Feasibility and limitations of deep learning-based coronary calcium scoring in PET-CT: a comparison with coronary calcium score CT

OBJECTIVE

This study aimed to determine the feasibility and limitations of deep learning-based coronary calcium scoring using positron emission tomography-computed tomography (PET-CT) in comparison with coronary calcium scoring using ECG-gated non-contrast-enhanced cardiac computed tomography (CaCT).

MATERIALS AND METHODS

A total of 215 individuals who underwent both CaCT and PET-CT were enrolled in this retrospective study. The Agatston method was used to calculate the coronary artery calcium scores (CACS) from CaCT, PET-CT(reader), and PET-CT(AI) to analyse the effect of using different modalities and AI-based software on CACS measurement. The total CACS and CACS classified according to the CAC-DRS guidelines were compared between the three sets of CACS. The differences, correlation coefficients, intraclass coefficients (ICC), and concordance rates were analysed. Statistical significance was set at p < 0.05.

RESULTS

The correlation coefficient of the total CACS from CaCT and PET-CT(reader) was 0.837, PET-CT(reader) and PET-CT(AI) was 0.894, and CaCT and PET-CT(AI) was 0.768. The ICC of CACS from CaCT and PET-CT(reader) was 0.911, PET-CT(reader) and PET-CT(AI) was 0.958, and CaCT and PET-CT(AI) was 0.842. The concordance rate between CaCT and PET-CT(AI) was 73.8%, with a false-negative rate of 37.3% and a false-positive rate of 4.4%. Age and male sex were associated with an increased misclassification rate.

CONCLUSIONS

Artificial intelligence-assisted CACS measurements in PET-CT showed comparable results to CACS in coronary calcium CT. However, the relatively high false-negative results and tendency to underestimate should be of concern.

CLINICAL RELEVANCE STATEMENT

Application of automated calcium scoring to PET-CT studies could potentially select patients at high risk of coronary artery disease from among cancer patients known to be susceptible to coronary artery disease and undergoing routine PET-CT scans.

KEY POINTS

• Cancer patients are susceptible to coronary disease, and PET-CT could be potentially used to calculate coronary artery calcium score (CACS). • Calcium scoring using artificial intelligence in PET-CT automatically provides CACS with high ICC to CACS in coronary calcium CT. • However, underestimation and false negatives of CACS calculation in PET-CT should be considered.

Validation of a Length-Adjusted Abdominal Arterial Calcium Score Method for Contrast-Enhanced CT Scans

BACKGROUND

The Agatston score on noncontrast computed tomography (CT) scans is the gold standard for calcium load determination. However, contrast-enhanced CT is commonly used for patients with atherosclerotic cardiovascular diseases (ASCVDs), such as peripheral arterial occlusive disease (PAOD) and abdominal aortic aneurysm (AAA). Currently, there is no validated method to determine calcium load in the aorta and peripheral arteries with a contrast-enhanced CT. This study validated a length-adjusted calcium score (LACS) method for contrast-enhanced CT scans.

METHOD

The LACS (calcium volume in mm³/arterial length in cm) in the abdominal aorta was calculated using four-phase liver CT scans of 30 patients treated between 2017 and 2021 at the University Medical Center Groningen (UMCG) with no aortic disease. Noncontrast CT scans were segmented with a 130 Hounsfield units (HU) threshold, and a patient-specific threshold was used for contrast-enhanced CTs. The LACS was calculated and compared from both segmentations. Secondly, the interobserver variability and the influence of slice thickness (0.75 mm vs. 2.0 mm) was determined.

RESULTS

There was a high correlation between the LACS from contrast-enhanced CT scans and the LACS of noncontrast CTs (R² = 0.98). A correction factor of 1.9 was established to convert the LACS derived from contrast-enhanced CT to noncontrast CT scans. LACS interobserver agreement on contrast-enhanced CT was excellent (1.0, 95% confidence interval = 1.0-1.0). The 0.75 mm CT threshold was 541 (459-625) HU compared with 500 (419-568) HU on 2 mm CTs (p = 0.15). LACS calculated with both thresholds was not significantly different (p = 0.63).

CONCLUSION

The LACS seems to be a robust method for scoring calcium load on contrast-enhanced CT scans in arterial segments with various lengths.

Variation in aorta attenuation in contrast-enhanced CT and its implications for calcification thresholds

Background

CT contrast media improves vessel visualization but can also confound calcification measurements. We evaluated variance in aorta attenuation from varied contrast-enhancement scans, and quantified expected plaque detection errors when thresholding for calcification.

Methods

We measured aorta attenuation (AoHU) in central vessel regions from 10K abdominal CT scans and report AoHU relationships to contrast phase (non-contrast, arterial, venous, delayed), demographic variables (age, sex, weight), body location, and scan slice thickness. We also report expected plaque segmentation false-negative errors (plaque pixels misidentified as non-plaque pixels) and false-positive errors (vessel pixels falsely identified as plaque), comparing a uniform thresholding approach and a dynamic approach based on local mean/SD aorta attenuation.

Results

Females had higher AoHU than males in contrast-enhanced scans by 65/22/20 HU for arterial/venous/delayed phases (p < 0.001) but not in non-contrast scans (p > 0.05). Weight was negatively correlated with AoHU by 2.3HU/10kg but other predictors explained only small portions of intra-cohort variance (R² < 0.1 in contrast-enhanced scans). Average AoHU differed by contrast phase, but considerable overlap was seen between distributions. Increasing uniform plaque thresholds from 130HU to 200HU/300HU/400HU produces respective false-negative plaque content losses of 35%/60%/75% from all scans with corresponding false-positive errors in arterial-phase scans of 95%/60%/15%. Dynamic segmentation at 3SD above mean AoHU reduces false-positive errors to 0.13% and false-negative errors to 8%, 25%, and 70% in delayed, venous, and arterial scans, respectively.

