Left ventricular ejection fraction (LVEF) is one of the most commonly reported measures of left ventricular (LV) systolic function. It is the ratio of blood ejected during systole (stroke volume) to blood in the ventricle at the end of diastole (end-diastolic volume). If the LV end-diastolic volume (EDV) and end-systolic volume (ESV) are known, LVEF can be determined using the following equation:
LVEF = stroke volume (EDV - ESV) ÷ EDV
LVEF can be determined using several invasive and non-invasive imaging modalities, either subjectively by visual estimation or objectively by quantitative methods. This article focuses on the common sources of error in the quantitative calculation of the LVEF using non-invasive methods – i.e., echocardiography, magnetic resonance imaging (MRI), computed tomography (CT), gated equilibrium radionuclide angiography (commonly referred to as multiple-gated acquisition [MUGA] scan) and gated myocardial perfusion imaging with either single-photon emission computed tomography (SPECT) or positron emission tomography (PET). LVEF can also be measured non-invasively using the ‘first-pass’ radionuclide technique, but this technique is rarely performed in the current era and will not be addressed further here.
Currently, there is no universally accepted ‘gold standard’ for measuring LVEF. Every method and modality used to measure the LVEF is subject to factors that may introduce error and/or variability into the calculated ejection fraction. Because there is no gold standard, the choice of modality used should depend on patient factors, local resources, other information desired from the study and the need for follow-up measurements. Knowing the methods and their pitfalls can help clinicians understand the sources of variability when measuring LVEF and choose the adequate method. For example, all techniques – except echocardiography – require obtaining information over several cardiac cycles, the image acquisition being ‘gated’ with the patient’s electrocardiogram (ECG), and calculated LVEF in patients with significantly irregular rhythms are often inaccurate if methods requiring gating are being used.
General Sources of Error
With any method, the endocardial border needs to be accurately detected to ensure accurate LV cavity detection and LVEF calculation. With methods requiring manual measurement of the LV cavity, errors can arise from differences in the perception of the endocardial border between different people or from variations in measurement technique. With automated or semi-automated methods, human variability is limited, but errors can arise from differences in the software algorithms used to detect the LV cavity border. Also important for LVEF calculation is the accurate detection of true end-diastole and end-systole.
LVEF will be underestimated if either the true end-diastole (underestimated EDV) or end-systole (overestimated ESV) is not measured. With manual techniques, this depends on accurate human detection of end-diastole and end-systole. With all techniques, this also depends on sufficient temporal sampling of LV volumes to detect the largest and smallest LV ones. Techniques with lower temporal sampling may not image the true end-diastole or end-systole.
Many methods have been developed to measure LVEF using echocardiography.1 The methods differ based on the type of echocardiographic image used (M-mode, 2D or 3D), the measurements needed and the equations/assumptions used to determine LV volumes. The measurements obtained can be linear (1D), area (2D) or volume (3D) measurements. Advantages of all echocardiographic techniques include lack of ionising radiation and portability.
M-mode and 2D Echocardiography
The modified Quinones method is commonly used and employs linear measurements.2 It uses single measurements of the LV cavity in the mid-ventricle in both end-diastole and end-systole, and can be employed using either M-mode (see Figure 1) or 2D imaging (see Figure 2). Because only two linear measurements are needed, the modified Quinones method often requires less time to perform than other methods and only needs adequate visualisation of the endocardium in the mid-ventricle. However, because this method only measures circumferential contraction in a single plane, significant assumptions about LV chamber geometry have to be made so that a change in a linear measurement can act as a surrogate for a change in volume. The contribution from longitudinal contraction on LVEF is not directly measured, but rather a correction factor is used to adjust the measured LVEF based on a visual grading of the apical contraction.2 In situations where there is asymmetry of contraction (e.g., wall motion abnormalities), such as in ischaemic LV dysfunction, this method has significant limitations. Other methods that use only linear measurements have also been previously described, but have fallen out of favour because of similar limitations.2,3 The biplane method of disks (modified Simpson method) is a 2D echocardiographic technique requiring area tracings of the LV cavity. This is the method recommended by the American Society of Echocardiography for measuring LVEF.4 It requires tracing the LV endocardial border in the apical 4- and 2-chamber views in both end-diastole and end-systole (see Figure 3).
