TO THE EDITOR:

We read with interest the article by Blagosklonov et al. (1), which suggests a possible explanation for motionlike artifacts (MA) associated with ²⁰¹Tl SPECT myocardial perfusion imaging (MPI). As they and others note (their references 4–8, 11, and 12), the history of rotational SPECT MPI is checkered with artifact issues prompting a variety of protocol, hardware, and software modifications intended as corrective measures, including dual-isotope imaging, prone imaging, breast binding, motion correction, resolution recovery, and attenuation correction. Still, in the presence of all of these remedial measures, the incidence of false-positive MPI studies remains in the 20%–40% range (2). Some of these proposed corrective measures have been found to actually increase the potential for false-positive MPI studies.

The work of Blagosklonov et al. (1) is intended to show that the rapid early washout (REW) of thallium is a significant cause of MA in myocardial perfusion images that are acquired on dual-head SPECT systems when imaging commences too soon after cessation of exercise. It is true that the 2-compartment model for uptake and redistribution of a diffusible tracer predicts REW of the tracer after the peak of maximum uptake, which is then followed 10–20 min later by transition to a more gradual or steady-state washout phenomenon. The 12-min delay in initiation of MPI proposed in the article of Blagosklonov et al. is intended to avoid initiation of image acquisition during this period of REW. Unfortunately, their work as presented does not conclusively demonstrate that this is the sole or even a plausible explanation for MA.

In particular, 2 aspects related to their own data could have been further explored to conclusively demonstrate the cause-and-effect relationship between REW and MA. First, the REW phenomenon in their data could be approximately corrected for by applying an image-by-image scale factor to their stress images based on the well-behaved difference between the pairs of stress/rest curves as shown in their Figure 6. The differences between these 2 curves could serve as a good estimate of the relative counting rate alterations in the stress images caused by the REW phase. Admittedly, this type of correction would be an approximation of what is actually going on in all regions of the image, but in normal myocardium all regions of the heart will demonstrate similar uptake and washout dynamics. Therefore, if the authors’ premise is correct, the application of this approximate corrective technique should demonstrate a reduction in the MA shown in the images of their Figure 1.

A second opportunity to clearly establish the proposed causal relationship between REW and MA would use the SPECT data that the authors obtained using the Mayo cardiac phantom filled with ^99mTc. Unfortunately, the only application of this phantom by the authors was to verify that imaging an object in the presence of a rapidly decaying count rate (simulating REW) would indeed result in a discontinuity between their images 16 and 17. If the authors had taken an additional step and reconstructed the 2 sets of SPECT phantom images (acquired with 25 s per frame vs. 120 s per frame), they should then have been able to demonstrate MA in the 120-s phantom images corresponding to those shown in their Figure 1. Again, we are left without conclusive proof.

The lack of definitive data to support the authors’ premise leads us to suggest an alternative explanation. It is well known that cardiac creep during sequential image acquisition has the potential to cause artifacts in tomographically reconstructed images (their references 11 and 12). Furthermore, the literature contains ample evidence that cardiac volume changes of 10%–15% are routinely seen during the 15 min after exercise, when the patient reclines for supine imaging (3,4). Furthermore, these changes have been demonstrated to be a function of position (5).

In general, tomographic reconstruction algorithms behave somewhat poorly in the presence of inconsistencies in the sequential image dataset to which they are applied. Therefore, changes in the actual size (volume) of the left ventricle could likely be another potential cause of MA. This fact also serves to explain the improvements in SPECT myocardial perfusion image quality observed when a prone imaging sequence is added after supine imaging, thus allowing time for cardiac volumetric equilibrium to be achieved. The authors present nothing to exclude this alternative explanation, which is also consistent with their observations. In other words, introducing a 15-min delay in the commencement of imaging after the patient reclines will coincidentally delay the acquisition of the rotational SPECT data to the time when volumetric changes in the left ventricle are minimized.

The authors go on to refer to a previous report of an increase in false-positive perfusion abnormalities detected by dual-head gamma cameras in comparison with single-head systems (their reference 9). The more rapid acquisition of MPI sequences is facilitated by multiple-detector SPECT systems, which serve to compress the total image acquisition interval into an earlier time frame, when poststress volume changes are more pronounced.

We routinely perform clinical nonrotational SPECT MPI studies in our laboratory using a multipinhole tomographic technique (6) that allows all images in the MPI dataset to be acquired coincidentally. This methodology allows us to monitor volume changes throughout the 20- to 30-min imaging interval after graded treadmill exercise. We have computed the ungated, average changes in the size of the heart chambers during the first 3 min versus the final 3 min of our acquisitions: percent volume change = (volume_{first 3 min}/ volume_{final 3 min} −1.0) × 100%. This protocol has been applied to 300 consecutive patients undergoing graded treadmill exercise testing and has shown the average ungated volume change in these patients to be +4.0%. Changes greater than 12.0% have been demonstrated in 13.5% of our patients. These findings are consistent with the published data (3–5).

The fact that ultimately emerges from these discussions is the inherent limitation associated with a methodology that uses a moving detector system to image an object that is dynamically changing in size, location, and intensity. It should not be surprising that the artifact issues associated with rotational SPECT systems are sometimes intensified with the desire to gate these images and correct for attenuation and creep, especially if the corrective methodology is incorrectly identified or applied. The simultaneous multipinhole SPECT MPI technique that we prefer (6) acquires all the images in a manner that is less affected by any temporal variables because changes in size and position of the heart affect all views equivalently. This simultaneous approach to SPECT MPI also takes better advantage of the increased sensitivity of multiple-detector systems.