Revisiting Perfusion in the Forgotten Right Ventricle: Artificial Intelligence–Enhanced Quantification on PET/CT

Visual Abstract

Abstract

Increased right ventricular (RV) radiotracer uptake on perfusion imaging has been recognized as a marker of increased cardiovascular risk. However, this uptake is challenging to quantify because of the variable intensity of uptake in a thin structure. We used a validated artificial intelligence–enhanced method for segmenting the right ventricle from CT attenuation correction (CTAC) imaging to automatically quantify RV activity and then evaluated its prognostic significance. Methods: We evaluated consecutive patients from 11 sites who underwent PET myocardial perfusion imaging with available CTAC. We segmented the RV and left ventricular myocardium from CTAC images using deep learning and then quantified RV activity measures on coregistered PET images. We evaluated associations between RV activity measures and the incidence of death or myocardial infarction (MI). Results: In total, 25,444 patients were included in our analysis (median age, 67 y). During a median follow-up of 4.1 y, 6009 patients (23.6%) experienced death or MI. Most RV activity measures were associated with the risk of death or MI. Higher maximum RV rest activity was associated with an increased risk of death or MI (unadjusted hazard ratio, 1.17 per SD for ¹³N-ammonia and 1.19 per SD for ⁸²Rb). These associations persisted after adjusting for age, sex, medical history, perfusion, function, and myocardial flow reserve. Conclusion: Deep learning can extract RV activity from hybrid PET/CT myocardial perfusion imaging. These measures are associated with myocardial flow reserve and provide complementary information regarding cardiovascular risk.

The right ventricle plays an important role in many cardiovascular diseases, remodeling dramatically in response to pressure or volume overload, regardless of the etiology (1). These adaptations could help identify patients with underlying cardiovascular disease when evaluated with nuclear cardiology studies. For example, increased right ventricular (RV) activity with stress has been identified as a marker of obstructive multivessel or left main coronary artery disease (CAD) on SPECT imaging (2). Similarly, RV quantification from PET myocardial perfusion imaging (MPI) can be used to identify patients with a higher probability of having obstructive CAD (3). RV uptake using nonperfusion radiotracers is associated with worse outcomes in both cardiac amyloidosis (4) and cardiac sarcoidosis (5). However, despite its known importance, the right ventricle has been mostly overlooked in nuclear cardiology. In part, this is related to its distinct morphology compared with the left ventricle, including a thin myocardium in the absence of disease. Right ventricle morphology can be difficult to quantify routinely in nuclear cardiology, traditionally relying on tedious manual segmentations that require a high degree of expertise or visual identification of abnormalities, which has inherent issues related to subjectivity.

Hybrid MPI, with PET or SPECT, includes a low-dose CT attenuation correction (CTAC) acquisition. CTAC is not only important for accurate image quantification; it provides potentially valuable anatomic information (6,7). We have previously demonstrated that deep learning–derived chamber volumes and mass from CTAC images can be used to improve prediction for cardiovascular events (8,9). These chamber segmentations can also be used as regions of interest to provide automated image quantification. This approach can be used to quantify radiotracer activity in disease with a variety of radiotracer distribution patterns, including cardiac sarcoidosis (10) and cardiac amyloidosis (11), which can be difficult to quantify with conventional software. Although we have previously performed automated quantification of hot-spot imaging (10,11), the same approach could be applied to quantify MPI radiotracer activity. This could be beneficial in quantifying radiotracer activity in the right ventricle, which is difficult to segment with conventional approaches because of its size and variability in activity (12).

Automated measurements of RV activity could make these measurements feasible for routine clinical practice. Accordingly, we evaluated the prognostic significance of automated RV quantification using a large, multicenter cohort of patients undergoing ⁸²Rb and ¹³N-ammonia PET MPI.

MATERIALS AND METHODS

Patient Population

We included consecutive patients from 11 sites from the REFINE PET who underwent ⁸²Rb and ¹³N-ammonia PET MPI with available CTAC. Details of the patient population are included in the supplemental materials, available at http://jnm.snmjournals.org (13). Clinical information was collected at the time of imaging and included age, sex, body mass index, family history of CAD, smoking status, history of previous myocardial infarction (MI), previous revascularization, hypertension, diabetes, and dyslipidemia. Prior CAD was defined as previous MI or revascularization (14).