The tracings are used to divide the LV cavity into a predetermined number of disks (usually 20) with disk volumes based on the tracings. The method requires fewer geometric assumptions of LV shape than the modified Quinones method, and it directly measures the contribution of longitudinal contraction; however, some geometric assumptions still need to be made, because not the entire LV cavity border is traced. Also, the four tracings take longer to record than the two measurements needed with the modified Quinones method. Other methods that use area measurements of the LV cavity have also been described.1
Sources of Error
The main limitation of both the 1D and 2D methods described above is that they are less accurate in patients with regional variation in systolic function, as the measurements can be obtained from a region of the LV cavity where the function is discordant from the overall ventricular function (see Figure 4). Both methods may also have reduced accuracy if the imaging planes used for the measurement are incorrect (i.e., off-axis or foreshortened). With the modified Quinones method, the cavity size can be overestimated if the ultrasound beam/image is not perpendicular to the ventricular cavity, or underestimated if the ultrasound beam/image is not in the centre of the cavity. With the modified Simpson method, the ventricle can be foreshortened on apical images, so that the true apex will not be imaged (see Figure 5). This can lead to errors if apical function is discordant, as can happen with an apical infarct.
All echocardiographic methods require acoustic windows that allow adequate visualisation of the blood/endocardial border to allow accurate measurement/tracing. Obese patients, patients with chronic obstructive pulmonary disease and patients with limited space between the ribs will often have poor image quality (see Figure 6A). Studies have reported that LVEF could not be determined using the modified Simpson method in 31–38 % of patients due poor image quality. If measurements are made in patients with poor image quality, they can be inaccurate because the endocardial border cannot be properly visualised and traced/measured. The use of echocardiography contrast has been shown to improve LVEF determination in patients with poor acoustic windows and reduce inter-observer variability (see Figure 6B).5
In addition to M-mode and 2D echocardiography, LVEF can also be calculated using 3D echocardiography. Several reconstruction techniques have been developed to acquire 3D data of the heart from which LV volumes can be calculated.6 Most 3D imaging techniques require that data be acquired over several heartbeats using special 3D imaging probes.
Unlike other echocardiographic methods, 3D methods make minimal assumptions of LV cavity shape. They have been shown to be less variable and more accurate than other echocardiographic methods, when compared with MRI as a reference standard.7 This is because, with 3D methods, the entire LV cavity will be detected. However, the acoustic window needs to be of sufficient quality to allow the delineation of the entire LV cavity endocardial border.
Sources of Error
Because the image data are usually acquired over several heart beats, an ectopic beat or breathing during the imaging time will lead to artefacts which can alter the endocardial border (see Figure 7) and different segments of the left ventricle will appear to contract at different times. Also, LVEF is usually calculated using commercially available software that is semi-automated, requiring the user to manually assign certain points (e.g., mitral annulus) in the left ventricle. If the points are improperly assigned, the LVEF may be inaccurate. The location of the endocardial border may also be incorrectly interpreted, either by the operator or the software due to poor definition, and a papillary muscle or trabecula may be interpreted as being the endocardial surface.8
Magnetic Resonance Imaging
LVEF can be calculated with MRI using manual, semi-automated or automated methods.9 The most commonly employed method is the Simpson disk summation method using short-axis cine steady-state free precession images of the left ventricle.9,10 LV endocardial borders are manually traced on each short-axis image obtained in both end-diastole and end-systole to determine the ventricular cavity area for each slice. The area of the tracing for each image slice is multiplied by the slice interval (slice thickness plus image gap) to determine a volume for that slice. The volumes of the slices are summed to determine an LV volume. This method requires few assumptions of LV shape because the entire LV cavity is traced. Because of the high contrast resolution and high signal:noise ratio of MRI, the endocardial border is usually well defined. MRI does not require the use of ionising radiation or contrast material for the calculation of LVEF. MRI is contraindicated in patients with implantable cardioverter defibrillators, most pacemakers and several other types of implanted devices.
