Synthetic Extracellular Volume in Cardiac Magnetic Resonance Without Blood Sampling: a Reliable Tool to Replace Conventional Extracellular Volume (2024)

Clinical Perspective

Cardiac magnetic resonance (CMR)-extracellular volume (ECV) mapping is the only noninvasive technique proven to provide a stable indicator of diffuse myocardial fibrosis. It has significant value in the clinical management and prognostic assessment of multiple diseases. However, the promotion of conventional ECV is limited by the need for blood tests at the time of scanning and the high variability of hematocrit. Synthetic ECV obtained using T1 time of the native blood pool omits the process of blood testing, while the results of synthetic hematocrit are in real time, thus avoiding apparent fluctuations in hematocrit due to time, body position, or other reasons. Excitingly, the results of synthetic ECV are almost identical to those of conventional ECV compared with conventional methods, which provides excellent efficiency in the assessment of diffuse CMR fibrosis. In our study, 2 types of models were derived: the specific model can be used in scientific studies or anemic patients because of its highest accuracy; the common model can be used for routine clinical practice due to its high efficiency and practicability. Synthetic ECV increases the potential of ECV in routine clinical CMR and allows for rapid clinical decision-making. Besides, the synthetic ECV provides a means to generate real-time ECV mapping during the scan automatically, an online tool on a CMR scanner to create an instant fully automated ECV map could be implemented on different CMR vendor platforms, which will be an attractive and promising development for each center that could greatly facilitate the broader use of CMR-ECV.

Patient Population

This study retrospectively analyzed 1101 patients who underwent CMR scans between August 2014 and November 2020 at German Heart Center Berlin that had pre- and postcontrast T1 mapping and point of care or laboratory hematocrit measurements from blood samples taken within 24 hours of CMR scanning available. Of these, 652 underwent CMR scans at 3.0T and 449 at 1.5T (Figure ​(Figure1).1). Additional clinical information including age, sex, and cardiac diagnoses was collected. Patients on both scanners were randomly split into equally sized derivation (n=550) and validation (n=551) subgroups. This study was approved by the ethics informed committee of the Charite-Universitatsmedizin Berlin (Ethic number: EA2/073/21), and all the subjects in the study gave informed consent.

CMR Protocol

Six hundred fifty-two subjects underwent CMR at a clinical 3T MRI scanner (Ingenia, Philips Healthcare, Best, The Netherlands) MRI with an anterior- and the built-in posterior coil array, where up to 30 coil elements were employed. Four hundred forty-nine subjects were examined at a 1.5T MRI (Achieva, Philips Healthcare, Best, The Netherlands) equipped with a cardiac 5-element phased array coil. Cine images were acquired using retrospectively gated cine-CMR in cardiac short-axis, vertical long-axis, and horizontal long-axis orientations using a balanced steady-state free precession sequence. Native and 15-min post-contrast T1 mapping was performed using a modified Look-Locker (MOLLI) 5s(3s)3s-scheme. Typical imaging parameters were as follows: Acquired voxel size=2.0×2.0×10 mm³, reconstructed voxel size=0.5×0.5×10 mm³, balanced steady-state free precession readout, flip angle=35◦, parallel imaging (SENSE) factor=2 and effective inversion times between 150 and 3382 ms. Patients received 0.15 mmol/kg of gadolinium-based contrast agent (Gadobutrol 1.0 mmol/mL, Gadovist, BayerAG, Leverkusen, Germany).

CMR Analysis

Image analysis was performed offline using commercially available postprocessing software (Philips Intellispace Portal, Philips Medical Systems Nederland BV, Best, The Netherlands). Left ventricular endocardial contours were drawn manually on short-axis cine images at end-diastole and end-systole for functional parameters, including ejection fraction, end-diastolic volume, end-systolic volume, stroke volume, and cardiac output with indexing for body surface area except for ejection fraction.

Native and postcontrast MOLLI images were corrected for in-plane motion using automatic motion correction. The T1 times were calculated using nonlinear fitting with a maximum likelihood estimator (MLE) by the postprocessing software. In case of extensive artifacts in an imaging sequence, the patient was excluded from the respective analysis at the discretion of the analyzing physician. Regions of interest were drawn conservatively in the ventricular septum¹³ and the left- and right-ventricular blood pool on the mid-ventricular short axis and transposed to the corresponding postcontrast image (Figure S1). ECV was calculated from native and postcontrast T1 relaxation times and the hematocrit (Figure ​(Figure22).

