Phantom-based correction for standardization of myocardial native T1 and extracellular volume fraction in healthy subjects at 3-Tesla cardiac magnetic resonance imaging

This prospective study was approved by the institutional review boards of the participating hospitals. Subjects were enrolled after obtaining written informed consent to participate in this study. From December 2019 to April 2021, healthy asymptomatic adult ( ≥ 20 years) volunteers of five different age groups (20–29 years, 30–39 years, 40–49 years, 50–59 years, and 60–79 years) were prospectively enrolled in three academic hospitals (Fig. 1). The target number of subjects was 14 (7 males and 7 females) per age group, which was determined by referencing the number of healthy subjects in similar researches (9–11 subjects per age group) [14, 15]. Prior to inclusion and CMR, all subjects underwent a clinical examination for symptoms of cardiovascular disease and assessment of medical history and cardiovascular risk factors, such as smoking, hypertension (systolic and diastolic blood pressure > 90mmHg/90 mmHg with home - based remedies or drug treatment), hyperlipidemia, atrial fibrillation, diabetes mellitus, obesity (BMI > 30 kg/m²), and family history of cardiovascular disease (acute coronary syndrome or coronary revascularization in first - degree relatives < 65 years old). For screening, all subjects underwent ECG and blood sampling with measurement of estimated glomerular filtration rate (eGFR, Modification of Diet in Renal Disease formula), hematocrit, cholesterol, and N-terminal pro - brain natriuretic peptide prior to inclusion. Exclusion criteria were (1) any evidence of heart disease as indicated by clinical history or physical examination, (2) presence of abnormal ECG or hyperlipidemia (total cholesterol > 240 mg/dL) on screening, (3) pregnancy, (4) contraindications to CMR or expectation of having degraded CMR image quality (ferrometallic cerebral aneurysm clips, pacemaker or implantable defibrillator, or severe claustrophobia), or contraindications to injection of gadolinium-based contrast agent (renal insufficiency with eGFR < 45 mL/min/1.73m²), and (5) diabetes. After 96 healthy Korean subjects were screened, 20 subjects were excluded due to screening failure (15 with hyperlipidemia, 2 with abnormal ECG, 2 with hypertension, and 1 who did not meet the age criteria), and 5 subjects withdrew their informed consents after study participation. The final population consisted of 71 participants (34 males, mean age 45.5 ± 15.5 years; 29 subjects from institution A, 16 subjects from institution B, and 26 subjects from institution C), with 14 subjects (7 males and 7 females) per age group between 20–29 and 50–59 years and 15 subjects (6 males and 9 females) in the age group of 60–79 years.

At all three participating institutions, CMR was performed using a 3-Tesla (T) system (Siemens 3T Prisma^fit for Institution A, Siemens 3T Verio for Institution B, and Siemens 3T SKYRA for Institution C). The CMR acquisition parameters for cine imaging and T1 mapping are described in Supplementary Material. To assess left ventricular (LV) myocardial function and mass, short-axis images of the LV were acquired using a cine balanced steady-state free precession (bSSFP) sequence [16]. Three short-axis Modified Look-Locker Inversion-recovery (MOLLI) images at the base, mid-cavity, and apex were acquired for native T1 mapping [16, 17]. Then, a total dose of 0.1 mmol/kg gadolinium agent (Uniray, gadoterate meglumine, Dongkook Pharmaceutical Co., Ltd.) was injected. Ten minutes after contrast injection, post-contrast MOLLI T1 mapping was acquired for T1 determination in an identical location as for native T1 mapping.

We developed a novel method using the T1MES phantom. The process of image acquisition and T1 correction based on phantom are shown in Fig. 2. The T1MES phantom was scanned at each institution within a month in December 2019, when the clinical study started. The phantom was employed to measure the error in T1 measurement in three institutions with the MOLLI T1 mapping protocol based on the default MOLLI protocol provided by the CMR manufacturer (Siemens Healthineers) with adjustment of field of view and without other voluntary modification of scan parameters.