Sources of Error
There are several limitations to the use of MRI to calculate LVEF. Cardiac MRI requires multiple breath holds, and image quality may be poor in patients who cannot hold their breath. If the level of inspiration is different during the acquisition of different levels, segments of the LV may not be imaged while other segments may be imaged twice. This may lead to variability in calculated volumes and LVEF. Because data are acquired over several cardiac cycles with ECG gating, image quality will be degraded in patients with cardiac arrhythmias or ectopic beats leading to decreased accuracy. LV volumes will vary based on whether the papillary muscles and/or trabeculations are included in the LV cavity (see Figure 8). One study showed that including trabeculae in the LV cavity volume led to significantly decreased LVEFs (-2 %, ±2 %) compared with excluding the trabeculae.11 Another area of potential error in LV volumes and LVEF calculation is the selection of the final ventricular basal slice (see Figure 9). In a study by Karamitsos et al., the selection of the end-systolic basal slice was a major source of error in LVEF calculation.12 In that study, the ESV was overestimated by 10.2 ml (±5.1) ml if an extra basal slice was included in systole (underestimating LVEF), and underestimated by 6.3 ml (±7.6 ml) if one basal slice was excluded from the ESV measurement (overestimating LVEF).
As with MRI, the most commonly used method for calculating LV volumes and LVEF using CT is the Simpson method – although other methods, including automated ones, are available.13 Iodinated contrast is required to differentiate the blood/endocardial border because of the poor differentiation on non-contrast images. The automated methods usually rely on the differentiation of the LV cavity from the endocardium based on Hounsfield unit measurements.13 With the Simpson method, reconstructed short-axis cine images of the heart are created and traced for LVEF calculation. Similarly to MRI, few assumptions of LV shape need to be made when using the Simpson method because the entire LV cavity is traced. As long as the contrast bolus timing is appropriate, there will be high contrast and spacial resolution resulting in a well defined endocardial border. An advantage of CT over MRI is that CT images can be obtained with a single breath hold. Disadvantages of CT are the exposure of the patient to ionising radiation and the need for iodinated contrast material. Iodinated contrast material should be not be used in patients with iodinated contrast allergies – unless they have been pre-medicated to avoid any allergic reaction – and should be used judiciously in patients with poor renal function.
Sources of Error
Many of the pitfalls of using CT to measure LVEF are similar to MRI. LV volumes will vary based on whether or not the papillary muscles and/or trabeculations are included in the LV cavity. Also, variability in the selection of the ventricular basal segment will cause variability in LVEF calculation when using the Simpson method. As with MRI, cardiac CT requires ECG gating for image reconstruction, so image quality will be degraded in patients with cardiac arrhythmias or ectopic beats (see Figure 10), which will reduce the accuracy of LVEF calculation. Breathing during image acquisition can also lead to artefacts, which can reduce accuracy. Because of the need for intravenous (IV) contrast to delineate the endocardium, a problem that is unique to CT is the need for proper coordination of the timing of contrast injection and scanning. Poor contrast enhancement of the left ventricle may prevent accurate differentiation and detection of the blood/endocardial border, especially with automated methods (see Figure 11).13 Also, depending on the CT scanner technology, temporal resolution may not be fast enough to precisely image end-systole.13 Image quality may be poor in obese patients.
Nuclear Cardiac Imaging
LVEF can be calculated by several methods using different nuclear cardiac imaging techniques. Two of the most common ways of calculating LVEF using radioisotopes are radionuclide angiography and gated myocardial perfusion imaging with either SPECT or PET.
Radionuclide angiography (also referred to as gated equilibrium blood pool method or MUGA scan) is performed by labelling a patient’s red blood cells (RBCs) with technetium 99m pertechnetate. Most commonly, planar images of the left ventricle are acquired for analysis, although SPECT images can also be acquired. If planar imaging is used, a left anterior oblique projection with best separation of the left and right ventricle is acquired for LVEF calculation. Radioactivity counts within a LV region of interest (ROI) are determined. MUGA scan measures changes in radioactivity in the left ventricle between end-diastole and end-systole, rather than truly measuring LV volumes. Assignment of LV ROI can be automated, semi-automated or manual, automated and semi-automated edge detection being most commonly used.14
Image acquisition is gated with an ECG and radioactive counts are acquired over multiple cardiac cycles. Each cardiac cycle is divided into a predetermined number of intervals (usually 16 or 32), corresponding to the number of frames (images) per cardiac cycle. The frame with the highest counts is considered end-diastole and the frame with the lowest counts is considered end-systole.