PET Image Acquisition

Patients underwent rest and stress PET MPI in accordance with clinical guidelines using ⁸²Rb (n = 16,942) or ¹³N-ammonia (n = 8502) (15). Most patients underwent vasodilator stress with regadenoson (n = 22,496, 88.4%), adenosine (n = 1529, 6.0%), or dipyridamole (n = 1200, 4.7%). The remaining patients (n = 219, 0.9%) underwent dobutamine stress.

Standard Imaging Quantification

Details are available in the supplemental materials (16–18). Coronary artery calcium was quantified from CTAC as previously described (19,20).

RV Quantification

CTAC scans were acquired in accordance with site-specific PET/CT protocols for all cases (21). We used our previously validated deep learning model to segment the RV and left ventricular (LV) myocardium from CTAC images (22). To avoid spillover from LV myocardium, both chambers were estimated at the same time to prevent overlapping segmentations. A rim was created around a shrunk RV cavity segmentation to quantify the RV free wall, and a rim was created around the LV myocardium to avoid spillover activity from the LV myocardium near the interventricular septum and apex. These segmentations yielded total volumes that have excellent correlation with volumes from contrast-enhanced CT (r = 0.922 for RV, r = 0.947 for LV myocardium), and good test–retest repeatability in populations external to the present study (23). The segmentations were then applied to coregistered rest and stress static perfusion images to quantify mean and maximum RV radiotracer activity. Coregistration was evaluated when studies were initially quality controlled in the core laboratory, with no additional evaluation for this study. Origin mapping was used, but no additional rigid or nonrigid registration was applied. Both mean and maximum activity reflect values from the entire 3-dimensional segmentation of the RV free wall. LV myocardium segmentations were applied to quantify mean and maximum radiotracer activity, and values reflected the full 3-dimensional segmentation. We evaluated mean and maximum RV activity at stress and rest as well as the ratio of stress-to-rest RV mean and maximum activity. Activity was evaluated to be consistent with prior work (3). We also calculated the RV uptake ratio as (maximum RV_stress/maximum LV_stress) − (maximum RV_rest/maximum LV_rest) as previously suggested by Abraham et al. (3). Lastly, we evaluated an alternative RV activity ratio, named RV uptake ratio2, which was calculated as (maximum RV_stress/maximum LV_stress)/(maximum RV_rest/maximum LV_rest). A representative example of RV segmentation and quantification is shown in Figure 1.

FIGURE 1.

Examples of RV quantification from ⁸²Rb PET/CT. Case 1 is 66-y-old male with history of hypertension, preserved stress myocardial blood flow (2.79 mL/min/g) and myocardial flow reserve (2.01). Maximum RV activity at rest was elevated (150 kBq/mL), which decreased with stress (121 kBq/mL). Patient died 138 d after scan. Case 2 is 63-y-old male with reduced myocardial flow reserve (1.99) and stress flow (1.92 mL/min/g). Maximum RV activity at rest was normal (87 kBq/mL) and increased appropriately with stress (127 kBq/mL). Patient reported no cardiovascular events at 1120-d follow-up.

Outcomes

The primary outcome was death or MI. Death was ascertained from the National Death Index or local administrative databases and included all-cause and cardiovascular-related mortality. The incidence of MI was determined using electronic medical records, with potential events adjudicated by an experienced physician.

Statistical Analysis

Details are available in the supplemental materials.

RESULTS

Patient Population

In total, 25,444 patients were included in our analysis (median age, 67 y). Of those patients, 8740 (34.3%) had reduced myocardial flow reserve (MFR) (Table 1). RV quantification measures stratified by the presence or absence of reduced MFR are shown in Table 2. Patients with reduced MFR had higher median maximum RV activity at rest compared with patients with preserved MFR (¹³N-ammonia: 50.0 kBq/mL vs. 43.7 kBq/mL, P < 0.001; ⁸²Rb: 104.1 kBq/mL vs. 89.5 kBq/mL, P < 0.001). Maximum RV stress-to-rest ratio was lower in patients with reduced MFR compared with preserved MFR (¹³N-ammonia: 1.318 vs. 1.642, P < 0.001; ⁸²Rb: 1.153 vs. 1.360, P < 0.001). Associations between RV activity measures and reduced MFR are shown in Supplemental Table 1. Most RV activity measures were associated with reduced MFR in adjusted analyses.

TABLE 1.

