Eleven normal adult mice (CD1, four females and seven males, Mouse Clinical Institute, Illkirch, France) weighing 21–47 g, were analyzed according to the French regulations concerning experimentation on small animals (authorization No. A 67-482-20). For 2D transthoracic echography and blood-pool gated SPECT, mice were anaesthetized with a gas mixture (air and 1–1.5% isoflurane). For blood-pool gated SPECT, mice preparation, conditions and routes of tracer administration were always conducted in the same manner. One extremity of the mice holder for SPECT was attached to a respiratory tube connected with a small animal anaesthesia device able to deliver the heated gas mixture (air and 1–1.5% isoflurane) as illustrated in Fig. 1 and able to maintain the temperature of the animal at 36 ± 1 °C. The caudal vein was catheterized for intravenous delivery of 550 MBq of ^99mTc human albumin (Vasculocis^® Cis-Bio International, Paris, France). Care was taken to keep the volume of injected tracer under 0.2 ml to avoid significant changes in whole blood volume of mice.

A small-animal one-head dedicated gamma camera (Gaede Medizinsysteme GMBH, Freiburg Germany) with a 6.5-mm-thick NaI(Tl) crystal equipped with 25 photomultipliers and a small field of view of 17 cm × 17 cm was employed in this study. The camera was equipped with a pinhole collimator of 12 cm in focal length and a tungsten insert of 1.5 mm in diameter (Fig. 1). The computer-driven circular rotation of the camera permitted to test the accuracy of the centre of rotation, and pre-processing correction was applied on the projection images before tomographic reconstruction using the line source method described in [24]. Cardiac acquisitions were started 30 min after injection of the radiopharmaceutical. The camera radius of rotation was adjusted as close as possible within the mouse thorax (2.5 cm in our case) to improve projection images magnification, corresponding in that case to a zoom factor of approximately 5. A 20% window centred on the 140-keV photoelectric peak of the ^99mTc was used. Mouse leading supine, a circular ventral 48-projection step and shoot acquisition over 180 ° with 64×64 word images from the left lateral to the right lateral side of the thorax was always used. For ECG gating, we used two paddle electrodes connected to an electrocardiograph (Physiogard RSM 784 ODAM, Wissembourg, France) and 350 cardiac beats were registered for each projection in a forward-backward mode with a 20% temporal window of the mean RR time and eight time bins per cardiac cycle. At the end of the exam, data were transferred to a computer (G4, Apple, Cupertino USA) for image reconstruction.

The EAR method (Efficient Algebraic Reconstruction) used for images reconstruction was developed in our laboratory and is described in details elsewhere [25]. The number of voxels and the dimensions of the field of view permit to predetermine the isotropic tomographic reconstructed voxel size, set in our case at 0.55×0.55×0.55mm3. The physical parameters, as attenuation and diffusion, were not taken into account in this study, but could be introduced in the algorithm. Before reconstruction, image projections were corrected for radioactive decay of ^99mTc due to long time acquisition. Previous studies of our group have shown that using the EAR algorithm and the pinhole gamma camera described before, a SPECT spatial resolution of 1 mm and a volume calibration linear relationship in the range 10–100 μl can be achieved with a 1.5-mm pinhole [19,25].

A 4D amplitude and phase analysis was applied to blood-pool gated SPECT images in order to separate the kinetic behaviour of atrium and ventricles chambers during the cardiac cycle. Matrix manipulation, measurements, thresholding and automatic contouring procedures were written using ImageJ software (W.S. Rasband, ImageJ, U.S. National Institutes of Health, Bethesda, MD, USA, http://rsb.info.nih.gov/ij/, 1997–2007) following four steps illustrated in Fig. 2.

2.4.1 Step 1: volume reorientation and filtering

2.4.2 Step 2: atrio-ventricular separation

2.4.3 Step 3: inter-ventricular separation

2.4.4 Step 4: definition of left and right boundaries and volumes measurements

A paired Student's t-test was used for direct comparison of LV an RV volume measurements. Linear regression analysis was used to test for correlation of LV and RV stroke volume measurements. Bland–Altman analysis was carried out by comparing differences in assessed RV and LV volumes with the mean of RV and LV volume comparisons. All results are given as means ± SD. Values of P<0.05 were considered significant.