Conclusion

CT contrast produces heterogeneous aortic enhancements not readily determined by demographic or scan protocol factors. Uniform CT thresholds for calcified plaques incur high rates of pixel classification errors in contrast-enhanced scans which can be minimized using dynamic thresholds based on local aorta attenuation. Care should be taken to address these errors and sex-based biases in baseline attenuation when designing automatic calcification detection algorithms intended for broad use in contrast-enhanced CTs.

Evaluation of Patients for Percutaneous Edge-to-edge Mitral Valve Repair: Comparison of Cardiac Computed Tomography Angiography With Transesophageal Echocardiography

PURPOSE

We sought to compare parameters derived from cardiac computed tomography angiography (CCTA) with those from transesophageal echocardiography (TEE) for the evaluation of patients with severe mitral regurgitation (MR) before percutaneous edge-to-edge mitral valve repair (PE2E). TEE is the mainstay for PE2E, although it has specific limitations. CCTA enables measurements in any arbitrary plane with high spatial resolution and offers good calcium visibility.

MATERIALS AND METHODS

Patients who underwent TEE and CCTA before scheduled PE2E at 2 medical centers were included in this retrospective analysis. Quantitative parameters relevant for PE2E were obtained from TEE and CCTA in a blinded manner and the intrareviewer variability was assessed.

RESULTS

All 30 patients (15 female, 76±10 y) had secondary MR attributable to ischemic (60%) or nonischemic cardiomyopathy (40%). On comparing parameters from TEE and CCTA, left ventricular end-diastolic diameter was 60±11 versus 64±11 mm (r=0.90), intercommissural mitral annulus was 35±5 versus 35±5 mm (r=0.88), long-axis annulus was 33±5 versus 33±5 mm (r=0.74), the distance between the fossa ovalis and the leaflet coaptation was 42±5 versus 41±5 mm (r=0.81), the anterior mitral leaflet was 21±6 versus 20±7 mm (r=0.81), the posterior mitral leaflet was 13±2 versus 13±2 mm (r=0.91), and the median mitral calcification was 1 (interquartile range: 0 to 2) versus 0 (interquartile range: 0 to 1; r=0.53), respectively. Intrareviewer agreement was good and excellent for continuous and categorical variables, respectively.

CONCLUSIONS

Our data suggest that evaluation of the mitral valve apparatus with CCTA in patients considered for PE2E is feasible, correlates well with TEE, and offers improved calcium visibility. In selected cases, additional information from CCTA may be helpful for achieving optimal interventional results.

Quantification of Calcium in Peripheral Arteries of the Lower Extremities: Comparison of Different CT Scanners and Scoring Platforms

OBJECTIVES

The aim of this study was to investigate the interscanner and interscoring platform variability of calcium quantification in peripheral arteries of the lower extremities.

MATERIALS AND METHODS

Twenty human fresh-frozen legs were scanned using 3 different computed tomography (CT) scanners. The radiation dose (CTDIvol) was kept similar for all scanners. The calcium scores (Agatston and volume scores) were quantified using 4 semiautomatic scoring platforms. Comparative analysis of the calcium scores between scanners and scoring platforms was performed by using the Friedman test; post hoc analysis was performed by using the Wilcoxon signed rank test with Bonferroni correction.

RESULTS

Sixteen legs had calcifications and were used for data analysis. Agatston and volume scores ranged from 12.1 to 6580 Agatston units and 18.2 to 5579 mm3. Calcium scores differed significantly between Philips IQon and Philips Brilliance 64 (Agatston: 19.5% [P = 0.001]; volume: 14.5% [P = 0.001]) and Siemens Somatom Force (Agatston: 18.1% [P = 0.001]; volume: 17.5% [P = 0.001]). The difference between Brilliance 64 and Somatom Force was smaller (Agatston: 5.6% [P = 0.778]; volume: 7.7% [P = 0.003]). With respect to the interscoring platform variability, OsiriX produced significantly different Agatston scores compared with the other 3 scoring platforms (OsiriX vs IntelliSpace: 14.8% [P = 0.001] vs Syngo CaScore: 13.9% [P = 0.001] vs iX viewer: 13.2% [P < 0.001]). For the volume score, the differences between all scoring platforms were small ranging from 2.9% to 4.0%. Post hoc analysis showed a significant difference between OsiriX and IntelliSpace (3.8% [P = 0.001]).

CONCLUSIONS

The use of different CT scanners resulted in notably different Agatston and volume scores, whereas the use of different scoring platforms resulted in limited variability especially for the volume score. In conclusion, the variability in calcium quantification was most evident between different CT scanners and for the Agatston score.

Fully Automated Artery-Specific Calcium Scoring Based on Machine Learning in Low-Dose Computed Tomography Screening

PURPOSE

 Evaluation of machine learning-based fully automated artery-specific coronary artery calcium (CAC) scoring software, using semi-automated software as a reference.

METHODS

 A total of 505 patients underwent non-contrast-enhanced calcium scoring computed tomography (CSCT). Automated, machine learning-based software quantified the Agatston score (AS), volume score (VS), and mass score (MS) of each coronary artery [right coronary artery (RCA), left main (LM), circumflex (CX) and left anterior descending (LAD)]. Identified CAC of readers who annotated the data with semi-automated software served as a reference standard. Statistics included comparisons of evaluation time, agreement of identified CAC, and comparisons of the AS, VS, and MS of the reference standard and the fully automated algorithm.