Derivation and Validation

Eight specific regression models for each combination of sex, field strength, and blood pool measurement site (right ventricle [RV] and left ventricle [LV]) were derived. Two common models for LV/RV blood pool with sex and field strength as factors were derived. The models were then used to calculate the synthetic ECV and assess its agreement and correlation with conventionally measured ECV in the validation group. The published model of Fent et al¹⁴ was used for further validation. Subgroup validations were performed in patients with amyloidosis (n=29) and anemia (n=185) from the whole cohort. The diagnosis of amyloidosis was confirmed by pathology as described previously.¹⁵ The diagnosis of anemia was based on laboratory hematocrit: hematocrit <41% in men and <36% in women was defined as anemia.¹⁶

Histological Analysis

Histological analysis was performed from myocardial tissue of 17 patients. Analysis of endomyocardial biopsy samples was performed in a specialized laboratory by experienced pathologists.¹⁷ HE and Masson trichrome staining were used for histological examination of myocyte necrosis and interstitial fibrosis. Fibrotic and artifact areas were digitally marked on Masson trichrome-stained endomyocardial biopsy sections and photographed at a magnification of x200 with a Zeiss Axioskop 40 microscope. The area of fibrosis was quantified by using the program Quantuepatho as previously described ¹⁷ (Figure ​(Figure3A3A and ​and3B).3B). The histology validation was performed in the best synthetic model after comparison. All samples were analyzed blinded to the CMR findings and ECV values.

Statistical Analysis

Continuous variables were expressed as mean and SD, and categorical variables were expressed as percentages. Comparisons between means were performed using Student t test for continuous values with normal distributions and a Wilcoxon signed-rank test for non-normal data. In the derivation cohort, 8 specific models for different groups (male/female, 1.5T/3T, LV/RV blood pool) were derived using Model II (Deming) linear regression analysis, in which R1 was predictor and hematocrit was the response. Two common models for LV/RV blood pool were derived using multiple regression analysis, in which R1, sex, and field strength were predictors and hematocrit was the response. The published model of Fent et al¹⁴ was used for further validation. Synthetic ECVs were calculated using specific and common models. A Bland-Altman analysis was performed for agreement between measured and synthetic ECV in the validation cohort. Correlations between the T1 time, hematocrit, and ECV were assessed using the Pearson or Spearman correlation coefficient. In the subgroup analysis of the amyloidosis cohort, nonamyloidosis patients were matched with amyloidosis patients by sex and scanner. In the ECV misclassification analysis, a new synthetic ECV cutoff value was derived using a receiver operating characteristic curve¹⁸ based on conventionally measured ECV and McNemar test was used in the validation cohort. Statistical tests were 2-tailed, and a P<0.05 was considered statistically significant. Data were analyzed with SPSS (version 26, Statistical Package for the Social Sciences, International Business Machines, Inc, Armonk, NY) and GraphPad Prism software (version 9.0.0, GraphPad Software, Inc, La Jolla, CA).

Patients Characteristics

One thousand hundred one subjects (669 male, 60.8%) with a broad spectrum of referral diagnoses were included (Table S1). The average age was 51.8±17.1 years, and the mean conventionally measured ECV was 27.2±6.0%. Patients in both scanners (1.5 and 3T) were randomly split into derivation and validation groups. The patient characteristics are shown in Table ​Table11.

Native T1 Relaxation Time in Different Subgroups

The blood pool native T1 relaxation time at 3T was higher than at 1.5T (1836±101.5 versus 1552±109.6 ms, P<0.0001 in LV blood pool; 1784±120.2 versus 1544±116.0 ms, P<0.0001 in RV blood pool). Blood pool native T1 relaxation time was higher in female than in male patients (1878±92 versus 1811±98 ms, P<0.0001 in 3T; 1590±101 versus 1525±108 ms, P<0.0001 in 1.5T). The LV blood pool T1 relaxation time was higher than that of the RV (1836±101.5 versus 1784±120.2 ms at 3T, 1552±109.6 versus 1544±116.0 ms at 1.5T, P<0.0001 and P<0.0001, respectively; Figure ​Figure4;4; Table S2)

Derivation

In the derivation cohort, there were 326 subjects (202 male, 62.0%) examined at 3T and 224 (132 male, 58.9%) at 1.5T. The correlation coefficients R² of measured hematocrit and native blood R1 ranged from 0.22 to 0.46 (P all <0.0001) in the different subgroups (Figure ​(Figure5).5). Based on the linear relationship, 8 specific models were derived. To simplify the workflow, common models for LV/RV blood pool were derived (Table ​(Table22).

