| Part Ⅰ Diagnostic performance of coronary CT angiography for the assessment of coronary stenosis in the presence of calcium using a dual-layer spectral detector CTObjective:To explore the diagnostic value of spectral reconstructions(virtual monoenergetic image,effective atomic number image and iodine density image)for the quantification of calcification-related coronary stenosis using a dual-layer spectral detector CT.Methods and materials:From July 2016 to November 2017,consecutive patients with known or suspected coronary artery disease who underwent coronary CT angiography(CCTA)on dual-layer spectral detector CT and invasive coronary angiography(ICA)in Peking Union Medical College Hospital were retrospectively enrolled.Conventional 120 kVp images,70 to 140 keV virtual monoenergetic images(at increment of 10 keV),the effective atomic number images and iodine density images were reconstructed.Bland-Altman analysis was used to evaluate the consistency of the results on different virtual monoenergetic images and ICA.Using ICA as the reference standard,the sensitivity,specificity,positive predictive value,negative predictive value and diagnostic accuracy of the optimal virtual monoenergetic images,effective atomic number images and iodine density images in detecting ≥50%or ≥70%calcification-related coronary stenosis were calculated.Parallel and serial testing were used to evaluate the incremental value of the effective atomic number or iodine density images to the best virtual monoenergetic images.Patients were then grouped according to Agatston score,and the diagnostic value of the combination test was analyzed for patients with ≥400 or ≥1000 Agatston score.Results:A total of 122 coronary lesions of 72 patients(49 men and 23 women;63.7±10.2 years)were enrolled in analysis.Bland-Altman analysis showed the highest consistency between 100 keV images and ICA results.The sensitivity,specificity,positive predictive value,negative predictive value and diagnostic accuracy of 100 keV images to identify ≥50%or ≥70%calcification-related coronary stenosis were 84%,70%,80%,76%,79%and 78%,98%,93%,91%,92%,respectively.The results of combined diagnostic tests showed that the combination of effective atomic number images and optimal virtual monoenergetic images achieved the highest diagnostic performance.Compared to using 100keV images alone,a serial combination(100 keV virtual monoenergetic images followed by the effective atomic number images)resulted in an improved specificity(from 70%to 80%)with a moderate loss of sensitivity(81%from 84%)in identifying ≥50%stenosis(P=0.021).For patients with ≥400 or≥1000 Agatston score,this combination could further reduce false positive cases and improve diagnostic accuracy.Conclusion:100 keV images provide optimal diagnostic performance for the detection of calcification-related coronary stenosis using a dual-layer spectral detector CT,with further improvements obtained with the combined use of the effective atomic number images.Part Ⅱ Automatic detection of coronary plaque and quantification of stenosis degree on virtual monoenergetic images of spectral detector CTObjective:To explore the value of a deep learning algorithm combined with virtual monoenergetic images for automatically identifying coronary plaque and stenosis.Methods and materials:From July 2016 to November 2017,consecutive patients with known or suspected coronary artery disease who underwent coronary CT angiography(CCTA)on dual-layer spectral detector CT and invasive coronary angiography(ICA)in Peking Union Medical College Hospital were retrospectively enrolled.The conventional 120 kVp images and 40 to 140 keV virtual monoenergetic images(at increment of 10 keV)were reconstructed.A deep learning-based algorithm automatically identified coronary plaques and calculated the degree of stenosis on conventional 120 kVp images and virtual monoenergetic images,which was further compared with ICA results.The overall assessment of deep learning algorithm in identifying coronary stenosis was performed at a segment,vessel and patient level,the plaque-specific assessment was performed at vessel level.Results:A total of 71 patients(52 men,19 women;63.3± 10.7 years)were enrolled in analysis.The median Agatston score was 532.8(quartile Q1-Q3:133.6-1180.7).Patient and vessel based analyses showed virtual monoenergetic images from 50 to 90 keV achieved optimal overall diagnostic performance,but revealed no significant difference with conventional 120 kVp images.The average sensitivity,specificity and diagnostic accuracy of 50 to 90 keV were 71.6%,90.7%and 81.6%for vessel-based analysis,and 92.4%,89.8%and 91.8%for patient-based analysis.Segment-based analysis showed no statistically significant difference with different virtual monoenergetic images and conventional 120kVp images in diagnostic performance.For plaque-based assessment,diagnostic performance of deep learning algorithm on 50-100 keV achieved optimal diagnostic performance in detecting stenosis caused by calcification,but revealed no significant difference with conventional 120 kVp images(91.4%vs.93.8%,P>0.05).For virtual monoenergetic images above 100 keV,the diagnostic accuracy dropped significantly.Conclusion:The overall performance of a deep learning algorithm based on 50-90keV virtual monoenergetic images outperformed other image levels,and the optimal image levels were 50-100keV for calcification related stenosis,but there was no significant difference between optimal image levels and conventional 120kVp imagesPart Ⅲ Diagnostic value of deep learning reconstruction-based subtraction CCTA for assessing calcification-related coronary stenosisObjective:To explore the impact of deep learning reconstruction(DLR)combined with subtraction technology on image quality of coronary CT angiography(CCTA)and subtraction CCTA,and to explore the diagnostic value in detecting calcification-related coronary stenosis.Methods and materials:From February 2020 to April 