RESULTS

 The machine learning-based software correlated strongly with the reference standard for the AS, VS, and MS (Spearman's rho > 0.969) (p < 0.001), with excellent agreement (ICC > 0.919) (p < 0.001). The mean assessment time of the reference standard was 59 seconds (IQR 39-140) and that of the automated algorithm was 5.9 seconds (IQR 3.9-16) (p < 0.001). The Bland-Altman plots mean difference and 1.96 upper and lower limits of agreement for all arteries combined were: AS 0.996 (1.33 to 0.74), VS 0.995 (1.40 to 0.71), and MS 0.995 (1.35 to 0.74). The mean bias was minimal: 0.964-1.0429. Risk class assignment showed high accuracy for the AS in total (weighed κ = 0.99) and for each individual artery (κ = 0.96-0.99) with corresponding correct risk group assignment in 497 of 505 patients (98.4 %).

CONCLUSION

 The fully automated artery-specific coronary calcium scoring algorithm is a time-saving procedure and shows excellent correlation and agreement compared with the clinically established semi-automated approach.

KEY POINTS

  · Very high correlation and agreement between fully automatic and semi-automatic calcium scoring software.. · Less time-consuming than conventional semi-automatic methods.. · Excellent tool for artery-specific calcium scoring in a clinical setting..

CITATION FORMAT

· Winkelmann MT, Jacoby J, Schwemmer C et al. Fully Automated Artery-Specific Calcium Scoring Based on Machine Learning in Low-Dose Computed Tomography Screening. Fortschr Röntgenstr 2022; DOI: 10.1055/a-1717-2703.

Computed tomography-based calcium scoring in cadaver leg arteries: Influence of dose, reader, and reconstruction algorithm

PURPOSE

Computed tomography (CT) might be a good diagnostic test to accurately quantify calcium in vascular beds but there are multiple factors influencing the quantification. The aim of this study was to investigate the influence of different computed tomography protocol settings in the quantification of calcium in the lower extremities using modified Agatston and volume scores.

METHODS

Fresh-frozen human legs were scanned at different tube current protocols and reconstructed at different slice thickness. Two different iterative reconstruction protocols for conventional CT images were compared. Calcium was manually scored using modified Agatston and volume scores. Outcomes were statistically analyzed using Wilcoxon signed-rank tests and mean absolute and relative differences were plotted in Bland-Altman plots.

RESULTS

Of the 20 legs, 16 had CT detectable calcifications. Differences between thick and thin slice reconstruction protocols were 129 Agatston units and 125% for Agatston and 78.4 mm³ and 57.8% for volume (all p ≤ 0.001). No significant differences were found between low and high tube current protocols. Differences between iDose⁴ and IMR reconstruction protocols for modified Agatston were 34.2 Agatston units and 17.7% and the volume score 33.5 mm³ and 21.2% (all p ≤ 0.001).

CONCLUSIONS

Slice thickness reconstruction and reconstruction method protocols influenced the modified Agatston and volume scores in leg arteries, but tube current and different observers did not have an effect. This data emphasizes the need for standardized quantification of leg artery calcifications. Possible implications are in the development of a more universal quantification method, independent of the type of scan and vasculature.

A correction score to compare aortic calcification in contrast enhanced and non-contrast measurements from computed tomography scans

BACKGROUND

Aortic wall calcification shows strong promise as a cardiovascular risk factor. While useful for visual enhancement of vascular tissue, enhancement creates heterogeneity between scans with and without contrast. We evaluated the relationship between aortic calcification in routine abdominal computed tomography scans (CT) with and without contrast.

METHODS

Inclusion was limited to those with abdominal CT-scans with and without contrast enhancement within 120 days. Analytic Morphomics, a semi-automated computational image processing system, was used to provide standardized, granular, anatomically indexed measurements of aortic wall calcification from abdominal CT-scans. Aortic calcification area (ACA) and aortic wall calcification percent (ACP) and were the outcomes of interest. Multiple linear regression was used to evaluate the relationship of aortic measurements. Models were further controlled for age and sex. Stratification of measurements by vertebral level was also performed.

RESULTS

A positive association was observed for non-contrast calcification in ACP β 0.74 (95% CI 0.72, 0.76) and ACA β 0.44 (95% 0.43, 0.45). Stratified results demonstrated the highest coefficient of determination at L2 for percent and L3 for area models [R² 0.91 (ACP) 0.74 (ACA)]. Adjusted lumber-level associations between non-contrast and contrast measurements ranged from (β 0.69-0.82) in ACP and (β 0.37-0.54) in ACA.

CONCLUSION

A straightforward correction score for comparison of abdominal aortic calcification measurements in contrast-enhanced and non-contrast scans is discussed. Correction of aortic calcification from CT scans can reduce scan heterogeneity and will be instrumental in creating larger cardiovascular cohorts as well as cardiovascular risk surveillance programs.