Validation

In the validation cohort, the correlation between synthetic hematocrit and conventionally measured hematocrit was moderate (R² range, 0.48–0.55), while the correlation between synthetic ECV and conventionally ECV was strong (R² range, 0.87–0.89; Table ​Table3;3; Tables S3 and S4). There was no statistical difference between measured and synthetic hematocrit/ECV (Table S5).

Using specific models, a slight bias and acceptable limits of agreement between measured and synthetic ECVs were observed, and the correlations between ECVs were strong (bias −0.16 and −0.10, R²=0.87 and R²=0.88 for LV and RV blood pool derived specific model, respectively; Figure ​Figure6A6A through ​through6D).6D). Using common models, Bland-Altman analysis also demonstrated slight bias, which was slightly higher than specific models, the correlation between measured and synthetic ECVs was strong (bias −0.21 and −0.18, R²=0.88 and R²=0.89 for LV and RV blood pool derived model, respectively; Figure ​Figure6E6E through ​through6H).6H). However, since the limits of agreement of ECV mostly were −4% to 4% for local models, the small bias may be attributable to the under and overestimation. Using published models,¹⁴ the results show a higher bias and lower correlation (Table ​(Table33).

Given the good performance of common model derived from the RV blood pool, which showed relatively low bias and high agreement, the RV common model was applied to the following subgroups for validation.

Histology Validation

In the cohort of 17 patients with histological analysis, the average collagen volume fraction (CVF) was 15.5±11.3%, and the average measured ECV was 31.9±11.2%. Synthetic ECVs correlated well with histology CVF, as conventional measured ECV (R²=0.63, P=0.0001, for synthetic ECV; R²=0.66, P<0.0001, for measured ECV). Synthetic ECV showed no statistical difference to measured ECV (31.6±10.1% versus 31.9±11.2%, P=0.89; Figure ​Figure3C3C and ​and33D).

Anemia Cohort Validation

The ECV bias in anemia cohort (n=185) was greater than in the nonanemia cohort (n=916) (−1.21 to −2.45 in the anemia cohort, −0.39 to 0.39 in the nonanemia cohort; Figure ​Figure7A).7A). Additional validation using specific models were performed, which showed the specific models could reduce the bias compared with common models (−1.2±2.2% versus −2.4±1.7%, LV specific and common models; −1.3±1.8% versus 2.5±1.5%, RV specific and common models).

Amyloidosis Cohort Validation

Further validation was performed in 29 patients with confirmed amyloidosis (Table S6), using 29 matched nonamyloidosis patients randomly drawn from the cohort as the control group. There was no statistical difference between the measured and synthetic ECV in the amyloidosis cohort, as in the control group (51.1±10.3% versus 51.9±10.9% and 27.6±4.7% versus 26.8±4.3%, P=0.18 and P=0.12, amyloidosis and control group, respectively), the bias of ECVs was only 0.77% in the amyloidosis group (Figure ​(Figure7B;7B; Figure S2).

Validation in Different LVEF Cohorts

Patients with ejection fraction >50% (n=779) had lower ECV values than those with ejection fraction ≤50% (n=322; 26.2±4.6% versus 29.5±7.9%). There was no statistical difference between the measured ECV and synthetic ECV in both groups (26.2±4.6% versus 26.1±4.3, P=0.84 in LVEF>50%, 29.5±7.9% versus 29.5±7.9%, P=0.25 in LVEF≤50%), and the bias were small (bias 0.11 and 0.03 in LVEF>50% and ≤50%; Figure S3).

Misclassification Analysis

Based on the conventional ECV cutoff value of 30%, new cutoff value of 29.5% for synthetic ECV was derived in the derivation cohort, misclassification analysis in the validation cohort showed 6% total misclassifications (4% false-negative and 2% false-positive, P=0.12; Tables S7 and S8).