LVEF equals net counts in the end-diastolic frame minus net counts in end-systolic frame divided by net counts in end-diastole. Net counts are calculated by subtracting counts from a background ROI (placed next to the left ventricle) from measured LV counts.14 With this technique, no geometric assumptions of LV cavity shape need to be made to calculate LVEF. It can be performed in patients whose body habitus might prevent or limit the reliable use of other techniques. There are no absolute contraindications to this technique.
Sources of Error
Because the acquisition of counts is gated with the patient’s ECG, assumes a regular R-R interval and is averaged over several cardiac cycles, arrhythmias (e.g., ectopic beats or atrial fibrillation) will lead to artificially reduced counts in frames later in the cardiac cycle (end-diastole). This will reduce the accuracy of LVEF calculations in patients with arrhythmias.14 The LV ROI measurement may be inaccurate if the left atrium, right ventricle, aorta, pulmonary artery or spleen is included within the LV ROI due to poor positioning or anatomic variations (see Figure 12). In addition, poor labelling of the RBCs can occur, which can lead to poor count rates within the blood pool and increased background counts. If the target:background ratio is too low, the LV ROI edge may be improperly detected, leading to reduced accuracy of the LVEF (see Figure 13).
Background counts are subtracted from the LV end-diastolic counts in the denominator of the LVEF calculation. If background counts are overestimated (e.g., background ROI over the spleen), the LVEF will be overestimated (see Figure 14). On the other hand, if background counts are underestimated (e.g., background ROI placed over a pleural effusion), the LVEF will be underestimated (see Figure 14C). Differences in software edge detection algorithms have also been shown to lead to variation in the calculated LVEF. In one study, the average LVEF showed considerable variation (up to 8 % difference in calculated means) between institutions using the same raw data but different processing systems.15
Gated Myocardial Perfusion Single-photon Emission Computed Tomography and Positron Emission Tomography
Gated myocardial perfusion SPECT is performed by injecting a patient with a radiolabelled myocardial perfusion agent such as technetium 99m radiolabelled sestamibi or tetrofosmin. LVEF can also be calculated in a similar manner using gated myocardial perfusion or viability PET. Ammonia, rubidium or fluorodeoxyglucose can be used as imaging agents.
The LV functional assessment (with LVEF calculation) is usually done in conjunction with a myocardial perfusion study,16–18 allowing function and perfusion to be evaluated with one test. After injection, the radiopharmaceutical is taken up by the myocardium. After allowing a sufficient amount of time for the radiotracer to be taken up by the tissues and cleared from the blood pool, ECG-gated images are acquired.18 The ECG gating is used to divide the cardiac cycle into a predetermined number of frames (images) per cycle (usually eight or 16).16,17
LVEF is determined quantitatively by analysing a reconstructed three-dimensional data set using software with automated edge detection. Because a three-dimensional data set is used, few geometric assumptions need to be made about the LV cavity shape. The software automatically determines the border between the count high LV myocardium and the count poor LV cavity. The volume within this border, corresponding to LV cavity volume, is calculated during each of the cardiac cycle frames to determine the EDV and ESV.16,18
Sources of Error
Because this method is ECG-gated, it is susceptible to problems due to arrhythmias just as MUGA scan, MRI and CT are. For this reason, LVEF calculation using gated myocardial perfusion is not recommended for patients with arrhythmias such as atrial fibrillation, frequent premature ectopic beats or heart block.16,18 In patients with perfusion defects (e.g., prior myocardial infarction), detection of the myocardial border may be limited. If the perfusion defect is so severe that the border cannot be accurately determined, the LVEF may be inaccurate (see Figure 15).19 Even with severe defects, however, the effect on LVEF may be minimal with the currently available automated programmes.20
In patients with small LV cavities, the LV ESV may be underestimated. This occurs because there is blurring of the LV cavity border due to the relative poor resolution of the gamma camera compared with LV wall thickness, and because of an increase in myocardial count density with contraction (see Figure 16). LVEF will be overestimated because of underestimation of the ESV.21,22 LVEF is often overestimated in women because women tend to have smaller LV cavities.
