Are SPECT measurements of myocardial blood flow and flow reserve ready for clinical use?

SPECT myocardial perfusion imaging (MPI) well-documented diagnostic accuracy for detecting coronary artery disease (CAD) has promoted the widespread clinical use of this modality. Nevertheless, SPECT MPI’s success has been attained using the basic SPECT camera design which is over 50 years old [1], using the basic filtered backprojection reconstruction algorithm which is even older, dating to over 90 years ago [2], and using Tc-99m based perfusion agents with limited extraction fractions [3]. Moreover, although SPECT MPI is inherently a digital, quantitative technique, our clinical approach to quantify hypoperfusion has depended on a database approach where a patient’s left ventricular perfusion pattern is statistically compared to an expected normal perfusion pattern generated from patients with low likelihood of CAD [4–6]. In spite of the success of SPECT MPI using these established conventional approaches, recent innovations are poised to bring SPECT MPI to the next level. In a recent article in EJNMMI, Hsu et al. [7] report on the use of dynamic SPECT with a SPECT/CT camera, to measure absolute myocardial blood flow (MBF) and myocardial blood flow reserve (MBFR) in patients. The authors conclude that their flow quantitation method is a clinically effective approach to enhance CAD detection. Thus it is pertinent to ask: Are SPECT measurements of myocardial blood flow and flow reserve ready for clinical use?

Perhaps in a patient-centered environment the first pertinent question should be how will these measurements help our patients? It has been posited that the role of the measurements of myocardial flow and flow reserve should not be limited as a gatekeeper to the catheterization laboratory but more importantly as a gatekeeper to revascularization [8]. So not only should this approach limit the referral of the 60% of patients who are found not to have obstructive disease in the catheterization laboratory [9] but for the 40% with obstructive disease guide the interventionist as to which vessels are truly flow limiting and have the potential for successful revascularization. As such, the measurements fulfill the mantra of today, i.e., an imaging test should not only have the attribute of yielding a correct diagnosis but should also guide a successful therapy thus being directly associated with a patient's outcome that can be used as evidence of the value of the test.

The next pertinent question should be: with today's SPECT instrumentation, radiopharmaceuticals and quantification software - are the measurements of absolute MBF and MBFR accurate and reproducible enough for clinical use? Limited by the lack of dynamic SPECT, early investigations of myocardial flow reserve measurements with SPECT used first-pass dynamic planar imaging of tetrofosmin [10] and sestamibi [11] to record the input function followed by non-AC SPECT for myocardial sampling. A simple microsphere model was used to measure flow and/or flow reserve. Dynamic SPECT and compartmental modeling was also investigated imaging with teboroxime [12] albeit it, at the time, there were very few SPECT cameras that could perform this fast dynamic acquisition. These investigations had in common that the methodology was applied to small patient populations or animal experiments. All of these studies showed the feasibility of measuring flow and particularly flow reserve with SPECT.

In recent years manufacturers have begun to break away from the conventional SPECT imaging approach to create innovative designs of dedicated cardiac imagers. These imagers’ designs have in common that all available detectors are constrained to imaging just the cardiac field of view. These new designs vary in the number and type of scanning or stationary detectors, and whether NaI or solid state detectors are used [13, 14] but have in common an increase in count sensitivity over conventional SPECT up to a factor of 10 [15]. Similarly, iterative reconstruction has significantly evolved allowing for the physical correction of scatter, attenuation, resolution changes with depth, and image noise. Although these might sound familiar today’s techniques are quite more accurate than what was used even 10 years ago. Now instead of implementing simple assumptions as in the past, the entire imaging process is modeled to better correct for these physical phenomena. Two groups have reported on the feasibility of using two of these new cardiac-centric CZT SPECT detector systems to measure flow [16] and flow reserve [16, 17].

Hsu et al [7], in this issue of the Journal, report on the use of dynamic SPECT with a SPECT/CT camera and iterative reconstruction with comprehensive correction methodology to measure absolute MBF and MBFR in humans. These investigators chose to use a standard dual-detector camera equipped with parallel whole collimators. Somewhat less conventional was the fact that the camera could perform a 180° arc acquisition, back and forth, every 10 seconds to record the input function as the first pass of the sestamibi tracer through the ventricles. Also not typical was their use of iterative reconstruction with comprehensive correction methodology which corrected for scatter, attenuation, resolution changes with depth, and image noise. These investigators approach was inspired by the success of PET measurements of myocardial flow and flow reserve but motivated by the need to provide flow measurements in situations where PET and solid state SPECT are not available. They also chose to use the FlowQuant program developed at the University of Ottawa validated for measuring flow with Rb82 [18] but modified to use a single compartment model and the sestamibi extraction fraction.

