Spatial Heterogeneity and Distribution of CT-Based Pulmonary Vascular Volumes in Chronic Thromboembolic Pulmonary Hypertension

INTRODUCTION:

Chronic thromboembolic pulmonary hypertension (CTEPH), classified as Group 4 in the World Symposium on Pulmonary Hypertension (WSPH) system, is a relatively uncommon form of PH that is characterized by persistent obstruction of the pulmonary vascular bed due to organized thrombus, neointima formation, and vascular webs[1,2]. Although there are a growing number of effective therapies for this condition (including surgical endarterectomy, endovascular methods, and pharmacologic agents), CTEPH remains underdiagnosed and undertreated[1–3].

Many patients experience long delays before they are diagnosed with CTEPH, as it typically presents with nonspecific symptoms and definitive diagnosis requires right heart catheterization (RHC)[2], an invasive procedure requiring an experienced operator which is not available at all centers. The use of non-invasive tools to identify a CTEPH signature may help raise the index of suspicion earlier in the disease course, expediting referral for RHC and reducing potentially avoidable morbidity. In addition, better methods of sub-phenotyping CTEPH may help improve treatment strategy. For example, when assessing surgical candidacy, it remains clinically challenging to differentiate those patients with predominantly proximal disease from those with more distal vasculopathy – a critically important distinction given that endarterectomy is potentially curative in the former group but is technically challenging and confers minimal benefit in the latter[2,4].

Image-based phenomarkers have been proposed as a complementary diagnostic tool in the evaluation of suspected CTEPH[5]. Computed tomography (CT) scans are frequently obtained in the workup for CTEPH, and quantitative metrics based on vascular volumes can be generated from clinical CTs using automated image analysis algorithms. Prior work has demonstrated that global differences in vascular volumes can be quantified in several forms of PH and may help distinguish PH subgroups[6,7]. Notably, CTEPH is characterized by a distinctively heterogeneous (“moth-eaten”) pattern of vascular abnormality[8] due to regional vascular defects[9], and prior work has suggested that the vascular volume in CTEPH may be redistributed towards the larger central vessels and away from the periphery[5]. In contrast, the primarily distal vasculopathy of Group 1 pulmonary arterial hypertension (PAH) is more globally homogenous in comparison[9]. Novel phenomarkers based on these features, derived from non-invasive imaging, could ultimately help to expedite the diagnostic process for individuals with CTEPH.

In this study, we sought to quantify measures of vascular heterogeneity and redistribution on CT, and hypothesized that patients with CTEPH would demonstrate greater variability and central distribution of vessel volumes compared to controls and patients with PAH.

MATERIALS AND METHODS:

RESULTS:

Of the 108 patients who comprised the study sample, 21 (19.4%) had CTEPH, 47 (43.5%) had PAH, and the remaining 40 (37.0%) formed the control group. Characteristics of the cohort, including mean CVs and central-to-peripheral ratios stratified by group, are presented in Table 1. The median interval between CTA and RHC was 1.4 months (interquartile range: 4.6 months). Examples of vascular reconstructions illustrating spatial heterogeneity and central-peripheral distribution are shown in Figure 2 and Figure 3, respectively.

DISCUSSION:

In this retrospective, single-center study of patients who underwent hemodynamic characterization with invasive RHC and CT angiography, we found that individuals with CTEPH had significantly greater spatial heterogeneity of the pulmonary vessels and greater central vascular redistribution compared to a control group, in both unadjusted and adjusted models. We also found evidence of greater heterogeneity and central redistribution of the small vessels (both arterial and venous) in patients with CTEPH compared to those with PAH. Our findings suggest that volumetric measures of heterogeneity and central distribution of pulmonary vessels can be quantified using CT techniques and may contribute to an image-based signature of CTEPH.

There is an urgent clinical need to improve non-invasive diagnostic tools for CTEPH. Many patients with CTEPH experience long delays prior to definitive diagnosis, preventing timely initiation of proven medical and surgical therapies. A major contributor to this delay is that CTEPH is often not clinically suspected until late in the diagnostic process, due to a prolonged asymptomatic period, nonspecific presenting symptoms, and the fact that a substantial minority of patients with CTEPH do not have any previously documented acute thromboembolic event[1,2,9]. While a definitive clinical diagnosis of CTEPH will always include RHC at a center with specialized expertise, the development and deployment of accurate, non-invasive tools may serve to increase the index of suspicion for CTEPH, thereby accelerating specialist referral, improving access to therapy, and ultimately improving outcomes in affected patients.