Differences in Native T1 Relaxation Times

In our study, the native blood pool T1 times were 65 to 93 ms higher in females than males, and 184 to 204 ms higher in the 3T scanner than in the 1.5T. The native blood pool T1 relaxation time was higher in the LV than the RV. The reason may be that deoxyhemoglobin is more abundant in venous blood than in arterial blood, which has a paramagnetic effect that lowers the T1 relaxation time. The difference between T1 times of venous blood and arterial blood was also observed before.¹⁹ A potential pitfall arises in patients with unusually low venous or arterial oxygen saturation, which might lower T1 relaxation times irrespective of the hematocrit and lead to false high synthetic hematocrit estimates.

Derivation and Validation

Based on the differences of native blood pool T1 values by sex, field strength, and measurement site, 8 specific models and 2 common models were derived. Both the common and specific models performed well in validation analysis. In terms of ECV bias, the specific model had a slight advantage over the common model.

As in previous studies,^{11,12,14,20,21} synthetic hematocrit and measured hematocrit were only moderately correlated. Nevertheless, synthetic ECV was highly correlated with measured ECV with only minor bias. In correlation to histological measurements, synthetic ECV performed equally to measured ECV.

Validation in the amyloidosis and anemia cohorts further demonstrated the clinical value of synthetic ECV. The ECV error did not increase significantly with the absolute ECV value in the amyloidosis cohort, suggesting that the synthetic ECV remained highly accurate. The synthetic ECV error was higher in the anemic cohort than in the nonanemic cohort, and the synthetic ECV tended to underestimate the measured ECV. This was consistent with the finding of Su et al.¹⁶ ECV errors were relatively more significant when measured hematocrit was lower. One possible reason is that the measured T1 relaxation time is dependent on more factors than the hematocrit alone, as evidenced by the R² below 1. However, the specific model minimized the ECV error in the anemic cohort. The myocardial ECV value in healthy control was 25.3±3.5%²² and 25.4±2.5%,²³ which was considered an acceptable ECV error of 2.5% to 3.5%. In our study, the ECV errors using all local models, even in the anemia cohort, were within the acceptable ECV error range. However, given the limit of agreement of synthetic ECVs were mostly −4 to 4, we have to acknowledge that small mean differences may be attributable to both under and over-estimation.

The conventional ECV was estimated based on venous blood hematocrit, whereas the ideal hematocrit for ECV calculation should be obtained from the LV blood pool representing arterial blood. As observed, there was a non-negligible difference between T1 values of venous and arterial blood. Therefore, we derived synthetic ECV models depending on RV and LV blood pool separately.

Notably, the validation results of RV models were slightly better than the results of the LV models in our cohort, either with conventional ECV as a reference or with histological results. Possible reasons to consider were that while there is some linear relationship between blood pool T1 time and hematocrit, the hematocrit values used in our conventional measurement of ECV are peripheral venous blood hematocrit, venous T1 time (RV blood pool) seems to be to reflect venous hematocrit better. Second, there was some error between measured hematocrit and RV-synthetic hematocrit, similarly, measured hematocrit and actual LV-synthetic hematocrit. There may be some degree of offset between the differences, instead of bringing the values of RV-synthetic hematocrit and actual hematocrit of the LV blood pool closer together, leading to a higher degree of agreement.

Shang et al²⁴ showed that synthetic ECV might result in 6% to 25% misclassification. In our study, 7% misclassification was observed. CMR-ECV as a new biological parameter for quantifying diffuse myocardial fibrosis does not have a universal standard cut-off definition to date. Besides, misclassification analysis was not based on the gold standard “tissue biopsy” but on conventional ECV. Therefore, the small proportion of miscategorization cannot offset the excellent overall performance of synthetic ECV.

Clinical Implications

CMR-ECV mapping is the only noninvasive technique proven to provide a stable indicator of diffuse myocardial fibrosis. It has significant value in the clinical management and prognostic assessment of multiple diseases.^25–28 However, the promotion of conventional ECV is limited by the need for blood tests at the time of scanning and the high variability of hematocrit. Synthetic ECV obtained using T1 time of the native blood pool omits the process of blood testing, while the results of synthetic hematocrit are in real time, thus avoiding apparent fluctuations in hematocrit due to time, body position, or other reasons. Excitingly, the results of synthetic ECV are almost identical to those of conventional ECV compared with conventional methods. It increases the potential of ECV in routine clinical CMR and provides a means to generate real time ECV mapping during the scan automatically; an online tool on a CMR scanner to create an instant fully automated ECV map could be implemented at 1.5 and 3T on different CMR vendor platforms. Synthetic ECV is an attractive and promising tool that could greatly facilitate the broader use of CMR-ECV.

