Introduction
Myocardial perfusion imaging (MPI) is known as one of the best validated noninvasive methods to test for ischemia,
1 although artefacts negatively influence the clinical accuracy. Introduction of patient-specific activities and CT-based attenuation correction have limited this influence.
2‐
4 However, artefacts still occur and may mainly be the result of motion.
5,
6 Artefacts resulting from patient movement, respiration, and myocardial contraction are difficult to distinguish from real perfusion defects and can lead to false positive studies.
6‐
10
Several types of motion tracking and techniques to correct for overall patient motion (PM) on conventional SPECT cameras have been introduced and validated.
9,
11,
12 However, these techniques cannot be applied in SPECT cameras equipped with stationary cadmium zinc telluride (CZT) detectors with pinhole collimators. Two recent studies showed that motion detection and correction seems feasible on a CZT-based SPECT camera.
8,
13 However, they did not compare their motion correction to a reference standard. It is therefore unknown whether the corrections improved the diagnostic outcomes.
A commercially available automatic motion detection and correction software has become available which can detect and correct for both respiratory motion (RM) and PM in all directions using a CZT-based SPECT camera. This program (MCD for Alcyone, GE Healthcare) has—to the best of our knowledge—not been described or validated in clinical practice before. Hence, the aim of our study was to quantify the motion and to determine the value of automatic correction using commercially available software.
Discussion
In this study, we have shown that patients undergoing an 8-minute CZT-SPECT scan barely move, as both respiratory motion and patient motion were limited. Nevertheless, applying automatic motion correction changed the SPECT interpretation in 11% of the stress scans. However, these changes were both deteriorations and improvements which led us to conclude that motion correction did not seem to improve the overall diagnostic outcomes of CZT-SPECT. Moreover, motion correction also changed semiquantitative outcome parameters, such as TPD and segmental uptake values, but neither these changes could be considered an overall improvement, as compared to the FFR measurements.
Both the RM and PM measured in this study were smaller than reported in previous studies using the same CZT-based SPECT camera.
8,
13 Ko et al reported a mean RM of 10.5 mm in the cranial-caudal direction using a pharmaceutical stress agent, which is much larger than the mean RM of 2.5 mm as we encountered in the present study.
8 The smaller motion in our study might be due to the following reasons: the use of longer timing bins in our study (1.0 second instead of 0.5 second); higher myocardial count rates due to the use of 99m-Technicium instead of Thallium-201, which decreases noise and increases the count statistics and resolution; and the exclusion of reproducibility errors in the masking and manual alignment prior to reconstruction. Although they performed a phantom study demonstrating the correctness of their motion tracking program, they filled their cardiac phantom with a fifth of the activity administered to their patients. Hence, they had far more count statistics during their phantom study than encountered in their patient studies. This might indicate that they detected more noise during their patient studies, possibly explaining the larger detected RM. They reported, based on their phantom study, that a RM larger than 15 mm could cause visual and quantitative image deterioration. However, this RM threshold was never reached in our study. Nevertheless, we applied RM correction in all patients, resulting in a changed SPECT interpretation in 11% of our scans. As we did not find a correlation between RM correction and changes in SPECT interpretation, TPD or segmental values, the changes we observed were possibly due to overcorrections by the automatic software because of inadequate count statistics. This is in agreement with a phantom study we performed in which we validated the motion detection software (this phantom study is described in more detail in the supplementary materials). A manually induced motion was detected within an uncertainty of typically 2 mm using 1-second time bins (corresponding to RM) and 1 mm using 20-second time bins (corresponding to PM). The maximum detection error was in the order of 8 mm for RM. Correction of the limited motion, within the margin of error, is therefore not expected to result in an improvement in the scan in relation to the diagnostic outcome. It may even lead to a deterioration of image quality. Hence, due to the limited detected motion, it is likely that RM correction is not useful in our patient population and application only resulted in overcorrections.
In contrast to the limited effect of correcting for respiratory motion in our study, a recent study by Clerc et al suggested that deleting respiratory motion by deep inspiration breath holding during MPI CZT-SPECT acquisition was beneficial.