The simplistic but pertinent answer to determine whether a diagnostic method is ready for clinical use is whether, for a given clinical application, in a given patient, the error of the measurement is such that we can reasonably separate, in this case, normal MBF and/or MBFR from abnormal flow and/or abnormal flow reserve then the technique is ready for clinical use. Hsu et al. [7] report that when the flow results of their 13 patients with CAD were grouped and compared to the results from the 8 patients with no significant lesions there was a statistically significant difference in stress MBF (p=.02) and MFR (p <.001) between the two groups. Even though the patient population is quite small these results are consistent in proving the feasibility of the technique. Yet, when they compare the flow results on a patient by patient basis there is clinically significant overlap between the CAD and non-CAD patients. This overlap is in spite of the 13 CAD patients having an unusually high frequency of multivessel disease (62%). In their hands, in their population, AUC from ROC analysis was significantly greater for MBFR than visual SSS, SDS analysis but stress MBF was not.

The more convoluted answer as to whether these SPECT MBF and MBFR are ready for prime time is whether both efficacy and effectiveness have been established and that we should try to avoid the mistakes of the past when introducing new technology. Because all of the reports on the measurement of MBF and MBFR with SPECT have dealt with very small patient populations or animal experiments there is no evidence of the efficacy and particularly effectiveness of these measurements in the clinical environment. What is clearly established to date with SPECT is the feasibility of performing these myocardial flow measurements. Feasibility is quite important and necessary so the field understands that there is no theoretically inherent limitation to the SPECT approach for measuring myocardial flow. Yet, feasibility alone is not enough to embark on clinical use.

This brings to light the mistakes of our past when introducing new technology. Two technologies come to mind, first-pass radionuclide angiography and SPECT attenuation correction. Even though most experts would agree that both of these techniques have been shown to be highly efficacious the frequency of their current use in patient studies is disappointing. The limiting factor in the use of first-pass studies was the complaint that it was too difficult to perform, in part because of the need for a bolus injection, synchronized start of the acquisition, fast counting cameras, fast framing rates and so on. The initial complain about SPECT attenuation correction was that it did not work, i.e., reduced the diagnostic yield. Later when the methodology matured and shown to work in the daily routine the complaint was that there was no financial incentive to perform this additional task. These technologies shared the common mistake of rushing to clinical use technology before they were ready. The readiness factors include the commercial availability of the right equipment, right radiopharmaceutical, easy to use and robust software, and perhaps most importantly appropriate training for the users. As we learned with SPECT attenuation correction, once a new technology develops a bad reputation it takes years to create the trust for clinical use, even after all the technical problems have been fixed. Moreover, have these techniques obtained wide acceptance, MBF and MBFR could be more readily measured with SPECT today. These flow SPECT measurements require first pass techniques to capture the input function, SPECT attenuation (and other physical phenomena) correction for measuring absolute concentration, cardiac-centric, multi-detector, high count sensitivity cameras to reduce noise, SPECT tracers with higher extraction fraction (such as teboroxime [12] and I-123 rotenone [19]) and flow quantification software similar to that used in PET as done by Hsu et al. [7]. Importantly, although comprehensive correction methodology considerably improves the accuracy of an absolute measurement of concentration, it tends to increase the error with which the measurement is made. This is particularly true for conventional SPECT cameras with low count density in the dynamic acquisitions and a clear disadvantage as compared to PET flow measurements. This is further compounded in SPECT with the use of tracers with low myocardial extraction fraction such as sestamibi and tetrofosmin. Although it is known how to correct for these limited extraction fraction, the correction, in the presence of image noise further propagates the noise and thus the ability to differentiate normal from abnormal flows in a specific patient.

Perhaps one answer would be to use higher radiopharmaceutical doses. Indeed Hsu et al use conventional protocol doses of 370 MBq/1000 MBq for the stress/rest sestamibi studies which results in approximately 12 mSv exposure to the patient. Because of concerns of patient risk due to radiation the trend is moving in the opposite direction, i.e., towards significant dose reductions. The American Society of Nuclear Cardiology published an information statement [20] recommending that laboratories use imaging protocols that achieve on average a radiation exposure of less than or equal to 9 mSv in 50% of the studies. Although there are many different protocols that may be implemented to accomplish this exposure goal, use of the more efficient hardware/software described above would greatly facilitate this goal and allow for improved SPECT measurements of myocardial flow over conventional cameras.

Today in SPECT cardiovascular imaging we have a fast changing field where the instrumentation hardware and reconstruction software have made major leaps in imaging performance consistent with the potential to accurately measure myocardial blood flow and flow reserve. Sestamibi and tetrofosmin radiopharmaceuticals suitable for blood flow measurements, albeit it with limited myocardial extraction fraction, are commercially available. Chemists know how and have in the past developed and validated SPECT tracers with higher extraction fraction than that of Rb-82, clearly showing that there is no inherent limitation of SPECT radiopharmaceuticals to measure myocardial flow. Hsu et al [7] have shown how the validated software used for measuring myocardial blood flow with PET agents can be adapted to model SPECT tracers. Finally, the answer to the posed question of whether SPECT measurements of myocardial blood flow and flow reserve are ready for clinical use is that although there are no real inherent limitations to SPECT to perform these measurements and that investigations like that of Hsu et al [7] have shown this feasibility, the efficacy and effectiveness (8) of these SPECT measurements must be established before the method is ready for clinical use.