Baseline Population Characteristics Stratified by MFR^*

Characteristic	All Patients	Preserved MFR	Reduced MFR
n	25,444	18,935	8740
Age (y)	67 (58–75)	66 (57–73)	70 (62–78)
Sex
Male	14,909 (58.6)	9560 (57.2)	5349 (61.2)
Female	10,535 (41.4)	7144 (42.8)	3391 (38.8)
Hypertension	19,991 (78.6)	12,599 (75.4)	7392 (84.6)
Diabetes	9196 (36.1)	5137 (30.8)	4059 (46.4)
Dyslipidemia	18,420 (72.4)	11,820 (70.8)	6600 (75.5)
Family history of CAD	6736 (26.5)	4837 (29.0)	1899 (21.7)
Smoking	4838 (19.0)	3472 (20.8)	1366 (15.6)
Log(CAC+1) value	2.15 (0–2.99)	1.83 (0–2.80)	2.78 (1.83–3.28)
Stress TPD	4.70 (1.78–11.1)	3.64 (1.41–8.08)	8.08 (3.08–19.3)
Rest TPD	0.86 (0.07–3.75)	0.58 (0–2.45)	1.88 (0.24–7.52)
Stress MBF	2.27 (1.68–2.89)	2.51 (1.99–3.08)	1.65 (1.22–2.20)
Rest MBF	0.93 (0.73–1.21)	0.88 (0.70–1.13)	1.09 (0.82–1.37)
Stress LVEF	67.5 (56.0–75.4)	69.9 (60.9–76.8)	61.5 (46.1–71.4)

^*P < 0.001 for all comparisons.

Log(CAC+1) = log scale for single coronary artery calcium measurements; TPD = total perfusion deficit; MBF = myocardial blood flow; LVEF = left ventricular ejection fraction.

Qualitative data are expressed as number and percentage; continuous data are expressed as median and interquartile range.

TABLE 2.

RV Quantification Measurements Stratified by MFR^*

	¹³N-ammonia		82Rb
Parameter	Preserved MFR	Reduced MFR	Preserved MFR	Reduced MFR
n	5532	2970	11,172	5770
Maximum RV activity (kBq/mL)
Stress	79.6 (54.3–115.5)	72.4 (49.4–106.8)	119.1 (63.7–169.4)†	118.6 (71.7–172.6)
Rest	43.7 (27.8–69.2)	50.0 (32.8–79.9)	89.5 (53.5–125.5)	104.1 (67.3–151.1)
Mean RV activity (kBq/mL)
Stress	24.7 (17.6–35.6)	23.8 (16.7–33.6)	44.1 (22.1–62.0)	45.3 (27.6–65.8)
Rest	15.4 (9.6–23.5)	17.4 (11.7. 26.7)	36.4 (21.5–50.7)	41.5 (27.5–59.8)
Maximum RV stress/RV rest	1.642 (1.298–2.513)	1.318 (1.059–1.696)	1.360 (1.205–1.531)	1.153 (1.022–1.299)
Mean RV stress/RV rest	1.493 (1.225–2.299)	1.262 (1.033–1.585)	1.236 (1.125–1.362)	1.113 (1.008–1.233)
RV uptake ratio‡	0.038 (−0.047–0.122)§	0.030 (−0.051–0.109)	0.079 (0.003–0.168)	0.052 (−0.018–0.129)
RV uptake ratio2‖	1.050 (0.939–1.171)¶	1.038 (0.935–1.153)	1.104 (1.004–1.239)	1.064 (0.977–1.177)

^*P < 0.001 unless otherwise noted.

^†P = 0.10.

^‡Calculated as (RV stress/LV stress) − (RV rest/LV rest).

^§P = 0.018.

^‖Calculated as (RV stress/LV stress)/(RV rest/LV rest).

^¶P = 0.012.

Qualitative data are expressed as number and percentage; continuous data are expressed as median and interquartile range.

Correlations between RV activity measures and PET measurements are shown in Figure 2. Only weak correlations were found between RV activity measurements and conventional PET imaging markers.

FIGURE 2.

Correlations between imaging findings. CAC = coronary artery calcium; EF = ejection fraction; max = maximum; TPD = total perfusion deficit.