Thoracic Aorta Calcium Detection and Quantification Using Convolutional Neural Networks in a Large Cohort of Intermediate-Risk Patients

Arterial calcification is an independent predictor of cardiovascular disease (CVD) events whereas thoracic aorta calcium (TAC) detection might anticipate extracoronary outcomes. In this work, we trained six convolutional neural networks (CNNs) to detect aortic calcifications and to automate the TAC score assessment in intermediate CVD risk patients. Cardiac computed tomography images from 1415 patients were analyzed together with their aortic geometry previously assessed. Orthogonal patches centered in each aortic candidate lesion were reconstructed and a dataset with 19,790 images (61% positives) was built. Three single-input 2D CNNs were trained using axial, coronal and sagittal patches together with two multi-input 2.5D CNNs combining the orthogonal patches and identifying their best regional combination (BRC) in terms of lesion location. Aortic calcifications were concentrated in the descending (66%) and aortic arch (26%) portions. The BRC of axial patches to detect ascending or aortic arch lesions and sagittal images for the descending portion had the best performance: 0.954 F1-Score, 98.4% sensitivity, 87% of the subjects correctly classified in their TAC category and an average false positive TAC score per patient of 30. A CNN that combined axial and sagittal patches depending on the candidate aortic location ensured an accurate TAC score prediction.

Head-to-Head Comparison of Different Software Solutions for AVC Quantification Using Contrast-Enhanced MDCT

Aortic valve calcification (AVC) in aortic stenosis patients has diagnostic and prognostic implications. Little is known about the interchangeability of AVC obtained from different multidetector computed tomography (MDCT) software solutions. Contrast-enhanced MDCT data sets of 50 randomly selected aortic stenosis patients were analysed using three different software vendors (3Mensio, CVI42, Syngo.Via). A subset of 10 patients were analysed twice for the estimation of intra-observer variability. Intra- and inter-observer variability were determined using the ICC reliability method, Bland-Altman analysis and coefficients of variation. No differences were revealed between the software solutions in the AVC calculations (3Mensio 941 ± 623, Syngo.Via 948 mm³ ± 655, CVI42 941 ± 637; p = 0.455). The best inter-vendor agreement was found between the CVI42 and the Syngo.Via (ICC 0.997 (CI 0.995-0.998)), followed by the 3Mensio and the CVI42 (ICC 0.996 (CI 0.922-0.998)), and the 3Mensio and the Syngo.Via (ICC 0.992 (CI 0.986-0.995)). There was excellent intra- (3Mensio: ICC 0.999 (0.995-1.000); CVI42: ICC 1.000 (0.999-1.000); Syngo.Via: ICC 0.998 (0.993-1.000)) and inter-observer variability (3Mensio: ICC 1.000 (0.999-1.000); CVI42: ICC 1.000 (1.000-1.000); Syngo.Via: ICC 0.996 (0.985-0.999)) for all software types. Contrast-enhanced MDCT-derived AVC scores are interchangeable between and reproducible within different commercially available software solutions. This is important since sufficient reproducibility, interchangeability and valid results represent prerequisites for accurate TAVR planning and its widespread clinical use.

Fully automated quantification method (FQM) of coronary calcium in an anthropomorphic phantom

Objective

Coronary artery calcium (CAC) score is a strong predictor for future adverse cardiovascular events. Anthropomorphic phantoms are often used for CAC studies on computed tomography (CT) to allow for evaluation or variation of scanning or reconstruction parameters within or across scanners against a reference standard. This often results in large number of datasets. Manual assessment of these large datasets is time consuming and cumbersome. Therefore, this study aimed to develop and validate a fully automated, open‐source quantification method (FQM) for coronary calcium in a standardized phantom.

Materials and Methods

A standard, commercially available anthropomorphic thorax phantom was used with an insert containing nine calcifications with different sizes and densities. To simulate two different patient sizes, an extension ring was used. Image data were acquired with four state‐of‐the‐art CT systems using routine CAC scoring acquisition protocols. For interscan variability, each acquisition was repeated five times with small translations and/or rotations. Vendor‐specific CAC scores (Agatston, volume, and mass) were calculated as reference scores using vendor‐specific software. Both the international standard CAC quantification methods as well as vendor‐specific adjustments were implemented in FQM. Reference and FQM scores were compared using Bland‐Altman analysis, intraclass correlation coefficients, risk reclassifications, and Cohen’s kappa. Also, robustness of FQM was assessed using varied acquisitions and reconstruction settings and validation on a dynamic phantom. Further, image quality metrics were implemented: noise power spectrum, task transfer function, and contrast‐ and signal‐to‐noise ratio among others. Results were validated using imQuest software.

Results

Three parameters in CAC scoring methods varied among the different vendor‐specific software packages: the Hounsfield unit (HU) threshold, the minimum area used to designate a group of voxels as calcium, and the usage of isotropic voxels for the volume score. The FQM was in high agreement with vendor‐specific scores and ICC’s (median [95% CI]) were excellent (1.000 [0.999‐1.000] to 1.000 [1.000‐1.000]). An excellent interplatform reliability of κ = 0.969 and κ = 0.973 was found. TTF results gave a maximum deviation of 3.8% and NPS results were comparable to imQuest.

Conclusions

We developed a fully automated, open‐source, robust method to quantify CAC on CT scans in a commercially available phantom. Also, the automated algorithm contains image quality assessment for fast comparison of differences in acquisition and reconstruction parameters.