Through the use of quantitative image analysis, biomarkers based on pulmonary vascular volumes can be generated which have shown promise in improving diagnostic strategy for patients with PH of various groups. For example, prior work from our group has demonstrated that lower volumes in the small pulmonary arteries on CT helped to discriminate between individuals with PAH, exercise-induced PH, and control subjects[6], and we recently reported that greater volume of the large pulmonary veins may be a more specific radiographic feature of Group 2 pulmonary venous hypertension (from left heart disease) as compared patients with PAH and control participants[7]. In CTEPH specifically, we previously showed that lower volumes of the small pulmonary vessels (analogous to the small vessel obstructive vasculopathy observed histologically in CTEPH) correlated with invasive hemodynamic measurements[5].

Pathophysiologically, CTEPH is caused by obstruction of the pulmonary vessels at both the micro- and macro-vascular levels[1]. This results in certain features associated with CTEPH, including greater heterogeneity/variability of the vasculature[5,8,9,13,14] and a particularly marked enlargement of the central pulmonary vessels[5,9,13,15].

However, very few studies have examined whether CT-based measures can be used to quantify vascular heterogeneity. A study by Koike et al. examined metrics of heterogeneity of pulmonary perfusion using dual-energy CT in a cohort of 33 patients with CTEPH who underwent balloon pulmonary angioplasty[16]. They found that heterogeneous pulmonary perfusion was a typical feature of CTEPH, and that treatment with balloon angioplasty substantially reduced the degree of heterogeneity, and this also correlated with improvement in hemodynamics and exercise capacity. These results suggest that heterogeneity of the vasculature that is characteristic of CTEPH can be quantified on imaging, and this may potentially be a marker of hemodynamic severity and a predictor of response in CTEPH.

Our study adds to the limited literature in this area by demonstrating that heterogeneity of the small and total vascular volumes on CTA can be quantified using automated techniques, and that these metrics may help distinguish those with CTEPH from controls who do not have pulmonary vascular disease. In addition, in adjusted models we found that patients with CTEPH exhibited greater heterogeneity of the small arteries and veins (although not the total volumes) compared to PAH. Importantly, these results persisted after adjusting for the baseline vascular volume – indicating that heterogeneity of the small vessels is a more specific feature of CTEPH regardless of the underlying severity of the vascular abnormality. Additionally, we found evidence of greater central re-distribution of small vessel volumes in CTEPH compared to the control and PAH groups across all four vascular compartments. Of note, while the term “central vessels” is typically used to refer to the large vessels, smaller vessels are also present in innermost region of the lung. Our study quantified these small (but still centrally located) vessels, and found similar evidence of greater central redistribution in CTEPH as with the large vessels. A strength of our study is that all participants underwent detailed clinical phenotyping at a center of clinical expertise and incorporated advanced radiomic techniques including arterio-venous segmentation using a neural network approach.

A particular advantage of the automated method employed in this study is that these metrics can be generated from clinical CTAs, which are routinely performed in virtually all patients in whom either CTEPH and/or acute pulmonary embolism (PE) are suspected. CT imaging has strong advantages compared to other non-invasive imaging tools. For example, promising techniques such as hyperpolarized Xenon MRI can provide dynamic assessment of pulmonary structure and function in CTEPH[17,18], and is now FDA-approved for use in the United States[19]. However, this method is limited by a substantial costs (both for initial setup and ongoing use), requires highly specialized expertise, and is only available at a handful of centers nationwide[19] – meaning that the vast majority of patients will not be able to benefit from this technology. In contrast, CT angiography is the imaging modality of choice in the diagnostic evaluation of CTEPH[2] and is widely available. Critically, CT-based image analysis methods can be automated, hosted on local workstations, and thus feasibly incorporated into clinical radiology workflows to help improve the diagnostic process for CTEPH. For most patients with suspected CTEPH, CT has already been performed as part of their evaluation and so no additional testing or radiation exposure would be required to generate these metrics.

