on diameter is known and reported in the literature for phantom measure- ments, and confirmed by patient data.7 60 60 40 40 Fig. 3: HU values of the two explanted 20 20 heart specimens (Heart 1 on the left, Heart 2 on the right). Only a very minor Houndsfield Units 0 0 downward slope towards lower tube voltages is visible in Heart 2 and the right ventricle ROI of Heart 1. The left 70 kV 70 kV 80 kV 80 kV 120 kV 120 kV 150 kV 150 kV 100 kV 100 kV SoftTissueLeft SoftTissueLeft ventricle ROI measurement of Heart 1 SoftTissueRight SoftTissueRight exhibits a slight offset including some Water Water upslope. This effect is most likely due to the residual iodine contrast medium Tube voltage Tube voltage found in this heart specimen. 7 White paper ∙ Agatston Score calcium quantification with arbitrary tube voltage 8 Fig. 4: Agatston Score values derived from the contributions for the 200 mg/cm3 density levels respectively (orange, berry, fixed base material threshold of 130 HU. the “soft tissue” base material is clearly caused by the strong HU dependency of reference (lower: 595, upper: 672, mean: The systematic slope of the Sa36f score visible. The stacked bar plots illustrate 630) based on inter-scanner variations uncorrected Qr36f (top) and corrected 400 mg/cm3 measurements of the calcium scoring Sa36f reconstruction (bottom) with a petrol). The red line provides a visual reference phantom based on the found by McCollough et al.5 HA, and 800 mg/cm3 HA, HA Agatston Score Agatston Score Agatston-equivalent calcium scoring Anthropomorphic thorax phantom The Agatston scoring method is based on a single 130 HU threshold for voxel count Scores since the insert location could be predefined by setting ROIs with a radius three acquired as described above using the standard default CaSc protocol. inclusion, combined with three further thresholds for lesion weighting.1 rected Sa36f reconstructions (bottom). measurements, it was possible to automatically extract and evaluate the Agatston times larger than the size of the actual insert (Figure 1, black circles). Images were Figure 4 shows the results of the uncorrected Qr36f reconstructions (top) and cor- using Sa36f reconstruction kernel 1000 1000 400 400 800 800 600 600 200 200 0 0 70 kV 70 kV 80 kV 80 kV 90 kV 90 kV Small phantom Small phantom 100 kV 100 kV 110 kV 110 kV 120 kV 120 kV 130 kV 130 kV 140 kV 140 kV 150 kV 150 kV Sn100 kV Sn100 kV 70 kV 70 kV 80 kV 80 kV 90 kV 90 kV Tube voltage Tube voltage Medium phantom Medium phantom 100 kV 100 kV 110 kV 110 kV 120 kV 120 kV 130 kV 130 kV 140 kV 140 kV 150 kV 150 kV Sn100 kV Sn100 kV 70 kV 70 kV 80 kV 80 kV 90 kV 90 kV Large phantom Large phantom 100 kV 100 kV 110 kV 110 kV 120 kV 120 kV In phantom 130 kV 130 kV 140 kV 140 kV 150 kV 150 kV Sn100 kV Sn100 kV Agatston Score calcium quantification with arbitrary tube voltage ∙ White paper Although corrected, the stacked plots of the Sa36f reconstructions (Figure 4, bottom) exhibit a clearly visible systematic slope of the Agatston Score as a function of tube voltage. The primary reason for this effect is the strong spectral dependency of the base material HU values, as previously shown (Figure 2, dotted line). This behavior influences scoring due to the correlation between a fixed 130 HU threshold for pixel counting and the edge spread function of a soft reconstruction kernel. To validate this hypothesis, the scoring procedure was adjusted so that the pixel-counting threshold is expressed as a distance from the measured base material’s HU value (T=Δ+HU(base)). A distance of Δ=95 HU was chosen, automatically leading to the established threshold of 130 HU for 120 kV in combination with the measured base material HU value of HU(base)=35 HU, where the variable HU(base) was adjusted for each combination of tube voltage and phantom size. The result of the adjusted scoring procedure is depicted in Figure 5, for both the Qr36f (top) and Sa36f reconstructions (bottom). After the adjustments, the corrected Sa36f reconstructions show a constant Agatston Score within the previously defined range for the individual tube voltages and phantom sizes. The only upward fluctuation observed relates to the 70 kV measurement in the large phantom. This is most likely due to the increase in noise for this particular setting in combination with the slight noise sensitivity inherent to Agatston scoring. 