Introduction

Cardiac masses are rare entities with high rates of morbidity and mortality. There are three basic types: tumour, thrombus, and vegetation. Primary tumours account for an extremely small proportion, with an autopsy frequency between 0.001 and 0.3% [1]. More than three-quarters of primary neoplasms are benign and almost half of the benign neoplasms are myxomas. About 20% of primary cardiac tumours are malignant, and 95% of these are sarcomas [2]. Metastatic disease, which typically originates from a primary pulmonary tumour [3], is much more common than primary cardiac tumours.

The therapeutic strategies and prognosis of cardiac masses not only depend on the nature of the mass (neoplastic or non-neoplastic), but, in the case of neoplasm, also on the tumour characteristics (benign or malignant, primary or secondary) [4]. Because of the location, even benign cardiac masses could lead to serious consequences, such as heart failure caused by heart cavity obstruction, pulmonary/systemic embolism, and even sudden death [4, 5]. With the development of modern medical technology, the resection rate of cardiac masses has been constantly improving [4]. Adjunctive therapy can significantly reduce any impairment of cardiac function [6]. In malignancies, postoperative radiotherapy and chemotherapy can significantly improve the short-term prognosis, but the long-term prognosis remains poor [6, 7]. Correct identification of the nature of a cardiac mass by non-invasive imaging is critical for diagnosis, management strategy, and prognosis.

Anatomical imaging modalities, including transthoracic and transoesophageal echocardiography, computed tomography (CT), and magnetic resonance (MR), are valuable to provide location and structural information. However, their ability to provide metabolic information is limited [8,9,10]. 18F-Fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG PET/CT) can detect both anatomical and metabolic information, which is applied in oncologic imaging for diagnosis, staging, and prognosis [11,12,13,14,15]. However, due to the rarity of cardiac masses, 18F-FDG PET/CT does not have an established role in their routine evaluation. In the present study, we retrospectively analysed the 18F-FDG PET/CT images of patients with cardiac masses and sought to evaluate whether this imaging modality could provide useful information on diagnosis and prognosis.

Patients and methods

Patients

The study was approved by the Institutional Review Board. We retrospectively reviewed imaging data of all the patients with cardiac masses who underwent 18F-FDG PET/CT in the PET Center of Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, from August 2012 to August 2018. The inclusion criteria were as follows: (1) age ≥ 18 years and (2) complete clinical and imaging data. The exclusion criterion was diabetic patients or patients with fasting blood glucose ≥ 200 mg/dL. All of the cases were divided into malignant and benign groups according to the histological or long-term follow-up results.

18F-FDG PET/CT protocol

18F-FDG was synthesized with 18F produced by a cyclotron (MINItrace®, GE Healthcare, Milwaukee, WI, USA), with radiochemical purity > 95%. All patients fasted for at least 8 h before the administration of 18F-FDG. A total of 3.70–5.55 MBq (0.10–0.15 mCi)/kg 18F-FDG was administered intravenously. PET/CT was performed approximately 60 min after 18F-FDG administration by a PET/CT scanner (Discovery VCT®, GE Healthcare). Both CT and PET were acquired from the top of the head to the upper thighs. CT scanning was performed after a CT scout view with a tube voltage of 120 kV, a tube current of 110 mAs, and scanning thickness of 3.75 mm. PET was acquired with 3 min per bed position.

PET/CT image interpretation

PET/CT images were visually interpreted by two experienced nuclear medicine physicians in consensus, with knowledge of the initial clinical data but blinded to the histology. The physicians reviewed the CT and PET images and recorded their findings, respectively. The CT diagnosis was mainly according to the location and morphological characteristics of the masses (margin, with/without necrosis), in addition to its relationship with surrounding tissues (with/without infiltration/involvement of epicardium, surrounding tissue, and vessel), with/without pericardial/pleural effusion, and whether there were extracardiac lesions. A semi-quantitative method was applied in the PET interpretation, which was based on several metabolic indices of 18F-FDG uptake (maximum standardized uptake value [SUVmax], mean standardized uptake value [SUVmean], metabolic tumour volume [MTV], and total lesion glycolysis [TLG]). All PET and CT data were processed using commercial software (PMOD PNEURO version 3.906, PMOD Technologies Ltd., Zurich, Switzerland) in DICOM format. The lesions were identified on PET images and segmented automatically using a 3D-area growing algorithm in the axial, coronal, and sagittal planes [16]. The measurement of the indices was performed within the regions of interest (ROIs), which were determined by the SUV threshold or from CT image. The SUV threshold was calculated by a background method as described previously [17] as follows:

$$ \mathrm{SUVthreshold}=S\mathrm{UVbkgd}+20\%\left(\mathrm{SUVmaxROI}-\mathrm{SUVbkgd}\right), $$
(1)

where SUVmaxROI refers to SUVmax of the lesion and SUVbkgd refers to the mean SUVmax of the background. Due to the variability of FDG uptake in the myocardium, we chose the erector spinae muscles as the background. The MTV and SUVmean were determined automatically by the software, and the total lesion glycolysis (TLG) was calculated according to the formula:

$$ \mathrm{TLG}=\mathrm{SUVmean}\times \mathrm{MTV} $$
(2)

Follow-up

Follow-up was performed after PET/CT scan. During the follow-up time, we recorded the therapeutic response of each patient, the time to first progression/recurrence or death, and the performance status of survivors. OS was defined as the time interval from the date of PET/CT imaging to death related to cardiac mass or the date of last follow-up and was chosen as an endpoint to estimate the prognostic value of clinical data and PET/CT parameters.

Statistical analysis

A commercial software package (SPSS 22.0, IBM Inc., Armonk, NY, USA) was employed for data processing. Continuous variables are expressed as mean ± SD. Categorical variables are expressed as number and percentage. For quantitative data, normal distribution and homogeneity of variances were tested first. Student’s t test was used for quantitative data with normal distribution and equal variances, while Welch’s t test was applied if variance was unequal. Quantitative data with non-normal distribution or the categorical variables were compared with the Mann-Whitney U test between two groups (benign and malignant groups), or with the Kruskal-Wallis H test among three groups (primary, secondary, and lymphoma groups). Receiver operating characteristic (ROC) analysis was used to determine the optimal cutoff values of parameters (SUVmax, SUVmean, MTV, and TLG), and the value with the highest sum of sensitivity and specificity was regarded as the cutoff value. Sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV) were calculated for each variable. The influences of gender, age, morphological features, and metabolic indices were examined in univariate analysis and multivariate Cox regression analysis for overall survival (OS). Kaplan-Meier’s method was applied to generate survival curves, and the log-rank test was used for comparison.

Results

The final study population comprised 64 patients (34 men, 30 women; mean age ± SD, 51.20 ± 17.50 years), with 65 cardiac masses (one man with two different benign cardiac masses [myxoma, ventricular aneurysm]). These cardiac masses were diagnosed as follows: (1) malignancy = 38 (58.50%; primary tumours [N = 17], metastases [N = 15], lymphomas [N = 6]); (2) benign = 27 (41.50%; neoplasms [N = 13], non-neoplastic lesions [N = 12], unknown [N = 2]). The information of pathology type, location, related history, and presence of extracardiac lesions is summarized in Table 1, and four representative cases are shown in Fig. 1.

Table 1 Cardiac lesion pathology, location, related history, and presence or absence of extracardiac lesions
Full size table
Fig. 1
figure 1

Representative cases. a A 54-year-old man presenting with malignant peripheral nerve sheath tumour in the left atrium. 18F-FDG PET/CT images showed the cardiac tumour with an SUVmax of 12.27 without extracardiac involvement (arrows). b A 83-year-old man presenting with left atrial myxoma with calcification. Images showed the benign mass with an SUVmax of 4.51 (arrows). c The 18F-FDG PET/CT images of a 63-year-old woman demonstrated a left atrial hypermetabolic mass with an SUVmax of 15.15 (arrows), which was secondary to a malignant mesenchymal tumour in the right lung (curved arrow). The MIP image also detected lymph nodes (arrowhead) and adrenal (dotted arrows) metastases. d18F-FDG PET/CT images of a 61-year-old man revealed a large cardiac mass with an SUVmax of 25.37 involving the left atrium (yellow arrows), right atrium (red arrows), and right ventricle (not shown). Bilateral adrenal involvement (dotted arrows) was also noted. The histological diagnosis was diffuse large B cell lymphoma

Full size image

PET/CT findings and diagnostic performance

The comparisons of CT features and FDG metabolic indices between benign and malignant groups, as well as among primary and secondary malignancies and lymphoma, are summarized in Table 2. There were statistically significant differences in multiple morphological features and metabolic indices between the benign and the malignant masses. For malignant masses, among patients with primary or secondary malignancies or lymphoma, there were no statistically significant differences in any of the aforementioned morphological features. The SUVmax and SUVmean of lymphoma were significantly higher than those of primary and secondary malignancies (Table 2, Fig. 2a).