Several studies have shown a systematically decreased LVEF (2–6 % across the LVEF range) when eight frames per cardiac cycle are used for LVEF calculation rather than 16 frames.23,24 This is believed to be due to temporal undersampling of LV volumes in which end-diastole or, more often, end-systole is not truly imaged or is ‘blurred’.
As with other automated or semi-automated methods of calculating LVEF, the calculated LVEF can vary between different processing systems, even when identical raw data are being analysed. Several studies have shown variation in LVEF using the most popular gated SPECT analytic software.25,26 In one study using identical raw data from normal controls, mean LVEF varied from 60.6 % to 53.2 % between different processing systems.25 LVEF may also vary with gated myocardial perfusion SPECT due to a true physiological change. The ‘resting’ LVEF with gated SPECT is often obtained 15–60 minutes following the stress portion of the test, especially with one day, low-dose rest, high-dose stress protocols. With other techniques, LVEF is calculated with the patient truly at rest. Some people with severe ischaemia have been shown to have a true decrease in LVEF due to prolonged stunning with post-stress imaging.27
Some of the variation in LVEF calculations may be intrinsic to the way in which the data are processed within an individual institution (e.g., software used or measurement technique). Because of this, each institution should determine its own normal ranges for each modality/method, so that clinicians understand and realise that the normal ranges for LVEF can vary. For example, a LVEF of 45 % from a SPECT study may be normal for one particular nuclear laboratory, but may be abnormal if obtained by echocardiography or gated SPECT in a different nuclear laboratory using different software. Because of the variation that can occur between modalities and institutions, if serial measurements of the LVEF are needed, the same modality should be used to limit intermodality variation.7,25,28 However, it should be realised that, even when the LVEF is calculated in the same manner at the same time, there can be variability when it is measured by different people or the same person twice (inter- and intra-observer variability). Also, LVEF measured by the same technique in the same patient can vary over short time periods, since it is dependent on loading conditions of the left ventricle, which may not be constant.29,30
Knowing the inter- and intra-observer variability of LVEF measurements is important, because these are used to define the magnitude of change in LVEF that is outside the accepted measurement error, and also because low inter-observer variability is often used as a surrogate for accuracy – as there is currently no reference gold standard with which non-invasive LVEF measurement can be compared.
However, it is very difficult to formally compare the inter- and intra-observer variability of different LVEF measurement techniques reported in the literature, as there is no widely accepted statistical methodology for measuring variability and because the variability of a given technique changes with the patient population. It is outside the scope of this article to formally compare the inter- and intra-observer variability of the techniques or the accuracy of one technique versus another.
In general, however, inter-observer variability increases with operator dependence and the use of geometric assumptions, making 1D and 2D methods most prone to variations due to geometric assumptions.7,31,32 Volumetric methods (3D echocardiography, MRI, CT, PET and SPECT) are thought of as being less prone to variations because they use few geometric assumptions. In patients with regular heart rates, the extent of variability of a volumetric technique depends on both its ability to accurately delineate the LV cavity and the amount of operator interaction, especially in defining the LV–left atrial interface. Among the volumetric imaging techniques, MRI is generally considered to have the least variability.8,33 Of all techniques (volumetric and non-volumetric), MUGA scan has the least operator dependence and is generally considered to have the least variability.19,29,34–36
There are many ways of measuring the LVEF and each method has limitations and potential for error. Because prognostic and therapeutic decisions are based on the LVEF, it is important to understand why errors occur so that all efforts can be made to limit them if possible. Referring clinicians should have a basic understanding of these methods so that they can make informed decisions about which modality is most appropriate for an individual patient. Technologists and interpreting physicians should also have an understanding of the way in which the LVEF is determined so they know where problems can arise and optimise studies accordingly. Image quality should be analysed with every study and, in cases where it is felt to be insufficient to accurately determine the LVEF, this should be reported.
Many factors should be taken into account when deciding which method is the most appropriate for an individual patient. These factors include what other information (such as valve function, viability, etc.) is desired from the study, cost, patient age, contraindications and local availability of techniques. Also, if a measurement by one method is inconsistent with the clinical findings, measuring the LVEF with another technique should be considered.