EAT volume is a well-known biomarker related to cardiovascular risk and disease which can be assessed from volumetric imaging techniques such as computed tomography or MRI without the need for additional scans or contrast agent administration [6].
The average values reported for EAT volumes in our population, with medians ranging from 81 to 93 cm3 among different readers and phases, appear within the normal limits for a population with no overt cardiovascular pathology, according to the cut-off of 125 cm3 proposed by Spearman et al. [14], in spite of our measurements being performed in an obese population. However, we should note that Shmilovic et al. [15] in a healthy population with low cardiovascular risk showed a higher 95th percentile set at 68.1 cm3. This would suggest that patients from our population present with a higher-than-average cardiovascular risk and could be prone to an earlier onset of CAD [16, 17], highlighting the value of EAT volume assessment in an obese population. Therefore, considering the peculiar needs of such patients, namely the necessity of adequately dimensioned MR units, validating the reproducibility of EAT volume assessment on open-bore units could allow a more widespread evaluation of EAT in a population at heightened cardiovascular risk.
Intra-reader reproducibility of EAT volume showed higher precision for the more experienced reader R1, while segmentations repeated by the less experienced reader R2 displayed a high SD implying a wider scattering of volumes around the bias and a minor overall precision. However, the less experienced reader R2 showed a relatively small bias which would most likely not provide clinical impact [18] (the largest being 3.9 cm3 at the diastolic phase). Therefore, it may be hypothesised that while segmentations performed by less experienced readers may not yield substantial differences on an average basis, they might not be entirely suitable for single patient assessment due to a higher intrinsic variability.
Concerning inter-reader reproducibility of EAT volume, the higher bias seen in the diastolic phase (-9.4 cm3) as compared to the systolic phase (-2.3 cm3) may be due to the fact that the segmentation of EAT and pericardium could be easier at systolic phase, as suggested by Malavazos et al. [19], as the contraction of the ventricles leads to EAT appearing thicker and more prominent.
The precision of inter-phase EAT evaluation between systole and diastole appeared to be influenced by reader experience, even though with biases that may suggest a non-substantial clinical impact [18], in spite of a SD of the differences of 12.1 cm3 for R2 which could lead to important variation. The significant difference between systole and diastole among values measured by R1 may be due to the intrinsic difference in the prominence of EAT at a ventricular level, which is highlighted by the higher precision and subsequent minor SD of the differences by R1. Such a finding may lead to questioning with regard to the best phase for EAT quantification at cardiac imaging. Nevertheless, previous works have shown that both systolic and diastolic EAT increase with the progression of CAD; therefore, both could be used as biomarkers of pathologic myocardial involvement [20]. Moreover, one additional study by Kang et al. [21] observed a relation between systolic EAT thickness and diabetes, but not with diastolic EAT thickness. This might play in favour of measuring EAT in the systolic phase, due to higher reproducibility and probable clinical impact.
To the best of our knowledge, there is no widespread evidence on the use of open-bore MRI for the assessment of EAT volumes, and thus no available data concerning its reproducibility. The reproducibility of EAT volumes was assessed by previous studies on acquisitions performed on closed-bore MRI systems. For instance, a work by Bettencourt et al. [18] analysing EAT volume reproducibility on 53 patients obtained a bias between 0 and 10 cm3, with a SD of the differences which neared 10 cm3. Another work by Flüchter et al. [22] performed on 43 patients with congestive heart failure, 24 with CAD and 28 healthy controls displayed an inter-reader bias around 0–5 cm3 with a SD of the differences nearing 5 cm3. A similar variability was also observed for EAT measurements on computed tomography scans by Commandeur et al. [23], with a bias of 4.35 cm3 and a SD of the differences only slightly lower than 10 cm3.
Our study presents some limitations. The first is related to its sample size, which is small and only composed of obese patients who met the criteria for bariatric surgery. However, the EAT volumes observed in this population did not differ greatly from those reported for other patients or healthy subjects; therefore, we may hypothesise that reproducibility was not enhanced by the presence of larger EAT volumes. Moreover, the observed reproducibility was comparable to that reported by previous studies [18, 22]. Secondly, a nontrivial number of patients (n = 3) were still unable to undergo MRI, even though performed on an open-bore system. Nevertheless, the use of such technology still allowed a considerable number of patients who would not have been able to fit in a traditional, closed-bore unit to undergo MRI imaging. Thus, assessing EAT volumes on scans acquired on open-bore units, despite not being the definitive answer for all obese patients, may provide a solution to increase the number of patients that may be screened for cardiovascular risk by EAT evaluation. Third, we did not include in this work clinical data for correlating EAT volumes and cardiovascular condition. However, we note that the main focus of this work was the evaluation of the precision and reproducibility of EAT volumes quantified on scans acquired on open-bore systems, and once this aim has been completed, the main project will include clinical correlations before and after bariatric surgery.
In conclusion, EAT volume can be quantified on cine acquisitions performed on an open-bore MRI scanner, with a seemingly more precise assessment in the presence of higher reader experience and the use of systolic phase scans. This added feature may prove beneficial, as it could allow the assessment of a subclinical cardiovascular risk biomarker from MRI sequences that are already routinely performed for the evaluation of cardiac function in obese patients. Moreover, as an increasing number of methods for automated segmentation of EAT are being proposed showing promising results also on MRI scans [24], images acquired on open-bore scanners may prove a suitable substrate for future clinical developments.