Patients
CMR was performed in 40 healthy volunteers (12 female/28 male, mean age 37 ± 19 y, range 7–78 y), some of them were part of a former study. No participant had history of cardiac disease and no systolic trans-mitral blood flow as sign of mitral valve regurgitation was observed in 4-chamber view cine images. Four individuals underwent 2–4 repeated CMR scans (between 40 and 156 months).
The institutional committee on human research approved the study, and all volunteers gave their written informed consent prior to examination.
CMR protocol
CMR was performed on a 1.5 T MR system (Symphony, Siemens Healthcare, Erlangen, Germany) using a four-element phased array coil. Single slice acquisitions were performed with breath hold and retrospectively ECG-gated cine SSFP sequences. Phase encoding steps (segments per view) were reduced for higher heart rates. Specifically, cardiac function was assessed from short axis cine-series (TE = 1.6 ms, TR = 50 ms, FA = 65°, bandwidth = 965 Hz/pixel, in-plane resolution = 1.5 × 1.5 mm2, slice thickness = 6, gap = 0/6 mm, cardiac phases = 25).
LV volumetry and diastolic function
LV volumetry was assessed by two radiologists with more than 10 years of CMR experience. LV volumetry was performed with: (1) inclusion of TPM into the myocardium and (2) inclusion of TPM into the LV blood pool.
LV end-diastolic volumes (EDV), end-systolic volumes (ESV), EF and myocardial mass (M) were determined by manual delineation of endocardial and epicardial borders in end-systolic, end-diastolic and mid-diastolic short axis views using dedicated software (CMRtools®, v. 2010, Cardiovascular Imaging Solutions Ltd, Cambs, UK) (Fig. 1). LV blood and myocardial volumes were calculated for each of the acquired 25 phases of the cardiac cycle by the following steps out of CMRtools’ workflow, each step characterized by a set of four images (Fig. 1I):
-
The longitudinal main axis of the LV chamber was set by defining the central symmetry axis of the LV chamber in three representative short axis slices (i.e. basal, mid-papillary and apical) in diastole and systole, respectively (Fig. 1A, B).
-
Delineation of LV endocardial borders was performed in a mid-papillary short axis slice in diastole and systole, respectively (Fig. 1C). The software automatically propagates endocardial contours for the other slices and phases of the short axis stack, respectively.
-
The resulting LV blood volume (including TPM) can be visualized in red color by activating the “shading” function, and its correct contour can be controlled as indicated by green circles in the long axis view of Fig. 1D.
-
Exclusion of TPM from the blood pool is automatically provided by the software based on different signal intensities of myocardium/TPM and blood (Fig. 1E). The individual threshold for precise identification of TPM can manually be adjusted on a color contrast scale (see ruler bar at the bottom of Fig. 1I) and was set in a mid-papillary diastolic short axis slice. The scale has to be moved until TPM are excluded from the LV blood pool. This step of TPM exclusion with threshold adjustment takes only several seconds.
-
The mitral valve plane was manually defined in diastolic and systolic 4- and 2-chamber views (see red arrow in Fig. 1F, left). This results in an oblique plane in the three-dimensional LV model (red rectangle in Fig. 1F, right) for precise separation of blood volumes of the left ventricle and atrium.
-
A third cardiac phase (mid-diastolic) was selected in a mid-papillary short axis slice, where the propagated endocardial contour showed the largest deviation from the real LV contour (Fig. 1G). Endocardial borders were manually corrected for this cardiac phase, resulting in a significantly different time volume curve (Fig. 1H, yellow symbols) in comparison with the curve without consideration of a third phase (Fig. 1H, white circles).
-
Finally, a fine-tuning procedure of the LV endocardial contour was performed in all three (diastole, systole, mid-diastole) phases by manually adjusting the endocardial border points (Fig. 1I, upper left) where necessary. This procedure results in a three-dimensional wire model showing all short axis slices below the basal plane (Fig. 1I, bottom right).