Although our study was not powered to detect differences within the CTEPH group, one potentially intriguing application of CT image analysis is to improve the identification of sub-phenotypes of CTEPH. A particular area of interest is to identify if the CT metrics utilized in this study correlate with defects in pulmonary perfusion, which are a key finding in CTEPH and play an important role in determining operability of disease[20]. One of the challenges in CTEPH care is in differentiating between those more classical proximal disease (for whom surgical pulmonary endarterectomy is considered for up-front and potentially curative treatment[4]) and those with more distal vasculopathy who may not benefit from surgical intervention, and indeed may even be harmed[2,4]. In recent years, dual/multi-energy CT methods have emerged, in which additional attenuation measurements (obtained from two or more energy spectra) allows for noninvasive mapping of pulmonary iodine uptake as a surrogate of pulmonary perfusion[21–23]. This method can visualize and quantify perfusion defects in CTEPH, and may even be able to identify patterns that differentiate acute PE from CTEPH[21]. Although we are not aware of any studies that correlate the vascular metrics utilized in this study with those derived from more modern multi-energy CT, this is an area we plan to pursue in future work. Ultimately, multi-modality sub-phenotyping of vasculopathy may help to individualize CTEPH therapy, given the rapidly evolving and increasingly nuanced therapeutic landscape, which includes surgical endarterectomy, directed balloon angioplasty, and pharmacologics[2].

Our study has several notable limitations. Our sample was derived from a single institution and all data were retrospectively collected. In addition, the RHCs and CTAs were not performed contemporaneously, and certain clinical factors (e.g. volume status, treatment status) may have changed from one test to the other; however, we would not expect systemic bias to be introduced based on these factors. Unfortunately, treatment status with pulmonary vasodilator therapy at the time of RHC and CTA is not available for this cohort, and our study is unable to account for the effect of treatment. We note that our control group was derived from a patient population that presented for clinical evaluation of unexplained dyspnea, and thus may not be fully representative of individuals without respiratory symptoms. However, our control group did not have any evidence of pulmonary or pulmonary vascular disease based on the normal results of their invasive hemodynamic assessment and CT scan. Residual or unmeasured confounding cannot be excluded in this observational study, although our models were adjusted for some potential confounders.

As none of the patients in our study had acute PE and all of the patients in the CTEPH group had PH, our study is not designed to determine if differences in CT-based vascular heterogeneity can distinguish between CTEPH, those with acute thromboembolism, and those with chronic thromboembolic disease but without PH (CTED). In addition, we do not have data regarding longer term clinical outcomes for our cohort, and are unable to determine if our findings carry any prognostic significance. Our study is also limited by the lack of other forms of PH beyond groups 1 and 4 (including PH associated with left heart disease and chronic lung disease), and therefore it remains unclear if the findings of our study can be extended to these other common forms of PH. However, in future work we plan to deploy these methods in longitudinal cohorts (with extended follow-up and adjudicated clinical outcomes) of patients with acute PE, CTED, CTEPH, and other etiologies of PH to investigate these important questions, which must be addressed prior to any potential clinical application of these image-based metrics. After such validation has been performed, such tools could conceivably be incorporated directly into clinical workflows. For example, in a patient who undergoes chest CT imaging for evaluation of exertional dyspnea, volumetric vascular reconstructions with measurement of vascular volume distribution and heterogeneity could be calculated in real-time using standard clinical workstations. While a standard visual inspection of a CT performed for dyspnea will often not demonstrate any notable findings, the automated radiomic classifier may assist the radiologist by identifying subtle decreases in the small arterial volume, greater central redistribution, and greater vascular heterogeneity. By interpreting this specific constellation of imaging findings as consistent with intermediate/high probability of CTEPH, the radiologist may then be able to include a stronger recommendation to obtain additional testing (e.g. V/Q scan) and help reduce the delay to diagnosis and treatment.

In conclusion, the results of our study suggest that spatial heterogeneity and central redistribution, particularly of the small vessels, may be a more specific imaging feature of CTEPH that could help differentiate this disease from individuals with PAH and those without PH. These findings must be confirmed in larger, prospective cohorts, including in patients with chronic thromboembolic disease but without PH. If confirmed, our results may support the application of CT-based quantitative methods, alongside existing diagnostic tools, to improve and expedite the diagnostic evaluation of this morbid disease.

Supplementary Material