Table 2 Comparison of morphological features and metabolic indices among groups and diagnostic performance
Full size table
Fig. 2
figure 2

a SUVmax of different types of cardiac masses. b ROC curve of SUVmax, SUVmean, TLG, MTV, and SUVmax and TLG

Full size image

Table 2 also illustrates the diagnostic performance of distinguishing benign masses from malignancies using the abovementioned CT features and PET metabolic cutoff values. Each CT feature alone showed relatively low diagnostic accuracy, whereas using a criterion containing more than three CT features for the diagnosis improved the diagnostic performance (sensitivity, 78.94%; specificity, 100%; accuracy, 87.69; PPV, 100%; and NPV, 77.14%). The ROC curves (Fig. 2b) show that using SUVmax of 6.75, SUVmean of 3.52, MTV of 61.32, and TLG of 126.07 as cutoff values resulted in the highest diagnostic performance with the areas under the ROC curves (AUC) of 0.9142, 0.8655, 0.8713, and 0.8187, respectively. Multiple logistic regression analysis resulted in

$$ \mathrm{logit}(p)=0.4092\times \mathrm{SUVmax}+0.0055\times \mathrm{TLG}-3.6223 $$
(3)

with a little better AUC than that of SUVmax alone (0.9279 vs. 0.9142) (sensitivity, 94.74%; specificity, 85.19%; accuracy, 90.77%; PPV, 90.00%; and NPV, 92.00%), but with no statistical difference.

Combining CT morphological features and SUVmax as a criterion for distinguishing benign masses from malignancies optimized the diagnostic performance (≥ 3 CT features and SUVmax ≥ 6.75 [sensitivity, 76.32%; specificity, 100.00%; accuracy, 86.15%; PPV, 100.00%; and NPV, 75.00%]; ≥ 3 CT features or SUVmax ≥ 6.75 [sensitivity, 94.74%; specificity, 88.89%; accuracy, 92.31%; PPV, 92.31%; and NPV, 92.31%]). These results were also shown in Table 2.

Follow-up and Survival Analysis.

Except for two patients with cardiac lipoma who did not undergo surgery, the main treatment for benign lesions was surgery. Surgery with/without adjuvant chemotherapy was the mainstay of treatment for patients with primary cardiac malignancy. The patients with secondary malignancy and lymphoma mainly received systemic chemotherapy.

Because 18 patients were lost to follow-up, the survival analysis was performed in 72.31% (47/65) of cardiac masses, with a mean age of 52.89 ± 17.73 years. The follow-up time ranged from 6 to 2289 days (534 ± 563 days). Of the 47 cardiac masses, 25 were malignant and 22 were benign. Of the 25 patients with malignant masses, 18 (72.00%) died of the disease, whereas only 2 (9.09%) patients with benign disease died. The 2-year OS rates of patients in the malignant and benign groups were 21% and 86%, respectively.

The optimal cutoff values were calculated by ROC curves and are summarized in Table 3. On univariate analysis (Table 3), risk of death was increased for patients with surrounding tissue infiltration, epicardial infiltration, necrosis, multiple chambers or vessel involvement, extracardiac lesion, SUVmax ≥ 6.715, SUVmean ≥ 3.16, MTV ≥ 54.54, and TLG ≥ 137.90. Irregular mass margins seemed to increase the risk of death but not to a statistically significantly extent (hazard ratio 4.15, p = 0.056). Gender, age, pericardial effusion, and pleural effusion were not predictive of overall survival. Multivariate Cox regression analysis showed SUVmax ≥ 6.715 was the only significantly independent prognostic factor (p < 0.001).

Table 3 Results of univariate analysis for predicting overall survival
Full size table

Kaplan-Meier curves illustrated that the cases with SUVmax < 6.715 had longer OS than the patients with SUVmax ≥ 6.715, with median survival times of 1460 and 342 days, respectively (p = 0.00016) (Fig. 3a). Seventeen of the 24 patients with SUVmax ≥ 6.715 (70.83%) died, whereas only three of the 23 lesions in 22 patients  with SUVmax < 6.715 died (13.64%), including one patient with myxoma who died during surgery, one patient with thrombus who died 1 year after heart transplantation due to restrictive cardiomyopathy, and another case with ovarian cancer metastatic to the right atrium. The 2-year OS rates were 29.17% (7/24) in the patients with SUVmax ≥ 6.715 and 86.96% (20/23) in the patients with SUVmax < 6.715, respectively.