Limitations

There are several limitations in our study. Only referral diagnoses were collected, the complete medical history was not analyzed due to time constraints; We used a single T1 sequence (MOLLI); however, different sequences of T1 mapping have been reported to yield different ECV values. Different T1 values for the RV and LV blood pool were found, which may arise from the different oxygen content. However, the exact reason needs further investigation. Also, we found that the values derived from the RV blood pool had lower bias and higher agreement than those derived from the LV blood pool, which requires more explanation. Multicenter studies with different vendors should be further performed to validate these results.

Conclusions

Synthetic ECV provides a reliable tool for clinical promotion of ECV, which avoids unnecessary blood draws and allows real-time assessment of hematocrit and the generation of online automated ECV mapping. Specific models could provide the most accurate value, while common models could be more suitable in routine clinical practice due to their simplicity while maintaining adequate accuracy.

Sources of Funding

Dr Kelle was supported by a grant from Philips Healthcare and received funding by the German Ministry of Education and Research. Drs Kelle, Doeblin, and Pieske were funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)—SFB-1470—B06.

Disclosures

None.

Supplemental Material

Tables S1–S8

Figures S1–S3

Supplemental Material is available at https://www.ahajournals.org/doi/suppl/10.1161/CIRCIMAGING.121.013745.

For Sources of Funding and Disclosures, see page 275.

1. Taylor AJ, Salerno M, Dharmakumar R, Jerosch-Herold M. T1 mapping: basic techniques and clinical applications.JACC Cardiovasc Imaging. 2016;9:67–81. doi: 10.1016/j.jcmg.2015.11.005 [PubMed] [Google Scholar]

2. Brilla CG, Funck RC, Rupp H. Lisinopril-mediated regression of myocardial fibrosis in patients with hypertensive heart disease.Circulation. 2000;102:1388–1393. doi: 10.1161/01.cir.102.12.1388 [PubMed] [Google Scholar]

3. Díez J, Querejeta R, López B, González A, Larman M, Martínez Ubago JL. Losartan-dependent regression of myocardial fibrosis is associated with reduction of left ventricular chamber stiffness in hypertensive patients.Circulation. 2002;105:2512–2517. doi: 10.1161/01.cir.0000017264.66561.3d [PubMed] [Google Scholar]

4. Lewis GA, Dodd S, Clayton D, Bedson E, Eccleson H, Schelbert EB, Naish JH, Jimenez BD, Williams SG, Cunnington C, et al.. Pirfenidone in heart failure with preserved ejection fraction: a randomized phase 2 trial.Nat Med. 2021;27:1477–1482. doi: 10.1038/s41591-021-01452-0 [PubMed] [Google Scholar]

5. Kwak S, Everett RJ, Treibel TA, Yang S, Hwang D, Ko T, Williams MC, Bing R, Singh T, Joshi S, et al.. Markers of myocardial damage predict mortality in patients with aortic stenosis.J Am Coll Cardiol. 2021;78:545–558. doi: 10.1016/j.jacc.2021.05.047 [PubMed] [Google Scholar]

6. Fröjdh F, Fridman Y, Bering P, Sayeed A, Maanja M, Niklasson L, Olausson E, Pi H, Azeem A, Wong TC, et al.. Extracellular volume and global longitudinal strain both associate with outcomes but correlate minimally.JACC Cardiovasc Imaging. 2020;13:2343–2354. doi: 10.1016/j.jcmg.2020.04.026 [PubMed] [Google Scholar]

7. Yang EY, Ghosn MG, Khan MA, Gramze NL, Brunner G, Nabi F, Nambi V, Nagueh SF, Nguyen DT, Graviss EA, et al.. Myocardial extracellular volume fraction adds prognostic information beyond myocardial replacement fibrosis.Circ Cardiovasc Imaging. 2019;12:e009535. doi: 10.1161/CIRCIMAGING.119.009535 [PMC free article] [PubMed] [Google Scholar]