End-to-End, Pixel-Wise Vessel-Specific Coronary and Aortic Calcium Detection and Scoring Using Deep Learning

Conventional scoring and identification methods for coronary artery calcium (CAC) and aortic calcium (AC) result in information loss from the original image and can be time-consuming. In this study, we sought to demonstrate an end-to-end deep learning model as an alternative to the conventional methods. Scans of 377 patients with no history of coronary artery disease (CAD) were obtained and annotated. A deep learning model was trained, tested and validated in a 60:20:20 split. Within the cohort, mean age was 64.2 ± 9.8 years, and 33% were female. Left anterior descending, right coronary artery, left circumflex, triple vessel, and aortic calcifications were present in 74.87%, 55.82%, 57.41%, 46.03%, and 85.41% of patients respectively. An overall Dice score of 0.952 (interquartile range 0.921, 0.981) was achieved. Stratified by subgroups, there was no difference between male (0.948, interquartile range 0.920, 0.981) and female (0.965, interquartile range 0.933, 0.980) patients (p = 0.350), or, between age <65 (0.950, interquartile range 0.913, 0.981) and age ≥65 (0.957, interquartile range 0.930, 0.9778) (p = 0.742). There was good correlation and agreement for CAC prediction (rho = 0.876, p < 0.001), with a mean difference of 11.2% (p = 0.100). AC correlated well (rho = 0.947, p < 0.001), with a mean difference of 9% (p = 0.070). Automated segmentation took approximately 4 s per patient. Taken together, the deep-end learning model was able to robustly identify vessel-specific CAC and AC with high accuracy, and predict Agatston scores that correlated well with manual annotation, facilitating application into areas of research and clinical importance.

Mitral annular calcification in the elderly - Quantitative assessment

PURPOSE

To determine the reliability of subjective and objective quantification of mitral annular calcification (MAC) in elderly patients with severe aortic stenosis, to define quantitative sex- and age-related reference values of MAC, and to correlate quantitative MAC with mitral valve disease.

METHODS

In this retrospective, IRB-approved study, we included 559 patients (268 females, median age 81 years, inter-quartile range 77-85 years) with severe aortic stenosis undergoing CT. Four independent readers performed subjective MAC categorization as follows: no, mild, moderate, and severe MAC. Two independent readers performed quantitative evaluation of MAC using the Agatston score method (Agatston_MAC). Mitral valve disease was determined by echocardiography.

RESULTS

Subjective MAC categorization showed high inter-reader agreement for no (k ​= ​0.88) and severe MAC (k ​= ​0.75), whereas agreement for moderate (k ​= ​0.59) and mild (k ​= ​0.45) MAC was moderate. Intra-reader agreement for subjective MAC categorization was substantial (k ​= ​0.69 and 0.62). Inter- and intra-reader agreement for Agatston_MAC were excellent (ICC ​= ​0.998 and 0.999, respectively), with minor inconsistencies in MAC involving the left ventricular outflow tract/aortic valve. There were significantly more women than men with MAC (n ​= ​227, 85% versus n ​= ​209, 72%; p ​< ​0.001), with a significantly higher Agatston_MAC (median 597, range 81-2055 versus median 244; range 0-1565; p ​< ​0.001), particularly in patients ≥85 years of age. Agatston_MAC showed an area-under-the-curve of 0.84 to diagnose mitral stenosis, whereas there was no association of Agatston_MAC with mitral regurgitation (p ​> ​0.05).

CONCLUSIONS

Our study in elderly patients with severe aortic stenosis shows that quantitative MAC scoring is more reliable than subjective MAC assessment. Women show higher Agatston_MAC scores than men, particularly in the elderly population. Agatston_MAC shows high accuracy to diagnose mitral stenosis.

Reproducibility of calcium scoring of the coronary arteries: comparison between different vendors and iterative reconstructions

Background

The coronary artery calcium scoring (CCS) has been widely used for cardiac risk stratification for asymptomatic patients.

Purpose

To assess the reproducibility of CCS performed on four different computed tomography (CT) scanners, and compare the variability between two reconstruction algorithms, filtered back projection (FBP), and iterative reconstruction (IR).

Material and Methods

A CCS phantom was made from agar and contained 23 pieces of chicken bones. The phantom was repeatedly scanned using four different CT scanners: Toshiba; GE; Philips; and Siemens. Images were reconstructed using FBP and IR. Agatston and volume scores of total bone fragments were calculated and the overall differences between the instruments were evaluated using the Friedman test. Comparison of the Agatston and volume scores between the two reconstruction algorithms, for each instrument, was evaluated using the Wilcoxon signed rank test.

Results

The difference in the Agatston scores was significantly different between the four machines (P = 0.001). The Toshiba scanner yielded the highest score followed by Philips, GE, and Siemens scanners. There was no difference in the CCS evaluated using the two reconstruction algorithms, except in case of the Siemens scanner (P = 0.032).

Conclusion

CCS performed on different scanners varied significantly. In the Toshiba, Philips, and GE scanners, there was no significant difference in the CCS determined using either an IR or the FBP algorithm. In the Siemens scanner, applying the IR algorithm resulted in a slightly different scores, which might not be clinically significant.

Evaluation of an AI-based, automatic coronary artery calcium scoring software

Objectives

To evaluate an artificial intelligence (AI)–based, automatic coronary artery calcium (CAC) scoring software, using a semi-automatic software as a reference.

Methods

This observational study included 315 consecutive, non-contrast-enhanced calcium scoring computed tomography (CSCT) scans. A semi-automatic and an automatic software obtained the Agatston score (AS), the volume score (VS), the mass score (MS), and the number of calcified coronary lesions. Semi-automatic and automatic analysis time were registered, including a manual double-check of the automatic results. Statistical analyses were Spearman’s rank correlation coefficient (⍴), intra-class correlation (ICC), Bland Altman plots, weighted kappa analysis (κ), and Wilcoxon signed-rank test.