In our experiments, an isotropic spatial resolution of 1 mm was obtained by pinhole SPECT with a 1.5-mm pinhole and a reconstructed voxel size of 0.55×0.55×0.55mm3, which has to be compared to the 0.75-mm spatial resolution with pinholes of 1 mm and a triple head camera with a reconstructed voxel size of 0.5×0.5×0.5mm3 obtained by Chin et al. [20]. However, isotropic spatial resolution of 0.5 mm can be actually achieved with different commercial small-animal-dedicated SPECT devices, leading in the future to a better delineation of the ventricles [27]. Regarding the time resolution that can be obtained in gated studies, we were able to measure eight time bins per cardiac cycle for a total acquisition time of 48 min with a simple pinhole gamma camera. As the sensitivity of the recently proposed multi-detector and/or multi-pinhole SPECT cameras is greatly improved compared to our single pinhole device, 16 phases per cardiac cycle are expected with reasonable acquisition times [3,14,16,20]. However, it is also important to remind for comparison that planar radionuclide ventriculography was able to provide ejection fraction curves with a high time resolution (up to 160 frames/s), as it was shown by Hartley et al. [17] in first pass studies using the short half live ¹⁷⁸Ta and a specialized multiwire gamma camera.

Adult healthy mice values of both left and right ventricular volumes, ejection fraction and cardiac output measured in vivo by different imaging methods spread out over a wide range according to the literature. This dispersion depends on the imaging techniques involved (2D/3D echography, MRI, CT, SPECT), but manual or automatic images analysis employed for contouring endocardial borders and ventricles' volume modelling using ventricular diameters or cross-sectional measurements are also concerned. Mice strains and reduced heart rates due to adverse impact of anaesthesia could also influence these values [28]. This is illustrated in Table 3, where for left ventricle volumes it can be observed in the literature that average heart rates are in the range 331 to 600 bpm, while average EDV and ESV vary from 18 to 68 μl and 4 to 24 μl, respectively. This dispersion is even more pronounced for the right ventricle with average EDV varying from 34 to 84 μl and ESV from 18 to 36 μl, as there is less corresponding imaging data available.

The mean values and standard deviation of left ventricular EDV, ESV, EF and cardiac output that we obtained are in close agreement with those published by Chin et al. [20], confirming the robustness of the blood-pool gated pinhole SPECT method.

Echocardiography based on high-frequency transducers is widely used and is able to measure internal left ventricular diameters in systole and diastole in 2D selected short- and long-axis views as in apical 4-chamber view. After manual tracing of LV diameters and cross-sectional surfaces, functional parameters could be derived using different formulas (shortening fraction on the one hand and Teicholz, prolate ellipsoid or Simpson's rule for volumes estimation on the other one) [5]. Tissue and colour Doppler are also added values for diastolic and flow parameters. The advantage of echocardiography is an excellent time resolution [7,29]. 3D echography by controlled translation of 2D probes, or in the future real time 3D echography, should be able to overcome the current limits of 2D echocardiography, entailing assumptions of left ventricular shape for direct volumes measurements but at the expense of a relative loss of resolution in two directions [6]. However and not surprisingly, echography, based on two different formulas and methods (Teicholz and Doppler), frequently overestimates the left cardiac output and shows also important differences on a mouse-per-mouse basis, illustrating the inaccuracy of 2D transthoracic echography for volumes assessment [30,31]. In comparison, blood-pool gated SPECT has an intrinsic isotropic spatial resolution enabling left volumes' measurement without any ventricular shape assumption. Cardiac cine magnetic resonance imaging (bright or black blood MRI) for assessment of left ventricular volumes and function in mouse are currently based on 2D multi-slice techniques with 12–16 bins per cardiac cycle and acquisition times ranging from 15 to 60 minutes for in-plane resolutions of 0.1–0.3 mm and slice thicknesses of 1–2 mm depending on the imaging protocols used [8,9]. The left ventricular volumes measurements are then obtained after manual delineation of endocardial borders of consecutive 2D slices encompassing the heart [32,33]. 2D multi-slice MRI techniques are limited by the necessary trade-off between high temporal resolution and high in-plane spatial resolution, leading in some experiments to the acquisition of only one mid ventricular slice with a thickness of 1 mm, but 30 images per cardiac cycle and 0.1×0.1mm2 pixel size [10,34]. The values of the left ventricular function we obtained in normal mice by blood-pool gated SPECT are in the range of MRI results obtained by different groups [8,9,30,33]. Cine MRI was recently improved by 3D isotropic gradient echo sequences yielding at 7 T a voxel resolution of 0.2×0.2×0.2mm3 and 12 time bins per cardiac cycle at the expense of a very long acquisition time (around 3 h), but leading in the future to more accurate left ventricle volumes' measurements [11]. However, an important hurdle to widespread implementation of MRI for small animal cardiac studies compared to echography and pinhole camera used in this study is the cost and availability of clinical or small-animal-dedicated high-field MR imagers [35]. Using a custom-built set-up, with the mouse placed vertical in a rotating holder and a delayed enhancement double-contrast protocol after administration of a long-lasting blood-pool contrast agent, cine micro-CT has recently shown a superior isotropic spatial resolution, with 0.1×0.1×0.1mm3 voxels and 15 ms of time resolution resulting in eight phases per cardiac cycle for a total acquisition time of about 8 min [12,13]. As for cine micro CT and isotropic 3D cine MRI, pinhole gated SPECT can provide the four-dimensional data (3D isotropic volumes over time) that are required for cardiac function in mice. Moreover, perfusion and blood-pool tracers with different isotopes (²⁰¹Tl and ^99mTc-blood pool) can be used with pinhole gated SPECT, as in dual-isotope human protocols, to simultaneously analyze wall perfusion and function of the left ventricle in a single exam [36].