25 Breath holding resulted in 12.5% more normal scans in their 40 patients when also using attenuation correction. Acquisition during breath holding causes a caudal shift of the abdominal structures which may prevent inferior wall artefacts and improve co-registration with the inspiration breath-hold CT used for attenuation correction. However, the reported results have not been compared with a reference standard. In addition, starting and stopping the acquisition after each breath hold should be perfectly timed by the operators and be available on the SPECT system. Repeated long breath holding can also be quite difficult for patients and it may even require the administration of higher tracer activities in order to limit the length of total acquisition, which in turn raises radiation dose again.
26,
27
The detected PM in the present study was also lower than the PM reported by Redgate et al, who even used a shorter acquisition time of 6 minutes.
13 Using data of 40 patients, they recently reported a mean PM of 0-4, 4-8, and ≥8 mm in 62%, 35%, and 3%, respectively. However, none of the 83 patients in our study had a mean PM larger than 4 mm. We only encountered a maximum PM of ≥8 mm in 2% (2) of the patients, in contrast to the 10% they reported. The lower PM in our study could be due to exclusion of reproducibility errors in the present study and the comfortable patient environment in combination with extensive patient preparation to calm the patients. Redgate et al concluded from their phantom study that PM should be corrected when it exceeds 10 mm for more than 60 seconds. The maximum PM encountered in this study was 8.9 mm in one time bin of 20 seconds and PM correction did not result in changed SPECT interpretation. Hence, these figures seem to confirm our finding that, similar to RM, PM correction is not necessary in our patient population.
Several assumptions underpinned this study. First, we used a retrospective cohort of patients; all referred for elective FFR measurements after invasive coronary angiography. The incidence of ischemia and irreversible defects was 4% and 11% higher in the present study, respectively, in comparison to what we commonly encounter in our population eligible for MPI CZT-SPECT.
28 Although the incidence of perfusion defects was not expected to influence RM or PM, it induces lower count statistics, possibly resulting in a higher tracking—and therefore also correction—error.
Second, we used FFR as a reference standard to assess the added value of motion correction on the diagnostic outcome. Although the accuracy of FFR is limited in patients with collateral circulation or serial stenosis,
29 it is nowadays considered as one of the most accurate tests to detect ischemia. We only compared the motion-corrected stress acquisitions with FFR, eliminating the possibility to distinguish reversible (ischemic) from irreversible defects as would have been possible when using both stress and rest acquisitions. Although occluded vessels were also interpreted as positive FFR, this might have led to a slight underestimation of the correspondence with FFR for negative motion corrections (normal scans corrected to equivocal or abnormal, or an increasing TPD or decreasing segmental values) and overestimation of positive motion corrections. Moreover, co-registration of MPI with coronary CT angiography was not performed. Therefore some of the changes in perfusion after RM correction may have occurred in a different coronary territory than the territory supplied by the vessel in which FFR measurement was done.
30 Nevertheless, the discrepancies between FFR and MPI were considered to be too limited to influence the outcomes of this study.
Third, we only compared the motion-corrected scans with nonattenuation corrected scans. It was not possible to apply motion correction to attenuation corrected scans and as attenuation correction is not expected to compensate for motion, it would not have contributed to our aim.
Fourth, the percentage of patients in which motion correction changed the diagnostic outcomes differed between the three endpoints: SPECT interpretation, TPD, and segmental uptake values. Although the influence of motion correction seemed limited for all three endpoints, one should be cautious in future studies when using only one of these semiquantitative endpoints. It appears that a segmental uptake difference of 5% is a very sensitive parameter for a change in defects but not a very specific one in comparison to the SPECT interpretation, which is still the reference standard in most institutions. Moreover, using a TPD difference of 7% appears not sensitive enough in detecting change in perfusion deficits in comparison to the visual SPECT interpretation.
Finally, there is a current trend towards lower activities and patient-specific dose protocols.
31,
32 Lowering the activity can easily be achieved by enlarging the scan time, as both are interchangeable within a certain range. In this study, we showed that a scan time up to 8 minutes does not introduce significant motion. The motion even decreased in the first minutes. Using longer scan times in combination with lower activities might therefore even decrease the influence of the higher motion in the first minutes of the acquisition. However, one should be aware that correction of observed motion will be harder with lower count statistics and that acquisitions should be repeated instead of corrected. The amount of motion depends on the calmness and relaxedness of the patients. We think it is of great importance to create a comfortable patient environment, provide clear instructions and to provide extensive patient information prior to MPI to reduce motion.