8. Haaf P, Garg P, Messroghli DR, Broadbent DA, Greenwood JP, Plein S. Cardiac T1 Mapping and Extracellular Volume (ECV) in clinical practice: a comprehensive review.J Cardiovasc Magn Reson. 2016;18:89. doi: 10.1186/s12968-016-0308-4 [PMC free article] [PubMed] [Google Scholar]

9. Robinson AA, Chow K, Salerno M. Myocardial T1 and ECV measurement: underlying concepts and technical considerations.JACC Cardiovasc Imaging. 2019;12(11 Pt 2):2332–2344. doi: 10.1016/j.jcmg.2019.06.031 [PMC free article] [PubMed] [Google Scholar]

10. Messroghli DR, Moon JC, Ferreira VM, Grosse-Wortmann L, He T, Kellman P, Mascherbauer J, Nezafat R, Salerno M, Schelbert EB, et al.. Clinical recommendations for cardiovascular magnetic resonance mapping of T1, T2, T2* and extracellular volume: A consensus statement by the Society for Cardiovascular Magnetic Resonance (SCMR) endorsed by the European Association for Cardiovascular Imaging (EACVI).J Cardiovasc Magn Reson. 2017;19:75. doi: 10.1186/s12968-017-0389-8 [PMC free article] [PubMed] [Google Scholar]

11. Treibel TA, Fontana M, Maestrini V, Castelletti S, Rosmini S, Simpson J, Nasis A, Bhuva AN, Bulluck H, Abdel-Gadir A, et al.. Automatic measurement of the myocardial interstitium: synthetic extracellular volume quantification without hematocrit sampling.JACC Cardiovasc Imaging. 2016;9:54–63. doi: 10.1016/j.jcmg.2015.11.008 [PubMed] [Google Scholar]

12. Raucci FJ, Jr, Parra DA, Christensen JT, Hernandez LE, Markham LW, Xu M, Slaughter JC, Soslow JH. Synthetic hematocrit derived from the longitudinal relaxation of blood can lead to clinically significant errors in measurement of extracellular volume fraction in pediatric and young adult patients.J Cardiovasc Magn Reson. 2017;19:58. doi: 10.1186/s12968-017-0377-z [PMC free article] [PubMed] [Google Scholar]

13. Schulz-Menger J, Bluemke DA, Bremerich J, Flamm SD, Fogel MA, Friedrich MG, Kim RJ, von Knobelsdorff-Brenkenhoff F, Kramer CM, Pennell DJ, et al.. Standardized image interpretation and post-processing in cardiovascular magnetic resonance - 2020 update: Society for Cardiovascular Magnetic Resonance (SCMR): Board of Trustees Task Force on Standardized Post-Processing.J Cardiovasc Magn Reson. 2020;22:19. doi: 10.1186/s12968-020-00610-6 [PMC free article] [PubMed] [Google Scholar]

14. Fent GJ, Garg P, Foley JRJ, Swoboda PP, Dobson LE, Erhayiem B, Treibel TA, Moon JC, Greenwood JP, Plein S. synthetic myocardial extracellular volume fraction.JACC Cardiovasc Imaging. 2017;10:1402–1404. doi: 10.1016/j.jcmg.2016.12.007 [PubMed] [Google Scholar]

15. Chamling B, Drakos S, Bietenbeck M, Klingel K, Meier C, Yilmaz A. Diagnosis of cardiac involvement in amyloid a amyloidosis by cardiovascular magnetic resonance imaging.Front Cardiovasc Med. 2021;8:757642. doi: 10.3389/fcvm.2021.757642 [PMC free article] [PubMed] [Google Scholar]

16. Su MY, Huang YS, Niisato E, Chow K, Juang JJ, Wu CK, Yu HY, Lin LY, Yang SC, Chang YC. Is a timely assessment of the hematocrit necessary for cardiovascular magnetic resonance-derived extracellular volume measurements?J Cardiovasc Magn Reson. 2020;22:77. doi: 10.1186/s12968-020-00689-x [PMC free article] [PubMed] [Google Scholar]

17. Klingel K, Sauter M, Bock CT, Szalay G, Schnorr JJ, Kandolf R. Molecular pathology of inflammatory cardiomyopathy.Med Microbiol Immunol. 2004;193:101–107. doi: 10.1007/s00430-003-0190-1 [PubMed] [Google Scholar]