Results

The correlation and agreement for the AS, VS, and MS were ⍴ = 0.935, 0.932, 0.934 (p < 0.001), and ICC = 0.996, 0.996, 0.991, respectively (p < 0.001). The correlation and agreement for the number of calcified lesions were ⍴ = 0.903 and ICC = 0.977 (p < 0.001), respectively. The Bland Altman mean difference and 1.96 SD upper and lower limits of agreements for the AS, VS, and MS were − 8.2 (− 115.1 to 98.2), − 7.4 (− 93.9 to 79.1), and − 3.8 (− 33.6 to 25.9), respectively. Agreement in risk category assignment was 89.5% and κ = 0.919 (p < 0.001). The median time for the semi-automatic and automatic method was 59 s (IQR 35–100) and 36 s (IQR 29–49), respectively (p < 0.001).

Conclusions

There was an excellent correlation and agreement between the automatic software and the semi-automatic software for three CAC scores and the number of calcified lesions. Risk category classification was accurate but showing an overestimation bias tendency. Also, the automatic method was less time-demanding.

Key Points

• Coronary artery calcium (CAC) scoring is an excellent candidate for artificial intelligence (AI) development in a clinical setting.

• An AI-based, automatic software obtained CAC scores with excellent correlation and agreement compared with a conventional method but was less time-consuming.

Current and future applications of CT coronary calcium assessment

INTRODUCTION

Computed tomographic (CT) coronary artery calcium scoring (CAC) has been validated as a well-established screening method for cardiovascular risk stratification and treatment management that is used in addition to traditional risk factors. The purpose of this review is to present an update on current and future applications of CAC. Areas covered: The topic of CAC is summarized from its introduction to current application with focus on the validation and clinical integration including cardiovascular risk prediction and outcome, cost-effectiveness, impact on downstream medical testing, and the technical advances in scanner and software technology that are shaping the future of CAC. Furthermore, this review aims to provide guidance for the appropriate clinical use of CAC. Expert commentary: CAC is a well-established screening test in preventive care that is underused in daily clinical practice. The widespread clinical implementation of CAC will be decided by future technical advances in CT image acquisition, cost-effectiveness, and reimbursement status.

Quantification of abdominal aortic calcification: Inherent measurement errors in current computed tomography imaging

Objective

Quantification software for coronary calcification is often used to measure abdominal aortic calcification on computed tomography (CT) images. However, there is no evidence substantiating the reliability and accuracy of these tools in this setting. Differences in coronary and abdominal CT acquisition and presence of intravascular contrast may affect the results of these tools. Therefore, this study investigates the effects of CT acquisition parameters and iodine contrast on automated quantification of aortic calcium on CT.

Methods

Calcium scores, provided in volume and mass, were assessed by automated calcium quantification software on CT scans. First, differences in calcium scores between the abdominal and coronary CT scanning protocols were assessed by imaging a thorax phantom containing calcifications of 9 metrical variations. Second, aortic calcification was quantified in 50 unenhanced and contrast-enhanced clinical abdominal CT scans at a calcification threshold of 299 Hounsfield Units (HU). Also, the lowest possible HU threshold for calcifications was calculated per individual patient and compared to a 130 HU threshold between contrast-enhanced and unenhanced CT images, respectively.

Results

No significant differences in volume and mass scores between the abdominal and the coronary CT protocol were found. However, volume and mass of all calcifications were overestimated compared to the physical volume and mass (volume range: 0–649%; mass range: 0–2619%). In comparing unenhanced versus contrast-enhanced CT images showed significant volume differences for both thresholds, as well as for mass differences for the 130 vs patient-specific threshold (230 ± 22.6 HU).

Conclusion

Calcification scoring on CT angiography tends to grossly overestimate volume and mass suggesting a low accuracy and reliability. These are reduced further by interference of intravascular contrast. Future studies applying calcium quantification tools on CT angiography imaging should acknowledge these issues and apply corrective measures to ensure the validity of their outcomes.

Coronary Artery Calcium Imaging in the ROBINSCA Trial: Rationale, Design, and Technical Background

RATIONALE AND OBJECTIVES

To describe the rationale, design, and technical background of coronary artery calcium (CAC) imaging in the large-scale population-based cardiovascular disease screening trial (Risk Or Benefit IN Screening for CArdiovascular Diseases [ROBINSCA]).

MATERIALS AND METHODS

First, literature search was performed to review the logistics, setup, and settings of previously performed CAC imaging studies, and current clinical CAC imaging protocols of participating centers in the ROBINSCA trial were evaluated. A second literature search was performed to evaluate the impact of computed tomography parameter settings on CAC score.

RESULTS

Based on literature reviews and experts opinion an imaging protocol accompanied by data management protocol was created for ROBINSCA. The imaging protocol should consist of a fixed tube voltage, individually tailored tube current setting, mid-diastolic electrocardiography-triggering, fixed field-of-view, fixed reconstruction kernel, fixed slice thickness, overlapping reconstruction and without iterative reconstruction. The analysis of scans is performed with one type and version of CAC scoring software, by two dedicated and experienced researchers. The data management protocol describes the organization of data handling between the coordinating center, participating centers, and core analysis center.

CONCLUSION

In this paper we describe the rationale and technical considerations to be taken in developing CAC imaging protocol, and we present a detailed protocol that can be implemented for CAC screening purposes.