Fig. 3
figure 3

Kaplan-Meier analysis of the relationship between OS and SUVmax (a), PET/CT findings (b), and mass types (c), respectively

Full size image

Compared with PET (SUVmax ≥ 6.715)- or CT (≥ 3 features)-positive findings alone, the cases with both PET-positive and CT-positive results had a worse prognosis (p = 0.0023) (Fig. 3b). For disease types, statistically significant differences were observed among the subgroups for survival probability, with best survival in the benign group, followed by the lymphoma group. Primary cardiac malignancies had the worst prognosis (p < 0.001) (Fig. 3c).

Discussion

18F-FDG PET/CT has been well documented in the diagnosis, staging, and monitoring of neoplastic and some non-neoplastic diseases [18,19,20]. Given the rarity and complexity of cardiac masses, the values of PET/CT in cardiac masses, especially the prognosis assessment value, remain to be confirmed. Scattered case reports have noted that FDG PET/CT can visualize cardiac masses and assess their glucose metabolism levels [21,22,23,24,25,26,27,28]. A few publications involved cardiac masses cohort showed that PET/CT or PET/MRI with 18F-FDG is promising for differentiating benign from malignant cardiac lesions [29,30,31,32]. However, the enrolled patient numbers were limited, ranging from 20 to 24, and prognostic assessment was not available.

In this study, we enrolled 65 cardiac masses in 64 cases for evaluating not only the diagnostic accuracy but also the prognostic value of FDG PET/CT. Considering that non-neoplastic lesions, including thrombus, vegetation, valvular lesions, and ventricular aneurysm, sometimes mimic tumours [33,34,35], we brought the patients with the abovementioned tumour mimics into this retrospective study. There are considerable differences in treatment strategies not only between malignant and benign entities but also among different types of malignancies. Therefore, in addition to diagnostic and prognostic assessment between malignant and benign entities, distinguishing among primary malignancies, secondary malignancies, and lymphoma was also carried out in our study.

Our results suggest that 18F-FDG PET/CT is a reliable tool in the diagnosis of cardiac masses. From the results of this study, we propose an 18F-FDG PET/CT diagnostic approach for cardiac masses shown in Fig. 4. Some reports have suggested that a location of the tumour outside the left heart, tissue inhomogeneity, the presence of pericardial effusion, infiltrative growth pattern, and lobulated margins are predictors of malignant cardiac masses [8,9,10, 36, 37]. Tumour size is not efficient to differentiate benign masses from malignancies [8, 32]. Following the morphological score system established by Rahbar et al. [32], we enrolled multiple morphological features for analysis (Table 2). Our data showed the diagnostic performance of combining ≥ 3 CT features was relatively superior to those using a single feature.

Fig. 4
figure 4

FDG PET/CT diagnostic approach for cardiac masses. First, using SUVmax ≥ 6.75 or ≥ 3 CT features as a differential criterion to distinguish malignancies from benign masses. Second, lymphoma may be distinguished from other malignancies by a higher FDG uptake. Moreover, further detection of extracardiac lesions can not only stage lymphoma but also identify primary and secondary malignant masses

Full size image

The lesion SUVmax values in this study were significantly higher in malignant cardiac masses than those in benign lesions, which is similar to those in the studies (imaging modality, patient numbers; SUVmax cutoff value; sensitivity, specificity, accuracy, PPV, and NPV for diagnosis) conducted by Shao et al. [29] (PET/CT; N = 23; 3.5–4.0; 100.0%, 90%, 95.7%, 92.9%, and 100.0%), Rahbar et al. [32] (PET/CT, N = 24; 3.5; 100%, 86%, 96%, 100%, and 100%), and Nensa et al. [30] (PET/MR, N = 20; 5.2; 100%, 92%, 95%, 88%, and 100%). The cutoff value in our study (6.75) was higher than those of the abovementioned three publications, whereas the diagnostic performance (92.11%, 88.89%, 90.77%, 92.11%, and 88.89%) using SUVmax ≥ 6.75 seemed to be slightly inferior to those studies. These differences might be due to the larger sample size and intra-individual/individual variability. Notably, we hold an identical view with these reports that despite an SUVmax of 2.5 being the conventional cutoff for differentiating benign and malignant lesions in other solid tumours, it is not suitable for cardiac masses. Additionally, SUVmean, MTV, and TLG also reflect glycometabolism and were used for diagnosis [38,39,40] as we did in our study. However, these values had relatively lower efficiency in diagnostic accuracy compared with SUVmax.