Clinical Outcomes

During a median follow-up of 4.1 y (interquartile range, 2.3–5.1 y), 6009 (23.6%) patients experienced death or MI. The first event was MI in 1415 patients. As shown in Figure 3, patients with increased maximum RV activity at rest were more likely to experience death or MI (unadjusted hazard ratio [HR], 1.42; 95% CI, 1.35–1.50; P < 0.001). Patients could be further risk stratified by considering both MFR and maximum stress/rest RV uptake (Fig. 4). Patients with reduced MFR and a low RV activity ratio had the highest risk (HR, 3.35; 95% CI, 3.14–3.57; P < 0.001), whereas patients with preserved MFR and a low RV activity ratio were at increased risk compared with patients with preserved MFR and a high RV activity ratio (HR, 1.31; 95% CI, 1.21–1.41; P < 0.001).

FIGURE 3.

Survival free of death or MI stratified by RV maximum activity at rest. Cutoffs for abnormal maximum RV rest activity were derived separately for each radiotracer. Multivariate model included sex, age, past medical history (hypertension, diabetes, dyslipidemia, family history of CAD, smoking, prior CAD), stress and rest total perfusion deficit, LV ejection fraction, coronary artery calcium, and myocardial flow reserve.

FIGURE 4.

Survival free of death or MI stratified as function of MFR and maximum RV activity ratio. Cutoffs for abnormal RV activity ratio were derived separately for each radiotracer.

A summary of the associations between RV activity measures as continuous variables and death or MI is shown in Supplemental Table 2. All RV activity measures were associated with the risk of death or MI. In patients imaged with ¹³N-ammonia, the maximum RV activity at rest had an unadjusted HR of 1.17 per SD increase (95% CI, 1.13–1.22; P < 0.001). In patients imaged with ⁸²Rb, the corresponding HR was 1.19 per SD increase (95% CI 1.16–1.23, P < 0.001). Multivariable models were adjusted for sex, age, past medical history (hypertension, diabetes, dyslipidemia, family history of CAD, smoking, prior CAD), stress and rest total perfusion deficit, LV ejection fraction, coronary artery calcium, and MFR. Results from the multivariable model are outlined in Table 3.

TABLE 3.

Associations Between RV Activity Measures and Death or MI by Radiotracer

Variable	¹³N-ammonia*		82Rb†
	Adjusted HR	P	Adjusted HR	P
Maximum RV activity
Stress	1.05 (1.00–1.10)	0.049	1.05 (1.02–1.08)	0.001
Rest	1.07 (1.02–1.12)	0.002	1.08 (1.05–1.11)	<0.001
Mean RV activity
Stress	1.07 (1.02–1.12)	0.005	1.06 (1.03–1.09)	<0.001
Rest	1.09 (1.04–1.14)	<0.001	1.07 (1.04–1.10)	<0.001
Maximum RV stress / RV rest	0.92 (0.86–0.98)	0.007	0.86 (0.82–0.90)	<0.001
Mean RV stress / RV rest	0.93 (0.87–0.98)	0.013	0.94 (0.90–0.98)	0.003
RV uptake ratio‡	0.95 (0.91–1.00)	0.033	0.92 (0.89–0.95)	<0.001
RV uptake ratio2§	0.96 (0.92–1.01)	0.083	0.90 (0.87–0.94)	<0.001

^*n = 8502.

^†n = 16,942.

^‡Calculated as (RV stress/LV stress) − (RV rest/LV rest).

^§Calculated as (RV stress/LV stress)/(RV rest/LV rest).

All HRs are expressed as per SD increase. Multivariable models were adjusted for sex, age, past medical history (hypertension, diabetes, dyslipidemia, family history of CAD, smoking, prior CAD), total perfusion deficit at stress and rest, LV ejection fraction, coronary artery calcium, and MFR.

Risk Reclassification

Risk reclassification after the addition of RV activity measures is shown in Figure 5. Mean RV activity at rest led to the greatest improvement in risk prediction (continuous net reclassification index [NRI] 0.415; 95% CI, 0.384–0.445), followed by mean RV activity at stress (continuous NRI, 0.347; 95% CI, 0.316–0.377). In comparison, the continuous NRI for LV end-diastolic volume was 0.187 (95% CI, 0.156–0.217). Receiver-operating-characteristic curves are shown in Supplemental Figure 1.

FIGURE 5.