From the time-volume curve observables (R-R time intervals (phases), temporal TPM masses and blood volumes), the LV diastolic function parameters PFRR, DVR, etc., can be semi-automatically assessed in an EXCEL template. The only manual adjustment has to be made for the minimum between the early and atrial peak (EAmin) (Fig. 2B).
The early and atrial (late) filling volumes (EFV and AFV) refer to the early diastolic LV filling due to passive LV relaxation, and the late LV filling due to active contraction of the left atrium (Fig. 2A). However, there is more information in this time pattern of ventricular volumes, which is revealed after differentiation. The temporally differentiated LV time-volume curve usually results in three peaks characterized by the systolic peak contraction rate (PCR) as well as the diastolic early (EPFR) and atrial (APFR) peak filling rate (Fig. 2B, C). In order to avoid artificial differentiation peaks, the data were smoothened by their next neighbors. As result, these two separated diastolic peaks can be fitted by a Gaussian dual-peak fit resulting in the early and atrial filling volumes (EFV, AFV) (Fig. 2B). The calculated filling volume ratio (FVR = EFV / AFV) is the corresponding volumetric (= area) counterpart to the phase-specific (= peak) peak filling rate ratio (PFRR = EPFR/APFR).
The EPFR and APFR assessed by CMR reflect the early (E) and atrial (A) transmitral peak filling velocities determined by echocardiography [13]. In analogy to echocardiography (E/A ratio), the peak filling rate ratio (PFRR = EPFR/APFR) is the equivalent to characterize diastolic filling patterns using CMR. Here, we have to point out that assessment of E/A ratios (PFRR, FVR) by CMR relies on filling volumes, while echocardiographic assessments address velocity measurements across the mitral valve [15].
The time interval between the early atrial phase peak and the atrial phase peak normalized to the RR-cycle (heart rate) is characterized by EAPT%, while EAmin describes the distinctiveness of the atrial peak relative to the minimum rate between EPFR and APFR. The corresponding temporal volume is LVpreA.
Further echocardiographic-derived diastolic indices that were determined included the time to peak filling rate (TPFR) and deceleration time (DT), describing the rise and decline of the early diastolic peak. The diastolic volume recovery (DVR) is defined according to Kawaji et al. [13], i.e., as proportion of diastole required to recover 80% of systolic volume, either absolute (DVR in ms) or as percentage of diastole (DVR%) (Fig. 2A).
Except for the deceleration time (DT%), all diastolic parameters were assessed from both volumetric approaches with either inclusion of TPM into the LV myocardium or the LV blood pool.
The influence of temporal resolution on diastolic measures was studied at temporal resolutions of 12 ms (= 64 phases), 16 ms (= 50 phases) and 30 ms (= 25 phases) in one healthy control (26 y).
Statistical and data analysis
Since many of the respective parameters had a skewed distribution (skewness > 1.0 ± 0.2), we applied nonparametric statistics reaching statistical significance by two-sided p < 0.05: median, 95% range, paired Wilcoxon test and Spearman rank correlation (rS) test (STATISTICA, Stat-Soft Inc., Tulsa, USA).
For adjustment (prediction) of diastolic LV parameters as a function of age, we fitted exponential models to the data using the Marquardt–Levenberg fit algorithm of Slide Write Plus for Windows (Advanced Graphics Software Inc., Encinitas, CA, USA), which resulted in adjusted coefficients of determination (r2), curve parameters ± standard errors. Similarly, the diastolic filling rates were fitted by a Gaussian twin peak model of type a0·exp(− 0.5·((x − a2)/a1)2) with a0 = amplitude (cm3/s), a1 = peak width/2 (ms), and a2 = peak position (ms) (fitted by Slide Write). The early or atrial filling volumes can then be calculated from the fitted parameters by a0·a1/1000·√(2π).
A semi-automatic EXCEL analysis template was developed for calculating critical time points within the differentiated time-volume curve (systole, early and atrial peaks) and derived non-trivial diastolic parameters (LVpreA, DVR, EAPT, EAmin, TPFR, DT, PFRR, FVR). This template can be obtained from the authors.