Influence of heart rate on coronary calcium scores: a multi-manufacturer phantom study

To evaluate the influence of heart rate on coronary calcium scores (CCS) using a dynamic phantom on four high-end computed tomography (CT) systems from different manufacturers. Artificial coronary arteries were moved in an anthropomorphic chest phantom at linear velocities, corresponding to < 60, 60-75 and > 75 beats per minute (bpm). Data was acquired with routinely used clinical protocols for CCS on four high-end CT systems (CT1-CT4). CCS, quantified as Agatston and mass scores were compared to reference scores at < 60 bpm. Influence of heart rate was assessed for each system with the cardiac motion susceptibility (CMS) Index. At increased heart rates (> 75 bpm), Agatston scores of the low mass calcification were similar to the reference score, while Agatston scores of the medium and high mass calcification increased significantly up to 50% for all CT systems. Threefold CMS increases at > 75 bpm in comparison with < 60 bpm were shown. For medium and high mass calcifications, significant differences in CMS between CT systems were found. Heart rate substantially influences CCS for high-end CT systems of four major manufacturers, but CT systems differ in motion susceptibility. Follow-up CCS CT scans should be acquired on the same CT system and protocol, and preferably with comparable heart rates.

Prevalence and severity of coronary artery calcification based on the epidemiologic pattern: A propensity matched comparison of asymptomatic Korean and Chinese adults

BACKGROUND

Lifestyle, environmental, and genetic factors substantially influence cardiovascular disease (CVD) risk. We aimed to explore epidemiologic trends in coronary artery calcium scores (CACS), as a marker of CVD, along with possible differences by geographic area and study period in separate East Asian populations.

METHODS

We generated 3 matched groups (n=702) using a propensity scoring approach derived from a Korean (N=48,901) and Chinese cohort (N=927) as follows: (1) A recent Chinese group and (2) recent Korean group, both of whom underwent CACS scanning from 2012-2014; and (3) a past Korean group who underwent CACS scanning 8-10years before the index group (2002-2006). We used logistic regression to generate odds ratios (OR) with 95% confidence intervals (95% CI) to estimate the likelihood of having CACS between the groups, based on CACS stratified by severity: >0 (any), >100 (moderate), and >400 (severe).

RESULTS

The prevalence of any, moderate, or severe CACS did not differ significantly between the recent Chinese and Korean groups. Notably, the odds of the presence of moderate CACS in the recent Chinese group (OR: 3.05, 95% CI: 1.49-6.71, P-value<0.001) and the presence of any CACS in the recent Korean group (OR: 1.58, 95% CI: 1.17-2.15, P-value<0.001) were significantly higher than in the past Korean group.

CONCLUSIONS

In this study involving separate East Asian populations, there were no geographic differences in the prevalence of CACS. However, changes in other unmeasured factors over time are likely the culprits for the elevated prevalence of CACS in asymptomatic East Asians.

Impact of Advanced Modeled Iterative Reconstruction on Coronary Artery Calcium Quantification

RATIONALE AND OBJECTIVES

To evaluate the influence of advanced modeled iterative reconstruction (ADMIRE) on the coronary artery calcium (CAC) scores by computed tomography (CT).

MATERIALS AND METHODS

Sixty patients underwent CAC imaging with dual-source 192-slice CT. Agatston, volume and mass score were calculated from filtered back projection (FBP) and iterative reconstructions with different levels of ADMIRE. Friedman test and Wilcoxon rank sum test were used for multiple comparisons of CAC values and the difference ratio among different ADMIRE groups using FBP as reference.

RESULTS

The median Agatston score (range) using FBP was 115 (0.1-3047) and significantly decreased with incremental ADMIRE levels 1-5: 96 (0.1-2813), 91 (0-2764), 87 (0-2699), 80 (0-2590), 70 (0-2440); all P < 0.001. In comparison with FBP Agatston, volume and mass scores significantly decreased with increasing ADMIRE levels 1-5 (P < 0.001): from -12% to -39%, from -14% to -41%, and from -13% to -40%, respectively. In four patients with low calcium burden, the use of ADMIRE 2 or higher resulted in the disappearance of calcium that was detectable using FBP or ADMIRE 1. The decrease of CAC in high-level ADMIRE resulted in a reassignment to a lower Agatston risk group in 27%.

CONCLUSIONS

ADMIRE causes a substantial reduction of the CAC scores measured by cardiac CT, which leads to an underestimation of cardiovascular risk scores in some patients.

Prevalence and Distribution of Coronary Artery Calcification in Asymptomatic United States and Korean Adults - Cross-Sectional Propensity-Matched Analysis

BACKGROUND

The incidence of coronary artery disease (CAD) varies depending on ethnicity, but the precise differences remain to be firmly established. This study therefore evaluated the disparity in coronary artery calcification (CAC), as a marker of CAD, in asymptomatic US and Korean adults.Methods and Results:CAC score was compared between asymptomatic Korean (n=15,128) and US (n=7,533) adults. Propensity score matching was performed according to age, gender, hypertension, diabetes, dyslipidemia, and current smoking, which generated 2 cohorts of 5,427 matched pairs. Both cohorts were categorized according to age group: 45-54, 55-64, and 65-74 years. Overall, the prevalence of CAC score >0, >100, and >400 in Korean adults was lower than in US adults (P<0.001, all). According to increasing age groups, the likelihood of CAC was most often lower in Korean adults, especially in Korean women. The odds of having CAC >400 in Korean adults aged 65-74 years was 0.66 (95% CI: 0.48-0.91) overall, 0.78 (95% CI: 0.52-1.19) in men, and 0.50 (95% CI: 0.29-0.86) in women, compared with US counterparts.