2021,consecutive patients with known or suspected coronary artery disease in Peking Union Medical College Hospital who underwent subtraction CCTA on a 320-rows CT scanner and subsequent invasive coronary angiography(ICA)witihin one month were prospectively enrolled.A two-breath-hold subtraction CCTA protocol was used.Based on hybrid iterative reconstruction(HIR)and DLR algorithm,images were reconstructed and subtracted.Four image groups including HIR based CCTA,HIR based subtraction CCTA,DLR based CCTA and DLR based subtraction CCTA were obtained.Subjective images quality comparison were performed by using a Likert 4 stage score and objective images quality parameters including image noise,signal-to-noise ratio and contrast-to-noise ratio were calculated.Using ICA as golden standard,the diagnostic performance of four image groups in the detecting obstructive coronary artery disease caused by calcified lesions was assessed and compared.Interobserver agreement was assessed by using the Cohen’s kappa coefficient.Results:There were 166 lesions in 42 patients(32 men and 10 women;62.9±9.3years)finally enrolled for analysis.The subjective and objective image quality of DLR based CCTA or subtraction CCTA were superior to those of HIR based CCTA or subtraction CCTA.The image noise of DLR based subtraction CCTA,DLR based CCTA,HIR based subtraction CCTA and HIR based CCTA were 25.25±4.43,18.00±3.62,34.23±7.64 and 26.56±4.27.Lesion based analysis showed the diagnostic accuracy of DLR based subtraction CCTA,DLR based CCTA,HIR based subtraction CCTA and HIR based CCTA to identify calcification-related obstructive diameter stenosis were 83.7%,69.3%,75.3%and 65.7%,respectively.The false positive rate of DLR based subtraction CCTA,DLR based CCTA,HIR based subtraction CCTA and HIR based CCTA for luminal diameter stenosis ≥50%were 15%,31%,24%and 34%,respectively.DLR based subtraction CCTA achieved optimal diagnostic performance among four image groups,the sensitivity and specificity to identify ≥50%luminal diameter stenosis were 90.9%and 83.2%.Interobserver consistency was good(kappa=0.82).Conclusion:DLR could improve image quality of CCTA and subtraction CCTA images significantly,DLR combined with subtraction technique could improve the diagnostic performance of CCTA for calcification-related obstructive coronary stenosis,which has good clinical application value.Part Ⅳ Diagnostic value of deep learning reconstruction-based subtraction CT-FFR in calcification-related hemodynamic stenosisObjective:To explore the impact of deep learning reconstruction(DLR)on image quality of coronary CT angiography(CCTA)and CCTA-derived fractional flow reserve(CT-FFR)values;and to explore the diagnostic value of subtraction CT-FFR combined with DLR in detecting calcification-related hemodynamically significant stenosis.Methods and materials:From March 2020 to March 2022,consecutive patients with known or suspected CAD in Peking Union Medical College Hospital who were scheduled for invasive coronary angiography(ICA)were included prospectively.Subtraction CCTA were provided to all patients and fractional flow revers(FFR)was measured during ICA in patients with moderate-to-severe coronary artery stenosis caused by calcification.Five image groups including filtered back projection(FBP),statistical-based iterative reconstruction(SBIR),model-based iterative reconstruction(MBIR)Cardiac,MBIR Cardiac sharp and DLR were reconstructed.The image quality and CT-FFR values based on different reconstruction approaches were compared.Image subtraction was performed on DLR-based CCTA image,and subtraction CT-FFR was measured based on subtraction image.Using FFR as the reference standard,the diagnostic performance of DLR-based CCTA,subtraction CCTA,CT-FFR,and subtraction CT-FFR in detecting calcification-related hemodynamically significant stenosis were evaluated.An FFR value of 0.8 or less was considered as hemodynamically significant stenosis.Results:A total of 33 patients(26 men and 10 women;63.2±8.3 years)with 182 lesions were included in this study.Among five reconstruction approaches,the image quality of DLR was superior to the others,with lowest image noise and highest signal to noise ratio(SNR)and contrast to noise ratio(CNR).There were no statistically significant differences in the CT-FFR values among these five approaches.The mean CT-FFR values were 0.86±0.14,0.87±0.13,0.85±0.13,0.87±0.13 and 0.87±0.14 for FBP,SBIR,MBIR Cardiac,MBIR Cardiac sharp and DLR images,respectively(P>0.05).Of the 33 patients,5 had no moderate-severe stenosis caused by coronary artery calcification,thus 52 calcification lesions of 28 patients were included in the diagnostic performance analysis.The diagnostic performance of CT-FFR was better than CCTA before and after subtraction,subtraction CT-FFR outperformed other approaches in detecting calcification-related hemodynamically significant stenosis.Lesion-based analysis showed area under curves(AUCs)for subtraction CT-FFR,CT-FFR,subtraction CCTA and CCTA were 0.90,0.70,0.59 and 0.61,respectively.Subtraction CT-FFR showed significantly higher sensitivity(100.0%vs.76.5%,P<0.05),specificity(71.4%vs.60.0%,P<0.05),positive predictive value(63.0%vs.48.2%,P<0.05),negative predictive value(100.0%vs.84.0%,P<0.05),and accuracy(80.8%vs.65.4%,P<0.05)than CT-FFR.Patient-based analysis showed AUCs for subtraction CT-FFR,CT-FFR,subtraction CCTA and CCTA were 0.77,0.47,0.46 and 0.50,respectively.Subtraction CT-FFR showed significantly higher sensitivity(100.0%vs.84.6%,P<0.05),specificity(33.3%vs.20.0%,P<0.05),positive predictive value(56.5%vs.47.8%,P<0.05),negative predictive value(100.0%vs.60.0%,P<0.05),and accuracy(64.3%vs.50.0%,P<0.05)than CT-FFR.Conclusion:DLR-based CCTA achieved optimal image quality but there were no significant differences in the CT-FFR values based on different reconstruction approaches.DLR-based subtraction CT-FFR improved diagnostic performance for calcification-related hemodynamically significant stenosis. |