To measure the gold-standard T1 (T1_GS) of the T1MES phantom in each institution, the inversion-recovery prepared turbo spin-echo [9] and the MOLLI sequence were used (Supplementary Table 1). Other scan considerations, such as setting the position in the iso-center, shim volume, and simulation ECG, were according to the instruction of the T1MES manual [18]. The ground-truth T1 (T1_GT) of this T1MES phantom was regarded as the T1 described in the manual by the manufacturer.

To reduce the variation in the T1, three correction methods were considered (Supplementary Table 2). First, a gold-standard T1-based correction function (GC) was calculated by multiple polynomial regression with T1_GS and T1_GT of the T1MES phantom. Second, a MOLLI T1-based correction function (MC) was calculated with the T1 on MOLLI (T1_ML) and T1_GT. Last, an internal reference T1-based correction function (IC) was calculated with the T1_ML and the T1_GS. In particular, the MC and IC methods were subdivided according to whether the RR interval (RRI) of the subject was considered. When the RRI of each subject was considered (adaptive RRI), the correction function was calculated using a T1_ML with an RRI close to the RRI of the subject (Supplementary Table 3). Otherwise, a T1_ML with an RRI of 900 ms (static RRI) was used (Supplementary Table 4).

First-, second-, and third-order correction equations were applied to each correction method as follows: where T1u is the uncorrected T1, T1c is the corrected T1, and a, b, c, and d are coefficients of the correction function according to each correction method (Supplementary Table 3 and 4). We chose the equation that showed the lowest coefficient of variation (CoV, standard deviation/mean) after correction.

CMR images were anonymized and analyzed independently by two experienced observers (Y.J.S. and B.W.C., cardiac radiologists with 8 and 21 years of CMR experience, respectively) who were blinded to the clinical data. Cine imaging and T1 map images were analyzed using commercial software (cvi42 image analysis software, Circle Cardiovascular Imaging Inc.) (Supplementary Methods).

On cine bSSFP images, the endocardial and epicardial contours of the LV were semi-automatically drawn with manual adjustments when needed. LV end-diastolic volume (EDV) and end-systolic volume (ESV) were calculated using the modified Simpson method and were indexed to body surface area. LV ejection fraction was calculated as (EDV-ESV)/EDV.

Native and post-contrast T1 map images were generated by fitting pixels to the equation s(t) = a – b exp. (t/T1*), and T1 = T1*((b/a−1), where a and b are constants, t is time, and s(t) is the signal intensity at time t. On T1 map images, endocardial and epicardial contouring of the LV was performed semi-automatically, and manual adjustments were applied when needed. Native and post-contrast blood T1 times were measured on a region of interest drawn in the center of the blood pool. Native and post-contrast T1 values were measured in 16 AHA segments in the short-axis view of LV [19]. The myocardial ECV was calculated using the following equation [20]:

Native T1 and ECV fraction were analyzed as the mean of the 16 entire segments (global) and mid septum, after the exclusion of segments with image artifacts (e.g., off-resonance or partial volume artifact) that caused significant deterioration in T1 measurement.

Statistical analyses were performed using MedCalc for Windows version 19.1.0.0 (MedCalc Software) and R (version 4.0.2, R Foundation). Continuous variables are expressed as the mean ± standard deviation or median with 25^th to 75^th percentile, and categorical variables are shown as counts and percentages. Categorical variables were compared using the χ² or Fischer’s exact test. Continuous variables were compared among groups using ANOVA for normally distributed data and the Kruskal-Wallis test for non-normally distributed data. T1 mapping results were excluded from the analysis if measured T1 values were estimated as outliers by the Tukey method [21, 22]. Variations in native T1 and ECV fraction among institutions were compared before and after phantom-based correction using a CoV. Inter-observer reproducibility of T1 times and ECV was assessed using an intraclass correlation coefficient. A probability value less than 0.05 was considered statistically significant, and a Bonferroni-adjusted post hoc probability value of less than 0.02 was considered statistically significant for comparison between the three institutions.