Addressing the RV structure and function is a difficult task and the invasive assessment of right ventricular volumes using the manual pull back of a transoesophagal ultrasound probe has shown significantly higher intra- and inter-observer variability, limiting therefore its application [22]. As previously described, isotropic 3D cine MRI at 7 T has recently demonstrated his efficacy for accurate measurements of right ventricular volumes despite longer acquisition times compared with 2D cine MRI [11]. To our knowledge, there are no data about RV volumes measurements by blood-pool contrast micro CT. We show for the first time, to our knowledge, that blood-pool pinhole gated SPECT is able to measure RV volumes in mice without any volume shape assumption. Values of RV volumes found are in the range of published data using other small animal imaging methods in healthy mice, but keeping in mind that very few data are available (see Tables 2 and 3). However, and as expected, we found on the one hand that left and right cardiac outputs we measured were not statistically different and on the other hand that right and left stroke volumes were linearly correlated. These results are in agreement with physiology and were also demonstrated using high-resolution MRI at 7 T by Wiesmann et al. [23].

For the particular anatomy and size of mice cardiac chambers, the use of a 4D harmonic analysis is a useful preliminary step for automatic atria and ventricles separation before functional ventricular analysis. We chose to apply temporal harmonic analysis on the reconstructed blood-pool voxels instead of temporal Fourier filtering of the gated projections before slice reconstruction because the amount of memory needed is actually not critical [37]. However, a reduction in noise in the images could be expected if temporal Fourier filtering of the projections is applied before reconstruction. 4D harmonic analysis of blood-pool gated images had also the potential of being a useful tool for qualitative analysis of impaired left ventricular wall motion [20].

Contrary to intuition, the amount of radioactivity used in this study was similar to that employed for humans, despite important differences in body mass between mouse and man. This is due to the size of the image voxel, which corresponds to mice scale, as well as to the low sensitivity of the single pinhole employed in this study. However, the amount of radioactivity we used was similar to that of Chin et al. [20]. Recent commercial multi-pinhole SPECT devices are now expected to dramatically lower this amount of radioactivity due to the large number of pinholes employed, while keeping a high spatial resolution, in order to reduce mice organs and whole body absorbed doses, which is an important issue in case of longitudinal examinations [14,15].

The principal limitation of this study was the lack of direct comparison of left and right volumes measured by ECG gated blood-pool SPECT with another method such as high-field MRI. However, for left ventricular volumes and global function in healthy mice, our results are in good agreement with previous data of our group on perfusion gated SPECT and those of Chin et al. [20] with a comparable technique and also with high-field MRI. For right ventricular volumes and function, further investigations are needed to confirm and validate the ECG gated blood-pool pinhole SPECT in mice by direct comparison with MRI or contrast ECG gated micro CT [13]. In this study, we used radiolabelled human albumin as a vascular tracer, instead of in vivo radio labelling of red blood cells, with the benefit of using only a single injection, but the disadvantage of potentially inducing an immune response.