CONCLUSIONS

Korean adults have a lower prevalence and severity of atherosclerotic burden as assessed on CAC, compared with US adults, but the disparity in CAC according to ethnicity may decline with older age. (Circ J 2016; 80: 2349-2355).

Feasibility of spectral shaping for detection and quantification of coronary calcifications in ultra-low dose CT

Objectives

To evaluate detectability and quantification of coronary calcifications for CT with a tin filter for spectral shaping.

Methods

Phantom inserts with 100 small and 9 large calcifications, and a moving artificial artery with 3 calcifications (speed 0–30 mm/s) were placed in a thorax phantom simulating different patient sizes. The phantom was scanned in high-pitch spiral mode at 100 kVp with tin filter (Sn100 kVp), and at a reference of 120 kVp, with electrocardiographic (ECG) gating. Detectability and quantification of calcifications were analyzed for standard (130 HU) and adapted thresholds.

Results

Sn100 kVp yielded lower detectability of calcifications (9 % versus 12 %, p = 0.027) and lower Agatston scores (p < 0.008), irrespective of calcification, patient size and speed. Volume scores of the moving calcifications for Sn100 kVp at speed 10–30 mm/s were lower (p < 0.001), while mass scores were similar (p = 0.131). For Sn100 kVp with adapted threshold of 117 HU, detectability (p = 1.000) and Agatston score (p > 0.206) were similar to 120 kVp. Spectral shaping resulted in median dose reduction of 62.3 % (range 59.0–73.4 %).

Conclusions

Coronary calcium scanning with spectral shaping yields lower detectability of calcifications and lower Agatston scores compared to 120 kVp scanning, for which a HU threshold correction should be developed.

Key points

• Sn100kVp yields lower detectability and lower Agatston scores compared to 120kVp

• Adapted HU threshold for Sn100kVp provides Agatston scores comparable to 120kVp

• Sn100 kVp considerably reduces dose in calcium scoring versus 120 kVp

Combined effects of exercise capacity and coronary atherosclerotic burden on all-cause mortality in asymptomatic Koreans

BACKGROUND AND AIMS

Both exercise capacity and coronary artery calcium score (CACS) are important prognostic factors in cardiovascular outcome. Yet, whether there is a significant interaction between these two factors in influencing clinical outcome is still uncertain. This study investigated the combined effects of exercise capacity and CACS on all-cause mortality in an asymptomatic population.

METHODS

From multicenter registry of health screening, a retrospective cohort of 25,972 asymptomatic subjects, who underwent both CACS and treadmill exercise test, was included in the final dataset for analysis. Outcome was defined as all-cause mortality, which was obtained from national mortality registry.

RESULTS

The mean age of study subjects was 53.7 ± 7.7 years and 81.5% of them were males. Median follow-up duration was 5.5 (IQR 3.6-7.5) years and 226 (0.9%) cases of all-cause mortality occurred. In multivariate Cox's proportional hazard model with interaction term, exercise capacity ≥10 METs (HR 0.684, 95% CI 0.483-0.971) and CACS ≥400 (HR 3.328, 95% CI 1.850-5.988) were significant predictors of all-cause mortality. In patients with higher exercise capacity, the effect of high CACS on all-cause mortality was significantly smaller than in those with lower exercise capacity. The HR for all-cause mortality of CACS ≥400, in those with lower exercise capacity, is estimated to be about three times of that in those with higher exercise capacity (HR 3.328 in <10 METs vs. 1.108 in ≥10 METs, p for interaction = 0.024) after adjustment for age, gender, fasting glucose, creatinine, alanine transaminase and albumin.

CONCLUSIONS

The effect of high CACS on all-cause mortality is lessened by good exercise capacity in the asymptomatic population. Good physical fitness may reduce the adverse effect of high coronary atherosclerotic burden.

Quantification of Aortic Valve Calcifications Detected During Lung Cancer-Screening CT Helps Stratify Subjects Necessitating Echocardiography for Aortic Stenosis Diagnosis

No study has been published on aortic valve calcification (AVC) extent at lung cancer screening low-dose CT (LDCT) and its relationship with aortic stenosis (AS). The purpose of this study was to estimate the cutoff value of AVC on LDCT for detecting AS in asymptomatic Asian subjects. Six thousand three hundred thirty-eight subjects (mean age, 55.9 years ± 8.6) self-referred to health-promotion center underwent LDCT, coronary calcium scoring CT (CSCT), and echocardiography. AVC was quantified using Agatston methods on CT. AVC extent on LDCT was compared with that on CSCT, and AVC threshold for diagnosing AS was calculated. Clinical factors associated with AS and AVC were sought.AVC was observed in 403 subjects (64.9 years ± 8.7) on LDCT (6.4%), and AVC score measured from LDCT showed strong positive correlation with that from CSCT (r = 0.83, P < 0.0001). Of 403 subjects, 40 (10%) were identified to have AS on echocardiography. Cutoff value of AVC score for detecting AS was 138.37 with sensitivity of 90.0% and specificity 83.2%. On multivariate analysis, age (odds ratio [OR] = 1.10, 95% CI: 1.09-1.12) and hypertension (OR = 1.39, 95% CI: 1.10-1.76) were associated with the presence of AVC, whereas AVC extent at LDCT (OR = 104.32, 95% CI: 16.16-673.70) was the only significant clinical factor associated with AS; AVC extent on LDCT (OR = 104.32, 95% CI: 16.16-673.70) was the significant clinical factor associated with AS.The AVC extent on LDCT is significantly related to the presence of AS, and we recommend echocardiography for screening AS based on quantified AVC values on LDCT.