Improved quantification of small hearts for gated myocardial perfusion imaging

In this study, a small heart was defined as a heart with an ESV of ≤20 mL as calculated using Quantitative Gated SPECT (QGS) software (Cedars Sinai Medical Center, Los Angeles, CA) [1, 14]. Patients with a small heart were further divided into two groups (SH20 group with ESV 11–20 mL and SH10 group with ESV ≤10 mL), and were compared with the group of patients with a normal-sized heart, i.e. LV with ESV >20 mL as calculated using QGS (NH group).

The SIMIND Monte Carlo program together with the XCAT phantom were used to generate projection data [15, 16]. A phantom representing a standard Japanese woman was assumed. Four different LV sizes with EDVs of 80, 60, 40 and 30 mL were used. For each LV size two different EFs of 65 % and 45 % were simulated, resulting in a total of eight different virtual patients. To simulate the beating heart and breathing motion, several static phantoms were created for each patient. A heart rate of 60 beats per minute and a 5-s breathing cycle were used. The entire breathing cycle was included, and the heart cycle was divided into eight frames, resulting in a total of 40 static phantoms, with the heart in five different positions for each heart frame. The five simulations for each heart frame were then merged. For each simulation, 32 projections in a 180° arc, starting at the 45° right anterior oblique position were created. A 128 × 128 matrix with a pixel size of 4.8 mm was used. In order to mimic real measurements, the simulations were performed with a sufficient number of photons to avoid Monte Carlo noise. Poisson noise was added after the simulations, corresponding to SPECT with an administered activity of 600 MBq, an uptake of 1 % in the heart and where every projection was measured for 20 s. Tomographic images were reconstructed using the ordered subsets expectation maximization algorithm with four iterations and eight subsets. A gaussian post-filter was applied with a full-width at half-maximum of 10 mm. Short-axis slices were created from the transaxial slices. Scatter and attenuation correction were not applied.

A standard database created in 2007 by the JSNM Working Group for Standardization based on myocardial SPECT studies was used [17]. Patients with only 16 frames per beat were selected (69 patients; 33 women, 36 men). The database included only exercise stress–rest projection sets. The subjects had no ECG abnormalities indicative of ischaemia and no underlying cardiac disease. Subjects with hypertension and diabetes mellitus who required medication and with inappropriate arrhythmia for ECG gating were excluded. Both ^99mTc -labelled hexakis-2-methoxy-isobutylisonitrile (MIBI) and tetrofosmin were included. Wall motion determined by gated SPECT was considered to be normokinetic. Patients who had a normal coronary angiogram as well as those without an indication for coronary angiography because of a low likelihood of ischaemic heart disease were included. Low-energy collimators were used in all subjects. The acquisition angle for each projection was 4–6° per step with a rotation of either 180° or 360°. No attenuation correction was used in any of the hospitals. Patients with ESV <10 mL were excluded.

Consecutive patients with suspicion of ischaemic heart disease and who were undergoing a stress–rest gated myocardial perfusion study were included. The studies were performed from October to December 2011. Of 116 patients, 79 men (mean age 59 ± 11 years) and 37 women (mean age 71 ± 11 years) were included. ^99mTc-MIBI and ^99mTc-tetrofosmin were used in 94 and 22 patients, respectively. The indications for stress myocardial imaging included evaluation of known coronary artery disease (45 %), screening for ischaemia due to multiple risk factors, preoperative evaluation of abdominal aortic aneurysm and peripheral vascular disease (38 %), chest symptoms and ECG abnormality (13 %) and other cardiac diseases including cardiomyopathies (4 %). The patients were classified into the three groups (NH, SH20 and SH10) based on the ESV determined by QGS software. The SPECT data were acquired with steps of 6° over 360°. ECG gating was performed with 16 frames per cardiac cycle. Since the purpose of the study was not to detect ischaemia, only resting gated SPECT data were used in the subsequent analysis.

Gated SPECT data were processed using a standard software package to reconstruct short-axis images. Butterworth and ramp filters were used for filtered back projection reconstruction. To quantify LV function including EF, EDV and ESV, a standard method using QGS software was applied. QGS also calculates the dV/dt parameters as discussed below [14].

The new method for delineation of the LV, EXINI heart (ExH) software (Exini Diagnostics, Lund, Sweden), was based on a previously described method in which a heart-shaped LV model and an active shape algorithm are used instead of geometrical approximations such as ellipsoid or hybrid cylindrical/spherical models [18]. After automated location of the LV, the heart-shaped LV model is adjusted in an iterative process to optimize the fit of the mid-myocardial surface to the three-dimensional image data of the first frame. The endocardial and epicardial surfaces are defined symmetrically on each side of the surface defined by the maximal pixel count along each sampling profile perpendicular to the LV wall. A similar iterative procedure was applied to subsequent frames separately, i.e. no constraints regarding LV basal motion were included. Finally, the LV volume was calculated using the endocardial surface and the LV valve plane. Then EDV, ESV and EF were determined. After the volume curve was fitted by Fourier transformation, the dV/dt curve was created. The summation from the fundamental frequency to the third harmonic was used for the fitting of the original curve and calculated differentiation curve.

The method was adjusted to more accurately process small hearts. In small hearts, due to the partial volume effect and the short distance to the opposite ventricular wall, the ventricular volume appears to be smaller than the actual volume. Therefore, the endocardial and the epicardial surfaces are shifted in the epicardial direction before LV volumes are calculated. The size of the shift is calculated using the mid-ventricular volume of the LV using a univariate polynomial equation of the second degree. The equation was adjusted to give an increasing shift rate for decreasing volumes and a smooth transition to volumes above which no adjustment was performed. The parameters of the equations were determined empirically based on experience from a previous study [6] and analysis of myocardial perfusion studies not used in this study. The final equation was designed to produce a shift of 3.5 mm at a mid-ventricular volume of 0 mL and a decreasing shift up to a volume of 85 ml for which the shift was 0 mm. For mid-ventricular volumes greater than 85 mL, no adjustment was done.

Diastolic dV/dt parameters were calculated [14, 17, 19]. The peak filling rate (PFR) was defined as the maximum dV/dt value divided by EDV (per second). The one-third mean filling rate (1/3MFR) was calculated as the average of dV/dt values in the first third of the filling time (per second). The time to PFR (TPFR) was measured from end-systole to PFR (milliseconds). The ratio of TPFR (milliseconds) to the R-R interval (milliseconds) was also calculated.

All data are expressed as means ± standard deviation (SD). Differences between groups were tested using the t test and analysis of variance. Differences in the data between the software algorithms were evaluated using a paired comparison test. Contingency analysis of categorical variables was also performed, and Fisher’s exact test and the Pearson test were used. Linear regression analysis of LV functional parameters was also performed. In the consecutive patients, comparison among multiple groups was performed for all pairs using the Tukey-Kramer method. A p value <0.05 was considered significant.