9 10 White paper ∙ Agatston Score calcium quantification with arbitrary tube voltage Fig. 5: Agatston Score values derived from density levels respectively (orange, berry, reconstruction after taking into account distance from the measured base “soft petrol). The bottom plot demonstrates uncorrected Qr36f (top) and corrected contributions for the 200 mg/cm3 Agatston Score of the Sa36f (ART120) 400 mg/cm3 measurements of the calcium scoring Sa36f reconstruction (bottom) with a the tube voltage independence of the T = HU(base) + 95 HU expressed as a same established 130 HU threshold. variable base material threshold of The stacked bar plots illustrate the For 120 kV this approach yields the the spectral shortcomings of the tissue” material in the phantom. reference phantom based on the HA, and 800 mg/cm3 base material. HA, HA Agatston Score Agatston Score 1000 1000 400 400 800 800 600 600 200 200 0 0 70 kV 70 kV 80 kV 80 kV 90 kV 90 kV Small phantom Small phantom 100 kV 100 kV 110 kV 110 kV 120 kV 120 kV 130 kV 130 kV 140 kV 140 kV 150 kV 150 kV Sn100 kV Sn100 kV 70 kV 70 kV 80 kV 80 kV 90 kV 90 kV Tube voltage Tube voltage Medium phantom Medium phantom 100 kV 100 kV 110 kV 110 kV 120 kV 120 kV 130 kV 130 kV 140 kV 140 kV 150 kV 150 kV Sn100 kV Sn100 kV 70 kV 70 kV 80 kV 80 kV 90 kV 90 kV Large phantom Large phantom 100 kV 100 kV 110 kV 110 kV 120 kV 120 kV 130 kV 130 kV 140 kV 140 kV 150 kV 150 kV Sn100 kV Sn100 kV Calcium Mass Score Factor value of the respective insert (Figure 6). was originally introduced to eliminate tube voltage dependency from the scoring by actual mass density of calcium. To support the score, a typically tube voltage- a calcium mass score in most calcium scoring applications. The calcium mass score dependent conversion factor is provided as a DICOM entry with each reconstruction. is derived from the measured HU values of the phantom and the nominal HA density introducing a calibration step. This step correlates the HU values measured to the In the case of a corrected Sa36f reconstruction, only a single average value of c=0.81 In addition, a calcium mass conversion factor can be derived to enable calculation of 0.4 0.8 0.6 0.2 1.0 1.2 0 70 kV 80 kV 90 kV Small phantom 100 kV 110 kV 120 kV 130 kV 140 kV 150 kV Sn100 kV 70 kV 80 kV 90 kV Tube voltage Medium phantom 100 kV 110 kV 120 kV 130 kV 140 kV 150 kV Agatston Score calcium quantification with arbitrary tube voltage ∙ White paper Sn100 kV 70 kV 80 kV 90 kV Large phantom 100 kV 110 kV 120 kV 130 kV 140 kV 150 kV Sn100 kV Sa36f reconstruction. of the individual inserts for the corrected Fig. 6: Calcium mass score conversion factor derived from the measured HU values 11 White paper ∙ Agatston Score calcium quantification with arbitrary tube voltage Human heart specimens To compensate for the shortcomings of Agatston scoring in the reference phantom described in the previous chapter, a decision was made to further validate the method in a realistic tissue environment. Given the need for repeated acquisition at different tube voltages, a test involving patients was considered unethical. Accordingly, two explanted human hearts were scanned using SOMATOM Force in collaboration with University Hospital Wuerzburg, Germany. All measurements reported in this section are based on a CARE Dose4DTM and CARE kV dose modulation approach using the attenuation information derived from the topogram of the measurement setup. This approach mimics clinical practice, where the applied dose is also dependent on the size of the patient to be imaged. The initial quality reference value used for 120 kV was set to 80 q.ref.mAs/rot, corresponding to the established clinical default value for SOMATOM Force. The values for all other tube voltages were derived semi-automati- cally using CARE kV based on exactly the same reference starting point, a target material contrast setting of “bone/calcium” and manual setting of the target voltage. The system automatically adjusts the tube current based on the changed tube voltage in combination with the user-defined material contrast target setting. This makes it possible to potentially save dose due to the contrast increase if scanning is performed at lower kV settings compared to the reference point. The results of the measurements were reconstructed using Qr36f (uncorrected) and Sa36f (corrected) with the standard 3 mm slice thickness and 1.5 mm increment. Manual Agatston scoring was performed with a commercially available standard software package (syngo.via VB20) on three selected groups of calcifications by an experienced professional. These three groups were assigned the labels RCA, LCA and CX in the software. They were not limited to calcifications of the vessels, but also included calcifications of the aortic valve and aorta. This results in relatively high total score values, but prevents the evaluation from being dominated by the statisti- cal uncertainties of small calcifications. Example image snapshots of the manual scoring procedure are shown side by side in Figure 7 (Heart 1, top; Heart 2, bottom). Fig. 7: Example image snapshots from the calcium scoring application (syngo.via VB20) of two explanted human cadaver hearts (Heart 1, top; Heart 2, bottom) at 70 kV using the uncorrected Qr36f reconstruction (left column) and the corrected Sa36f reconstruction (right column). 