18. Mandrekar JN. Receiver operating characteristic curve in diagnostic test assessment.J Thorac Oncol. 2010;5:1315–1316. doi: 10.1097/JTO.0b013e3181ec173d [PubMed] [Google Scholar]

19. Shimada K, Nagasaka T, Shidahara M, Machida Y, Tamura H. In vivo measurement of longitudinal relaxation time of human blood by inversion-recovery fast gradient-echo MR imaging at 3T.Magn Reson Med Sci. 2012;11:265–271. doi: 10.2463/mrms.11.265 [PubMed] [Google Scholar]

20. Lim EH, Le TT, Bryant J, Chung YC, Su B, Gan J, Hausenloy DJ, Cook SA, Chin CWL. Importance of Sex-Specific Regression Models to Estimate Synthetic Hematocrit and Extracellular Volume Fraction.JACC Cardiovasc Imaging. 2018;11:1366–1367. doi: 10.1016/j.jcmg.2017.11.035 [PubMed] [Google Scholar]

21. Kammerlander AA, Duca F, Binder C, Aschauer S, Zotter-Tufaro C, Koschutnik M, Marzluf BA, Bonderman D, Mascherbauer J. Extracellular volume quantification by cardiac magnetic resonance imaging without hematocrit sampling: Ready for prime time?Wien Klin Wochenschr. 2018;130:190–196. doi: 10.1007/s00508-017-1267-y [PMC free article] [PubMed] [Google Scholar]

22. Sado DM, Flett AS, Banypersad SM, White SK, Maestrini V, Quarta G, Lachmann RH, Murphy E, Mehta A, Hughes DA, et al.. Cardiovascular magnetic resonance measurement of myocardial extracellular volume in health and disease.Heart. 2012;98:1436–1441. doi: 10.1136/heartjnl-2012-302346 [PubMed] [Google Scholar]

23. Kellman P, Wilson JR, Xue H, Bandettini WP, Shanbhag SM, Druey KM, Ugander M, Arai AE. Extracellular volume fraction mapping in the myocardium, part 2: initial clinical experience.J Cardiovasc Magn Reson. 2012;14:64. doi: 10.1186/1532-429X-14-64 [PMC free article] [PubMed] [Google Scholar]

24. Shang Y, Zhang X, Zhou X, Wang J. Extracellular volume fraction measurements derived from the longitudinal relaxation of blood-based synthetic hematocrit may lead to clinical errors in 3 T cardiovascular magnetic resonance.J Cardiovasc Magn Reson. 2018;20:56. doi: 10.1186/s12968-018-0475-6 [PMC free article] [PubMed] [Google Scholar]

25. Cui Y, Chen Y, Cao Y, Liu J, Song J, Zhang S, Kong X, Han P, Shi H. Myocardial extracellular volume fraction measurements with MOLLI 5(3)3 by cardiovascular MRI for the discrimination of healthy volunteers from dilated and hypertrophic cardiomyopathy patients.Clin Radiol. 2019;74:732 e9–732 e16. [PubMed] [Google Scholar]

26. Wong TC, Piehler K, Meier CG, Testa SM, Klock AM, Aneizi AA, Shakesprere J, Kellman P, Shroff SG, Schwartzman DS, et al.. Association between extracellular matrix expansion quantified by cardiovascular magnetic resonance and short-term mortality.Circulation. 2012;126:1206–1216. doi: 10.1161/CIRCULATIONAHA.111.089409 [PMC free article] [PubMed] [Google Scholar]

27. Wong TC, Piehler KM, Kang IA, Kadakkal A, Kellman P, Schwartzman DS, Mulukutla SR, Simon MA, Shroff SG, Kuller LH, et al.. Myocardial extracellular volume fraction quantified by cardiovascular magnetic resonance is increased in diabetes and associated with mortality and incident heart failure admission.Eur Heart J. 2014;35:657–664. doi: 10.1093/eurheartj/eht193 [PMC free article] [PubMed] [Google Scholar]

28. Khan MA, Yang EY, Nguyen DT, Nabi F, Hinojosa J, Jabel M, Nagueh SF, Graviss EA, Shah DJ. Examining the relationship and prognostic implication of diabetic status and extracellular matrix expansion by cardiac magnetic resonance.Circ Cardiovasc Imaging. 2020;13:e011000. doi: 10.1161/CIRCIMAGING.120.011000 [PMC free article] [PubMed] [Google Scholar]