Introduction
Electrocardiography (ECG)-gated myocardial perfusion single-photon emission computer tomography (SPECT) is a standard technology used today, and composite information on perfusion and function has been utilized for both diagnostic and prognostic purposes [
1‐
3]. A number of studies since the 1990s have dealt with quantification of SPECT, and functional parameters such as ejection fraction (EF) and left ventricular (LV) volumes correlate well with those from left ventriculography, gated blood-pool studies and MRI. The interinstitutional reproducibility of these parameters is also excellent [
4,
5]. However, it is well known that in subjects with a small LV volume LV end-systolic volume (ESV) is underestimated and EF is overestimated, and the errors are greater in women [
6‐
12]. The so-called “small heart” effect has been found in various multicentre studies particularly in Japan. For example, in a multicentre investigation performed in Japan (J-ACCESS study), three-quarters of women with a low likelihood of coronary artery disease had a small heart [
13]. This finding implies the necessity to use different thresholds between normal and abnormal subjects depending on the size of the heart.
The purpose of this study was first to develop a new method for delineation of the LV that more accurately quantifies small hearts. The new method was then evaluated using digital XCAT phantoms with Monte Carlo simulation, a normal database produced by the Japanese Society of Nuclear Medicine (JSNM) Working Group, and a clinical series of consecutive Japanese patients. EF, end-diastolic volume (EDV), ESV and volume curve differentiation (dV/dt) parameters calculated with the new method and with widely used cardiac quantification software were compared.
Discussion
Inaccuracy of volume and EF calculations in patients with a small heart has been noted, but few solutions or practical correction methods have been proposed. We propose here a new method for LV delineation using a three-dimensional heart-shaped LV model, an active shape algorithm and a volume-dependent algorithm for delineation of the endocardial surface. The important advantage of this approach is that a uniform EF can be used in men and women as well as in small and normal-sized hearts.
The frequency of small hearts is relatively high in some populations. In the Japanese J-ACCESS population who showed a low likelihood of coronary artery disease and no cardiac events, a small heart was observed in 74 % of women and 13 % of men [
13]. In an Iranian study, 85.4 % of subjects had ESV <25 mL, and most of them were women (112/123, 91 %) [
20]. Although the frequency may differ among populations, the number of patients who may potentially require correction is not negligible. The finding that more than 80 % of patients have a suboptimal ESV indicates that the algorithm should be corrected rather than indicating a limitation of the software.
Digital phantoms were used to test the accuracy of volume determination. The digital phantoms were made so that clinical image characteristics were included. Although we created simple “cylindrical plus spherical” models in a previous study [
6], the model used in this study was more realistic and was more reliable for validation. The ExH algorithm, however, still showed a slight underestimation of the volumes. The volume was strictly defined by the phantom, but due to some “blurring” effects caused by scatter, attenuation, depth-dependent resolution degradation, the beating of the heart and respiratory motion, to some extent underestimation of volumes could not be avoided. The volume of half a prolate spheroid with an equatorial radius of 25 mm and a centre to pole distance of 30 mm is 39 mL. If the radius is only 2 mm shorter, the volume is 31 mL, that is a 21 % underestimation. In this study, we could see significant improvement compared with the use of the QGS algorithm.
Normal values have been investigated in various populations including European, North American and Asian [
11,
13,
14,
21‐
24]. In most of the studies, women showed EFs 5–11 % higher and smaller LV volumes than men. Differences between the EFs in men and women are relatively large in Japanese populations. The higher EFs in women are also associated with an EF limit that is as much as 10 % higher than in men. Although physiological differences between the genders might be one of the causes, the relatively higher frequency of small hearts might also be a factor. In this study after correction for the effects of small hearts, no significant differences, or at least only minor differences, were noted between the genders. In some multicentre studies, the threshold values for EF and ESV have been used as prognostic predictors. Therefore, a uniform EF that could be used in both genders and in small and normal-sized hearts would be useful for clinical practice and interinstitutional comparisons.
The dV/dt parameters are affected by the size of the heart. Diastolic parameters have been investigated in an American study, but subjects with a small heart (ESV<20 mL) were excluded [
14]. However, filling rate parameters calculated using QGS showed higher values in women than in men. The PFR was also slightly higher in patients with a small heart. In contrast, the ExH algorithm gave comparable filling parameters between the genders and between normal-sized and small hearts. From a physiological point of view, higher rapid filling in small hearts and in women could not be explained, and was more likely due to a technical problem. Although the magnitude of dV/dt may have differed between genders, the timing parameters TPFR and TPFR/RR were identical between ExH and QGS software.
Some methods for overcoming the effect of heart size have been suggested. A high cut-off frequency of the Butterworth filter during processing could help improve visualization of the endocardial border [
6,
8]. This approach has also been used in paediatric studies. However, changing the filter settings according to the chamber size is not practical for clinical studies, and the filter type used can affect the apparent myocardial tracer distribution in diseased hearts. Zoomed projection images during data acquisition has also been used. Zoom factors, such as 1.2 and 1.5 times, are possible. In clinical settings, however, changing the zoom factor before starting the study is not used in daily practice. Consequently, overestimation of EF and small ESVs are usually reported.
The effect of a small heart may differ depending on the algorithm used for ventricular edge determination [
9,
10]. Common software types have individually developed algorithms to increase the correlation between EF and volumes determined by conventional methods and also to enhance the stability of edge tracing [
25‐
27]. Although the precise steps in the algorithm have not been published, volume-dependent shifting of the edge detection algorithm has not been used. This study indicated the usefulness of volume-dependent correction since the effect of a small heart not only occurs below ESV <20 mL, but occurs gradually as the volume decreases [
6].
The initial application of this algorithm to the clinical setting indicated that EF seemed to be more independent of ventricular volumes. In other words, when QGS software is used, whether or not the studied patients have a small heart should be considered to best judge the abnormality. Although additional studies are required, the results with the ExH algorithm indicate that the correction for a mean volume of <85 mL will work effectively in various populations.
There were some limitations to this study. The true volume was not determined from the patient data. Probably MRI might have been the first choice for this purpose, but it was not performed routinely in our hospital. Second, the algorithm for a small heart could have been improved by adjusting the range of volume for correction. Since the original images were blurred, various methods for improving the resolution might have been added, but this requires further investigation.
Conclusion
The LV volume-dependent correction algorithm for subjects with a small heart provided a more uniform EF between genders as well as between small and normal-sized hearts. Normal values of EF and dV/dt variables were also stable. The application of this software to the clinical setting shows promise.