12 Agatston Score calcium quantification with arbitrary tube voltage ∙ White paper Figure 8 shows the result plots illustrating the Agatston Score for Heart 1 using the uncorrected Qr36f (left) and corrected Sa36f reconstructions (right). Figure 9 provides the same plots for Heart 2. As shown by the uncorrected distribution, the score behaves as expected with a strong increase towards low voltages below 120 kV, while decreasing sharply at Sn150 kV. Here, the combination of high tube voltage and addi- tional tin filtration yields a very hard spectrum. By contrast, the plots of the corrected Sa36f reconstructions clearly show that a kV-independent Agatston Score is achieved across the entire tube voltage spectrum, ranging from 70 kV to Sn150 kV, with approxi- mately 10% relative uncertainty. This is well within the current precision limits achieved in Agatston Score-based calcium scoring for different scanner systems. It should also be noted that this uncertainty level is derived from a limited dataset from two heart specimens involving a manual scoring step. Accordingly, a higher degree of precision may be possible with this method. However, demonstrating it would require a much larger dataset. 12000 12000 10000 10000 8000 8000 Agatston Score Agatston Score 6000 6000 4000 4000 2000 2000 RCA RCA CX CX 0 LAD 0 LAD Fig. 8: Agatston scoring results as a function of tube voltage, derived from measurements performed in an explanted human cadaver 70 kV 70 kV 80 kV 80 kV 120 kV 120 kV 100 kV 100 kV heart (Heart 1) used to validate the method Sn150 kV Sn150 kV Tube voltage Tube voltage (Qr36f, left; Sa36f, right). 6000 6000 5000 5000 4000 4000 Agatston Score Agatston Score 3000 3000 2000 2000 1000 1000 RCA RCA CX CX Fig. 9: Agatston scoring results as a function 0 LAD 0 LAD of tube voltage derived from measurements performed in an explanted human cadaver 70 kV 70 kV 80 kV 80 kV 120 kV 120 kV 100 kV 100 kV heart (Heart 2) used to validate the method Sn150 kV Sn150 kV Tube voltage Tube voltage (Qr36f, left; Sa36f, right). 13 White paper ∙ Agatston Score calcium quantification with arbitrary tube voltage Conclusion The findings show that applying the dedicated Sa36f reconstruction method to non- contrast CT data enables generation of artificial 120 kV equivalent CT images suitable for Agatston-equivalent calcium scoring based on the established fixed set of scoring thresholds, regardless of the tube voltage of the original CT acquisition. Moreover, a calcium mass score conversion factor of c=0.81 can be derived based on phantom HU measurements regardless of the tube voltage for the dedicated Sa36f reconstruction. References 1 Agatston AS, Janowitz WR, Hildner FJ, Zusmer NR, Viamonte M Jr, Detrano R. Quantification of coronary artery calcium using ultrafast computed tomography. J Am Coll Cardiol. 1990 Mar;15(4):827-832. 2 Alluri K, Joshi PH, Henry TS, Blumenthal RS, Nasir K, Blaha MJ. Scoring of Coronary Artery Calcium Scans: History, Assumptions, Current Limitations, and Future Directions. Atherosclerosis. 2015 Mar;239(1):109-17. 3 Al-Mallah MH, Aljizeeri A, Villines TC, Srichai MB, Alsaileek A. Cardiac computed tomography in current cardiology guidelines. J Cardiovasc Comput Tomogr. 2015 Nov-Dec;9(6):514-23. 4 Hoff JA, Chomka EV, Krainik AJ, Daviglus M, Rich S, Kondos GT. Age and gender distributions of coronary artery calcium detected by electron beam tomography in 35,246 adults, Am. J. Cardiol. 2001 Jun 15;87(12):1335-9. 5 McCollough CH, Ulzheimer S, Halliburton SS, Shanneik K, White RD, Kalender WA. Coronary artery calcium: a multi-institutional, multimanufacturer international standard for quantification at cardiac CT. Radiology. 2007 May;243(2):527-38 6 Parikh S, Budoff MJ. Calcium Scoring and Cardiac Computed Tomography. Heart failure clinics. 2016 Jan;12(1):97-105. 7 Willemink MJ, Abramiuc B, den Harder AM, van der Werf NR, de Jong PA, Budde RP, et al. Coronary calcium scores are systematically underestimated at a large chest size: A multivendor phantom study. J Cardiovasc Comput Tomogr. 2015 Sep-Oct;9(5):415-421. 8 Willemink MJ, Vliegenthart R, Takx RA, Leiner T, Budde RP, Bleys RL, et al. Coronary artery calcification scoring with state-of-the-art CT scanners from different vendors has substantial effect on risk classification. Radiology. 2014 Dec;273(3):695- 702. 9 Zatz LM, Alvarez RE. An inaccuracy in computed tomography: the energy dependence of CT values. Radiology. 1977 Jul;124(1):91-7. 10 Zatz LM. The effect of the kVp level on EMI values. Selective imaging of various materials with different kVp settings. Radiology. 1976 Jun;119(3):683-8. 14 Agatston Score calcium quantification with arbitrary tube voltage ∙ White paper 15 Siemens Healthineers Headquarters Siemens Healthcare GmbH Henkestr. 127 91052 Erlangen, Germany Phone: +49 9131 84-0 siemens.com/healthineers Published by Siemens Healthcare GmbH · 5812 · 0418 online · ©Siemens Healthcare GmbH, 2018