Continuous NRI for addition of RV quantification measures to multivariate model. Baseline multivariate model included sex, age, past medical history (hypertension, diabetes, dyslipidemia, family history of CAD, smoking, prior CAD), stress and rest total perfusion deficit, LV ejection fraction, coronary artery calcium, myocardial flow reserve, and radiotracer.

DISCUSSION

In a large, multicenter cohort, quantitative summary measures of RV activity from hybrid PET/CT MPI were associated with reduced MFR. Furthermore, RV activity measurements were independently associated with an increased risk of death or MI. Importantly, these associations persisted in the subset of patients with preserved and reduced MFR. Lastly, RV activity measures improved patient risk classification when added to age, sex, medical history, perfusion, function, coronary artery calcium, and MFR. This approach could be integrated within existing workflows to provide physicians with additional physiologic insights for PET MPI and could be translated to other nuclear perfusion imaging.

In our study, several RV parameters were associated with the presence of abnormal MFR. In particular, stress activity tended to be lower in patients with reduced MFR, and rest activity was higher. The lower activity at stress may reflect an inability to augment flows attributable to the presence of epicardial CAD or microvascular disease. Consistent with this, previous studies have shown that reversibility of RV activity on SPECT MPI can be used to detect patients with RV ischemia (24). Furthermore, Abraham et al. demonstrated that patients with obstructive CAD had less of an increase in RV activity from rest to stress compared with patients without obstructive CAD (3). Their work relied on visual evaluation of RV activity and manual placement of regions of interest. In contrast, our approach, with automated quantitation, is more feasible for most clinical workflows. We expanded on these prior studies by demonstrating that the associations with abnormal MFR were independent of age, sex, medical history, and other perfusion findings. This suggests that RV quantitation could be helpful in estimating myocardial blood flow measurements from exercise PET scans (25), where dynamic acquisitions cannot be acquired, or for SPECT MPI.

Another significant finding in the present study was that RV quantification was associated with the risk of death or MI. Although the nature of these associations may differ among patients, a few previous studies provide some pathophysiologic insights. The presence of increased RV activity at rest, assessed qualitatively, is associated with RV systolic pressure estimates from echocardiography (26). Jose et al. similarly demonstrated that incidental RV activity identified patients with pulmonary hypertension and an increased risk of all-cause mortality (27). Their study also relied on visual identification of RV hypertrophy or dilation and did not differentiate between rest and stress values. These studies suggest that RV hypertrophy from pulmonary hypertension may be a key underlying factor. However, it is also worth noting that reduced MFR is associated with diastolic dysfunction even in the absence of regional perfusion abnormalities (28). Furthermore, left-sided heart disease remains the most common cause of pulmonary hypertension (29). As such, the association between RV quantitation and death or MI may be partly mediated through diastolic dysfunction, given the association with reduced MFR (as outlined above) and our finding that the risk persisted after adjustment for MFR.

Importantly, we demonstrated the prognostic utility of these measures using a large, multicenter dataset incorporating the two most commonly used PET perfusion radiotracers. In the overall population, we found that mean RV activity at rest provided the highest improvement in NRI, followed by mean RV activity at stress. NRI was evaluated using patients held out from model training, suggesting that the results would be generalizable to new populations. In comparison, the NRI for mean RV uptake at stress and rest was significantly higher compared with LV end-diastolic volume, an established risk marker (30). Those values were also higher compared with reclassification considering all cardiac chambers and mass (NRI 0.231) (8), but lower compared with deep learning coronary artery calcium (NRI 0.494) (20). In contrast, improvement in the area under the receiver-operating-characteristic curve was modest compared with the full baseline multivariate model. In unadjusted analyses, RV uptake at stress was associated with death or MI in patients imaged with ⁸²Rb but not ¹³N-ammonia. However, after multivariable adjustment, RV uptake at stress was associated with death or MI in both populations. These results suggest that the association with RV uptake at stress may be confounded by other factors, such as LV perfusion or MFR. Additionally, the longer positron range of ⁸²Rb may lead to increased LV myocardial spillover to RV segmentations, making it a more sensitive measure of CAD. However, in patients undergoing PET with ⁸²Rb, there was a stronger association with maximum stress-to-rest ratio compared with ¹³N-ammonia (adjusted HR, 0.86 vs. 0.92 per SD), suggesting that this may be a more useful measure in those patients. In contrast, more complex activity ratios were associated with risk of death or MI but did not improve overall risk classification. The automated approach to measuring RV activity could help bring the importance of its evaluation to clinical practice. Although previous studies demonstrated potential benefits from RV quantification, many centers still rely on visual identification of abnormal uptake. Automated quantification would remove subjectivity, improve reproducibility, and ensure that the RV is evaluated on all studies—so that the RV is no longer the forgotten ventricle in nuclear cardiology.

