Several studies evaluated PET-based LV volumes and cardiac function with ECG gating and indicator dilution techniques, but, to the best of our knowledge, this is the first study combining the two methods for quantitation of mitral regurgitation severity. The different functional assessments required for regurgitation calculation were possible to perform with a high degree of automation on data from a single PET acquisition, using either a freely diffusible (15O-water) or a metabolically trapped (11C-acetate) PET tracer with similar results. PET measurements of regurgitation magnitude correlated strongly with CMR, and using PET-based RegVol it was possible to discriminate the controls from patients with high accuracy. These results indicate that PET is able to detect and roughly quantify mitral regurgitation.

All gated PET reconstructions used in the current study can be performed automatically on the PET console, if the time window for ECG gating is pre-defined. For 11C-acetate, this is straightforward, as the myocardial uptake is used to delineate the cavity and calculate volumes. We chose 11C-acetate data between 2 and 7 min, as has been described previously [14]. During this time window, the myocardial uptake is expected to be clearly visualized, but a number of different time settings would likely be applicable. 15O-water blood pool gating is more sensitive, since for adequate contrast between myocardial tissue and cavity, the acquisition window has to be short enough for the bolus to remain intact during first pass, which limits the amount of counts available for calculation. The acquisition window might need adjustments for patients with unusually low or high CO [18]. A window of 0–50 s worked for the analysis of all the subjects included in the current study, and CO did not differ between patients and healthy volunteers. If the scans are acquired in list-mode, as in this study, it would be possible to reconstruct the gated series using a different acquisition interval retrospectively.

PET-based LV volumes correlated strongly with CMR, while the correlation for LVEF was only moderate. This was most likely due to the narrow range of LVEF values, and in good agreement with previously reported measurements using ECG gated PET [9]. The accuracy of gating-based volume measurements is dependent on temporal resolution and no published data have been presented where more than 8 gating bins were used for 15O-water PET, likely due to the use of data from older, non-digital PET scanners with relatively low sensitivity and resolution. Increasing the number of bins to 16 improved the agreement of PET-based RegVol and RegF towards CMR. A slightly higher change was seen in ESV as compared with EDV when increasing the number of gating bins. This was expected since the end-systolic phase is shorter, and the ESV is more likely overestimated when using a low temporal resolution.

PET systematically underestimated LV volumes in comparison to CMR for both tracers. However, when performing an additional CMR analysis with removal of papillary muscle volume from EDV, the bias was no longer present between 15O-water and CMR. Also, the underestimation for RegVol and RegF was eliminated. In hindsight, this observation suggests that CMR-based evaluation of LV regurgitation using standard circular chamber delineations overestimates RegVol. The bias for LVSV, RegVol and RegF remained between 11C-acetate and CMR, primarily explained by the larger underestimation of EDV in 11C-acetate. This underestimation for 11C-acetate EDV was similar to prior results [14].

The 15O-water and 11C-acetate arterial and venous based CO correlated strongly, with no bias between tracers or clusters. This was in line with the high accuracy presented in previous studies assessing CO from PET data with older scanner types [12, 13], and suggests that the method is robust and reproducible. With a recorded HR, the FSV is thereafter easily derived from the CO. Using ECG during the scan automated this process for 15O-water PET, since the HR registration corresponded to the first pass during which CO was measured. For 11C-acetate PET, the ECG-derived HR corresponded to 2–7 min into the scan, and thus it is possible that the HR recorded differs from the first pass. This was the case in 2 subjects, where the gating-based and manual HR differed 10 and 13 heart beats. Therefore, the manual recording was used for calculation of 11C-acetate FSV, even though there were no significant mean differences between the two methods on the group level.

As in all cases of indicator dilution techniques, correct PET-based CO calculation requires the knowledge of exact amount of injected radioactivity. Since 15O-water has a half-life of 2.03 min, precise measurements are challenging. The scans conducted in the current study utilized a standardized semi-automatic tracer injection. Radioactivity was measured in a syringe in a dose calibrator 2 min before injection, aiming for 830 MBq to account for isotope half-life and residual activity in the syringe and tubing, and resulting in an injected dose of approximately 400 MBq. It is difficult to obtain the precise amount for every injection and the residue depends somewhat on the concentration of the tracer produced, and the manual handling of the syringe. Using more automated boluses would likely ensure an even more robust CO calculation for 15O-water. With 11C-acetate, the residual activity in the syringe and tubing could be accounted for by direct measurements after injection. However, the actual amount of radioactivity reaching the main circulation during the first pass is still somewhat uncertain as PET images frequently show remaining activity at the injection site and brachial veins, even when flushing with 30 mL saline. This may contribute to the overestimation of FSV as shown in Figure S3.

In line with the findings in this study, earlier experiments have shown PET-based FSV utilizing 5 s frames during first pass to be overestimated in comparison to CMR [13]. The overestimation was scanner dependent, but no evaluation has previously been performed for the type of scanner used in this study. Therefore, we performed the scanner specific calibration of the FSV towards CMR using linear regression. It is a methodological limitation that cardiac output and FSV calculations currently require scanner-dependent corrections. The calibration should preferably be based on invasive thermodilution or local CMR devices and analysis tools.

In order to utilize either 15O-water or 11C-acetate, an on-site cyclotron is required, which is a barrier for widespread clinical implementation. However, as shown here with 15O-water PET, a technique based on first-pass analysis might allow most PET tracers to be used, provided that dynamic scanning is acquired, reliable estimates of injected dose are available, and the radiopharmaceutical is delivered as a fast and standardized bolus. Our results indicate that PET could be used for simultaneous evaluation of primary or secondary mitral regurgitation, LV dilatation and PET-specific parameters such as metabolism or ischemia. This might lead to a more comprehensive diagnosis from cardiac PET scans, while possibly speeding up the patient management. Considering the increasing burden of valvular diseases, allowing for calculation of mitral regurgitation using PET could be valuable. PET-based measurements of regurgitation could be particularly useful to discover secondary mitral regurgitation in perfusion imaging on rare occasions when echocardiography has not been performed prior to the PET-examination.

Some limitations of the present study should be noted. The accuracy of quantifying low levels of RegVol and RegF was not assessed as the healthy volunteers were not examined with CMR, and few patients with mild and moderate mitral regurgitation were included. Only five of the healthy volunteers performed echocardiography, and thus for 13 of the controls included it is unknown whether any of the volunteers had symptom-free, undiagnosed valve disease that could have affected the results. The 15O-water based mean RegVol and RegF was relatively high and likely overestimated in the healthy volunteers (21 mL and 22% respectively). One methodological explanation is that identical settings were used in the gating analysis for patients and controls, although LV volumes are expected to be significantly higher in patients. A further explanation is that the FSV calibration was only performed on data from patients, as only they underwent both CMR and PET. These findings indicate that further development of the methods is needed, incorporating larger cohorts with variation in cardiac function and size. However, the PET measurements were able to separate with confidence patients with substantial mitral regurgitation from normal controls.

An inherent limitation of the PET method is the current inability to distinguish between aortic and mitral regurgitation, since only the total amount of blood not moving forward in the system is taken into account when calculating RegVol and RegF. This suggests that elevated levels of left ventricular regurgitation found with PET should be complemented with further cardiac imaging.