To our knowledge, this study was the first comparative study of PMOD and HeartSee for relative uptake and absolute quantitation. We found that the rest relative uptake and stress relative uptake of HeartSee was higher than that of PMOD. Besides, the rest MBF of PMOD was higher than that of HeartSee, but there was no statistically signifcant difference between stress MBF, and the MFR of HeartSee was higher than that of PMOD. The correlation between PMOD and HeartSee for absolute quantitative MBF and MFR was good, but there was no correlation in relative uptake. Besides, we found that values of HeartSee global MFR revealed an accuracy of 72% by a cutoff value of 2.885 to assess an abnormal MFR documented by PMOD.
Relative uptake could be obtained by SPECT and PET measurements, that is, assessment of the percentage of local myocardial perfusion uptake relative to the region of highest perfusion tracer uptake [1, 2]. In our study, we found significant differences in rest and stress relative uptake between PMOD and HeartSee. We considered that this may be due to the difference in the method of finding and segmenting the myocardium in the two software programs and the difference in the placement of the ROI, causing differences in the calculation input and temporal activity curves [23, 24].
Previous studies showed that absolute quantification of MBF from dynamic PET images was meaningful for clinicians because it could provide relevant complementary information for relative MPI [25]. However, there are many software packages and models for quantifying MBF and MFR, and two of the most common are PMOD's 1TCM and HeartSee's simple retention model [26]. Comparing the data derived from PMOD and HeartSee, we found a statistical difference of > 10% for the rest MBF and MFR. Therefore, variability due to methodology could be ruled out and the difference was most likely caused by software differences [27, 28]. Similarly, Renaud et al. [29] studied 14 normal volunteers and found that the simple retention model had a significantly lower rest MBF than the compartment model. The reasons for software differences may be as follows: first, the kinetic models of the two software programs were different. The 1TCM of PMOD depended on tracer uptake (K1) and tracer washout (K2) parameters; it was based on plasma activity and total uptake in the myocardium. In this model K1 was linked to MBF through Renkin-Crone relation that corrected for MBF bias and reduction to tracer extraction in different patients during the stressed state [26]. However, HeartSee used a simple retention model, which was easy to use and had a very short usage time [30, 31]. However, it needs to be combined with an extraction correction in the estimation of the MBF. The accuracy of results was also highly dependent on the time taken to evaluate the model and the accuracy of the assumed partial volume and spillover correction factors [32]. And Lortie et al. [32] concluded that a smaller correction factor was required to obtain the MBF with the 1TCM compared to the simple retention model. For 1TCM, DeGrado et al. recommend to only use the first four minutes of data after injection of the tracer to reduce the effects of metabolite buildup and washout [17]. In contrast, the simple retention model suggested that 13N-ammonia was taken up, washed out, and metabolized in cardiomyocytes more than 10 minutes. Additionally, orientation of the adjustment, outlining of the left ventricle contour, and sampling of the left ventricle could also cause these differences. However, our study found no significant difference in stress MBF values, probably because when the stress PET-MPI was induced by adenosine, stress MBF usually increased 4–5 times that of the original, so the difference between stress MBF might be neglected [33]. As MFR is defined as the ratio between the stress and rest MBF, error propagation might cause variance in the MFR measurements. Thus, there was a statistical difference in the MFR.
In this study, we found that the relative uptake results were not consistent with the absolute quantification results. This was because relative uptake used the point with the highest uptake as the reference point, and the uptake values of other regions were calculated relative to this highest point [1]. However, in patients with multiple vascular lesions or CMD, this highest uptake point might be the lesion location, resulting in inaccurate lesion diagnosis results [10]. In contrast, absolute quantification was applied to the 1TCM model or simple retention model algorithm for TAC to obtain the uptake parameters, and the MBF in each region could be quantified [26]. Therefore, the results of relative uptake and absolute quantification were inconsistent.
For all patients in our study, the MBF and MFR values of PMOD and HeartSee correlated to a certain degree. Likewise, Chang et al. [15] found that the simple retention model highly correlated well with the compartment model for 14 healthy volunteers. There was a higher correlation for MFR than for MBF. We thought the reason may be that MFR is a relative indicator and possibly eliminated the systematic error due to different types of software. Moreover, for relative uptake, there was no correlation between PMOD and HeartSee software. This was understandable, because relative uptake has disadvantages for absolute quantification [1]. Comparing relative intake with absolute dosing might underestimate the severity of the disease, especially in multi-vessel coronary artery disease and diffuse microvascular lesions [10].
Finally, PMOD has accurate thresholds and has been used in many studies, so we used PMOD as a criterion to obtain HeartSee's MFR threshold by ROC curve with Youden’s index, and the accuracy was 72%. It indicated a high diagnostic agreement with PMOD at a HeartSee MFR threshold of 2.885; however, our study only included patients with non-obstructive CAD, and further studies are needed to determine whether the obtained thresholds can be used for later diagnosis of obstructive CAD. Therefore, it can be recommended that adequate stress MBF and MFR thresholds be established for each post-processing software to diagnose ischemia in patients referred for suspected impaired myocardial perfusion, preferably with a common database [34].
Limitations
Our study had some limitations. First, the gold standard microsphere method for MBF quantification was unavailable in humans. Hence, we were unable to conclude with certainty which software system was more accurate than the other in quantifying MBF. Second, this was a retrospective study with a small sample size and a single disease type, which may lead to bias. Future studies with larger sample sizes and various diseases are needed to validate our results. Third, we did not perform follow-up for these patients; therefore, we could not provide information for prognosis evaluation and risk stratification.