Simultaneous consideration of CT results (≥ 3 features) and SUVmax (≥ 6.75) could further optimize the diagnostic performance (Table 2). If both CT- and SUVmax-positive results were regarded as positive criteria, the specificity and PPV would increase, avoiding false-positive results. In contrast, either CT- or SUVmax-positive results regarded as positive criteria could increase the sensitivity and NPV, reducing false-negative results.

In addition to providing morphological characteristics and metabolic parameters, 18F-FDG PET/CT clearly illustrated the locations of the lesions, which was critical for diagnosis and staging. Our results suggest cardiac masses may occur in any single chamber or multiple chambers, with or without vascular infiltration. The right atrium is the most common location for both malignant and benign cardiac masses. Multi-chambers or vessel involvement often occurred in malignant disease. In benign diseases, only one case of myxoma and one case of intravenous leiomyomatosis occurred in the right atrium and inferior vena cava. Some diseases had specific locations, such as lipomas usually occurring in the myocardial wall, valvular lesions in the atrioventricular junction area, and ventricular aneurysms in the ventricle (more common in left ventricle) [4]. One of the biggest advantages of PET/CT is panoramic imaging, which is especially useful for detecting tumours in patients with metastasis with unknown primary and for staging. In this study, PET/CT detected extracardiac lesions in 30 cases, contributing to correct staging.

Diagnostic CT and MR (especially cardiac MR) provide excellent anatomical visualization [41, 42], and some criteria (surrounding tissue infiltration, epicardial infiltration, necrosis, multiple chambers or vessel involvement, and extracardiac lesions) were associated with poorer OS. However, multivariate Cox regression analysis showed all of the abovementioned morphological criteria were not independent prognostic factors. In addition to morphological criteria, PET/CT offers semi-quantitative metabolic indices to aid prognostic assessment. Due to the panoramic view of PET/CT, the detection of extracardiac lesions allows prognostic evaluation. High glycometabolism levels indicate worse outcome, as is reported in other solid tumours [43, 44]. Our data demonstrated that SUVmax is the only significant independent prognostic factor. The patients with cardiac masses with SUVmax ≥ 6.715 had a significantly worse prognosis.

In this study, patients with benign masses had the best prognosis, because gross total resection could alleviate the heart burden and greatly improve the quality of life, obtaining satisfactory long-term survival [45, 46]. Primary cardiac lymphoma had a relatively favourable prognosis due to most patients having a good response to chemotherapy. Petrich et al. [47] described poor survival with left ventricular involvement by lymphoma. However, no left ventricular involvement was observed in our study, which may be one of the reasons for the relatively good prognosis. Whether primary or secondary, cardiac malignancies have a very dismal prognosis because of their high degree of aggressiveness [46]. Some authors suggest right heart malignancies tend to be bulky, infiltrative, cause late symptoms, and metastasize early, all of which are closely related to poor outcome [48, 49]. The majority of our cases had right cardiac involvement, which could partly explain the poor prognosis. Superior survival was observed in secondary malignancies in comparison with primary ones, which might be due in part to the pathological type of primary tumour and relatively good response to therapy.

One of the main limitations of this study is the background interference coming from physiological uptake of 18F-FDG in myocardium in some patients. A low-carbohydrate/high-fat diet before 18F-FDG administration can suppress the tracer accumulation within the myocardium [50, 51]. However, this diet regimen was not instituted due to the lack of suspicion of a cardiac mass. Prolonged fasting before 18F-FDG PET is another suitable method to depress normal myocardial 18F-FDG uptake [51, 52], which was performed in our study. Other limitations include inexact image fusion due to the respiratory and cardiac motion [53], the relatively short follow-up time, and the heterogeneity of treatment modalities. In addition, although our study enrolled 64 cases, which is a relatively larger cohort compared with previous studies, the sample size remains relatively small due to the extremely low incidence of cardiac masses. Hence, we need a prospective randomized trial involving a larger number of patients with gated 18F-FDG PET/CT to further confirm our results.

Conclusions

18F-FDG PET/CT is of great value for diagnosis and prognostic assessment in patients with cardiac masses before treatment and has optimal detecting ability for extracardiac lesions due to the whole-body images. As a semi-quantitative metabolic criterion, SUVmax ≥ 6.75 displayed significant positive predictive value for cardiac malignancies, and SUVmax ≥ 6.715 predicted poor overall survival.