The baseline characteristics of the 71 subjects are shown in Table 1. No subject had a history of stroke, chronic obstructive pulmonary disease, or sleep apnea syndrome or presented hemochromatosis or anemia on the blood test.

The CMR examination was successfully completed in all subjects, but 2 patients (a 54-year-old male in institution A and a 29-year-old female in institution B) were excluded from the analysis of T1 mapping because their measured T1 values were estimated as outliers due to image artifacts. All participants had normal LV function on cine imaging (Supplementary Table 5). Artifacts precluded analysis of native T1 in 67 segments of 25 subjects and 83 segments of 35 subjects in post-contrast T1 maps, resulting in missing ECV in 111 segments of 41 subjects. Inter-observer agreement of measurements on T1 map was excellent for both native T1 and ECV (intraclass correlation coefficient 0.993 [95% confidence interval 0.989–0.996] and 0.998 [95% confidence interval 0.997–0.999]), respectively).

The MC₂ method (second-order correction equation with MC method) with static RRI was used for the correction of native and post-contrast T1 map images because the method showed the greatest decrease in the CoV of native T1 map in healthy human subjects among various correction methods (Fig. 3). Correction equations for myocardial native T1 and post-contrast T1 map for each institution are provided in Table 2.

The mean global native T1 and ECV fraction in 69 subjects are shown in Table 3. Mean RRI during the acquisition of native T1 and post T1 mapping sequences was 918.8 ± 144.1 ms (range 659.0–1261.3 ms) and 922.3 ± 129.7 ms (range 714.3–1247.5 ms), respectively. Before the phantom-based correction, the global native T1 was significantly different between the three institutions (1198.7 ± 32.1 ms for institution A, 1217.7 ± 39.9 ms for institution B, and 1232.7 ± 31.1 ms for institution C; p = 0.002), but ECV was not significantly different between the institutions (26.6 ± 1.8% for institution A, 27.5 ± 3.6% for institution B, and 27.4 ± 2.5% for institution C; p = 0.355). After phantom-based correction, the global native T1 was not significantly different between the three institutions (1284.5 ± 31.5 ms for institution A, 1296.5 ± 39.1 ms for institution B, and 1291.3 ± 29.3 ms for institution C; p = 0.440), and ECV was also not significantly different between the institutions (24.4 ± 2.2% for institution A, 25.9 ± 3.7% for institution B, and 25.4 ± 2.6% for institution C; p = 0.078). The mean of the corrected native T1 and ECV in 69 subjects were 1289.4 ± 32.4 ms and 25.0 ± 2.7%, respectively. After phantom-based correction, the CoV for the native T1 between the three institutions decreased from 3.0 to 2.5%.

In the measurement of the mid septum, native T1 was significantly different between the three institutions (1210.7 ± 35.0 ms for institution A, 1227.1 ± 34.8 ms for institution B, and 1244.2 ± 31.5 ms for institution C; p = 0.004) before phantom-based correction, but ECV was not significantly different between the institutions (26.7 ± 1.9% for institution A, 27.4 ± 3.1% for institution B, and 27.8 ± 2.5% for institution C; p = 0.079). After phantom-based correction, the mean native T1 of the mid septum was not significantly different between the three institutions (1296.2 ± 34.2 ms for institution A, 1305.6 ± 33.9 ms for institution B, and 1302.1 ± 29.7 ms for institution C; p = 0.498), and ECV was also not significantly different between the institutions (24.2 ± 1.9% for institution A, 25.6 ± 3.1% for institution B, and 25.8 ± 2.7% for institution C; p = 0.056). After phantom-based correction, the CoV for the native T1 between the three institutions decreased from 3.0 to 2.6%.