Our study had a few important limitations. Most importantly, we do not have information on the presence of pulmonary disease, including pulmonary hypertension, in these patients. However, on the basis of prior studies we would expect that at least a portion of the associated risk was driven by these disorders. Second, given the large scale of our analysis, we did not verify the contours that were derived automatically to facilitate RV quantification. Furthermore, our previous validation of RV and LV contours was limited to the comparison of volumes derived by expert segmentation rather than pixel-by-pixel comparisons. The contours for RV and LV myocardium could be verified by technical staff at the time of study processing, which may further improve the clinical utility of this approach. A prospective observational study evaluating the impact of RV quantification on diagnosis and clinical management is warranted.

CONCLUSION

Quantitative summary measures of RV activity from hybrid PET/CT MPI are associated with MFR and provide complementary information regarding cardiovascular risk.

DISCLOSURE

This research was supported in part by grant R35HL161195 from the National Heart, Lung, and Blood Institute/National Institutes of Health and R01EB034586 from the National Institute of Biomedical Imaging and Bioengineering (principal investigator: Piotr Slomka). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Robert Miller reports consulting fees from Alnylam and Bayer and research support from Alberta Innovates. Daniel Berman and Piotr Slomka report software royalties from QPS software at Cedars-Sinai Medical Center. Daniel Berman, Damini Dey, and Piotr Slomka report equity in APQ Health Inc. Daniel Berman reports research grant support from The Dr. Miriam and Sheldon G. Adelson Medical Research Foundation and consulting fees from GE HealthCare. Piotr Slomka reports research grant support from Siemens Healthineers and consulting fees from Synektik SA and Novo Nordisk. Panithaya Chareonthaitawee reports consulting for Clario. Marcelo Di Carli reports consulting fees from MedTrace, Valo Health, GE HealthCare, Bitterroot Bio, and IBA; investigator-initiated research support from Amgen; and institutional research grant support from Sun Pharma, Xylocor, Alnylam, and Intellia. Andrew Einstein reports speaker fees from Ionetix, consulting fees from Artrya and W.L. Gore & Associates, and authorship fees from Wolters Kluwer Health. Andrew Einstein has also served on scientific advisory boards for Canon Medical Systems and Synektik S.A. and received grants paid to Columbia University from Alexion, Attralus, BridgeBio, Canon Medical Systems, Eidos Therapeutics, Intellia Therapeutics, International Atomic Energy Agency, Ionis Pharmaceuticals, National Institutes of Health, Neovasc, Pfizer, Roche Medical Systems, Shockwave Medical, and W.L. Gore & Associates. René Packard serves as a consultant for GE HealthCare. Mouaz Al-Mallah reports research support from Siemens Healthineers and GE HealthCare and is a consultant to Jubilant, Medtrace, GE HealthCare, and Pfizer. Leandro Slipczuk reports grant support/consulting honoraria from Amgen and Philips and served as site principal investigator for V-INITIATE and Ocean(a) trials. Viet Le reports research grant support from Johnson & Johnson/Janssen and honorarium from the American College of Cardiology for the Editor-in-Chief role at Cardiosmart and has served on advisory boards for Amgen, Amarin, Bayer, Boehringer Ingelheim, Esperion, Idorsia, iRhythm, Merck, Novartis, Novo Nordisk, and Pfizer. No other potential conflict of interest relevant to this article was reported.

KEY POINTS

QUESTION: Can quantitative measures of RV activity, provided by deep learning, improve cardiovascular risk assessment?

PERTINENT FINDINGS: We used deep learning to quantify RV activity in 25,444 patients from 11 sites. RV activity measures were associated with the risk of death or MI, with higher RV rest activity associated with an increased risk of death or MI in both unadjusted and adjusted analyses.

IMPLICATIONS FOR PATIENT CARE: RV activity can be quantified on PET MPI to provide insights into cardiovascular risk, potentially improving risk classification. Studies evaluating whether automated RV quantification improves diagnosis or patient management are warranted.