Multipath Curved Planar Reformations of Peripheral CT Angiography: Diagnostic Accuracy and Time Efficiency

Peripheral arterial disease (PAD) and its associated morbidity place a significant burden on both patients and healthcare systems. Based on demographic projections and PAD particularly affecting the elderly [1, 2], the incidence of PAD is expected to further increase [3]. While the diagnosis is made clinically, the guidelines of the Cardiovascular and Interventional Radiological Society of Europe on endovascular treatment in aortoiliac arterial disease recommend a comprehensive radiological assessment for accurate treatment planning [4]. Digital subtraction angiography (DSA) is still considered the standard of reference because it offers the highest spatial and temporal resolution. However, over the last 20 years, computed tomography angiography (CTA) has evolved into an accurate and cost-effective imaging alternative [5‐7], with high clinical impact for patient management [8, 9]. However, the improved spatial resolution has resulted in an increasing number of axial slices [10]. Thus, axial image evaluation is a progressively time-consuming and cumbersome task in the clinical routine [11], which bears the risk of missing small pathologies [12, 13]. 3D reformations provide a better overview and facilitate the visualization of complex anatomical structures. This might result in superior diagnostic performance and shorter evaluation times. For treatment planning, not only the severity, but also length and number of stenoses within a vascular segment are essential parameters according to the Trans-Atlantic Inter-Society Consensus (TASC) guidelines [14]. By providing an angiogram-like view of the peripheral vessels, 3D reformations facilitate the assessment of all treatment-relevant parameters. Thus, many different techniques have been developed. Maximum intensity projections (MIPs) provide a distortion-free display of the arterial tree, but are inherently unsuitable for the assessment of calcified vessels [15]. In addition, automated algorithms for bone segmentation, which is a prerequisite for vascular MIPs, tend to fail when vessels are not separated by soft tissue from the bones, which happens frequently in the path of the anterior tibial artery [16, 17]. Portugaller et al. [18] showed that semitransparent volume rendering (STVR) of CTA provided accuracy superior to that of MIPs in calcified vessels. However, both MIP and STVR failed to provide sufficient diagnostic performance when used alone [19], so that these techniques could be used only as a supplement to axial images. Curved planar reformations (CPRs) were shown to depict the cross-sectional profile of a vessel along its length while preserving the relative X-ray attenuation information [20], resulting in a higher accuracy than MIPs in calcified vessel segments [18]. However, CPRs are limited to a single vessel path, by design [21]. Multipath CPRs (mpCPRs) were developed to overcome this limitation [22]. These reformations were shown to facilitate the assessment of the peripheral arterial tree according to the TASC guidelines [14], when used in conjunction with axial images [23]. However, the diagnostic accuracy of CTA for the assessment of PAD, based exclusively on mpCPRs, has not been evaluated. Thus, the purpose of this study was to compare the diagnostic performance and the time efficiency between mpCPRs and axial images of peripheral CTA for the assessment of PAD, with DSA as the standard of reference.

This prospective single-center study was approved by the local ethics committee. Written informed consent was obtained from all patients prior to recruitment. Patients were referred from the Division of Angiology and Vascular Surgery. The following served as inclusion criteria: Rutherford Category 2–6; age ≥ 18 years; and normal TSH levels. The following served as exclusion criteria: patient on dialysis (estimated glomerular filtration rate < 30 ml/min); pregnancy; and breastfeeding. The time interval between CTA and DSA had to be equal to or less than 30 days to avoid a bias attributable to disease progression. Patients who did not ultimately undergo DSA, or who exceeded the required 30-day interval between CTA and DSA, were excluded from the study.

During the study period of 12 months, 183 PAD patients were referred to CTA of the peripheral arteries. A total of 121 patients had to be excluded because no endovascular, but surgical or conservative treatment was performed after CTA. In another 22 patients, endovascular therapy was not performed within 30 days after CTA, so that the final study cohort consisted of forty patients (30 males, 10 females; mean age 72 ± 11 years; range 44–101) as displayed in Fig. 2. For these patients, the average interval between CTA and DSA was 6 days, with a range of 1–28 days. Six patients of our study cohort had a total of eight orthopedic implants, and 17 patients had a total number of 47 stents in the segments that were evaluated in this study. Demographic characteristics, including risk factors for PAD, are displayed in Table 1.

The main finding of our study is the high diagnostic accuracy of mpCPRs for the assessment of PAD. Whereas the expert reader reached a similar level of accuracy using either axial images or mpCPRs, the diagnostic accuracy of the less-experienced reader was even higher with mpCPRs. This is in contrast to studies on other 3D reformation techniques, such as MIP and STVR, which failed to reach the diagnostic accuracy of axial images when used exclusively [18, 19]. With regard to specificity, MIPs have been shown to be clearly inferior to axial images as well [19]. In this study, however, the specificity of mpCPRs significantly surpassed that of axial images. Only the sensitivity of mpCPRs was, similar to STVRs [19], inferior to that of axial images. Overall inter-reader agreement for mpCPRs was slightly higher than that of axial images, confirming the robustness of this reformation technique.

As previously reported, mpCPRs provide a significantly greater viewing range of the peripheral arterial tree than MIPs [22]. This might be one explanation for the superior diagnostic performance of mpCPRs in this study, compared to the previously reported performance of MIPs [19]. The second advantage of mpCPRs over MIPs is their capability to visualize the vessel lumen inside stents or in the presence of severe vessel wall calcifications, as demonstrated in Fig. 3. Interestingly, both readers declared more vessel segments too calcified for evaluation on axial images than on mpCPRs. This was especially true for vessels below the knee, where the vessel diameter is represented by a few voxels only, so that the identification of contrast media next to severe calcifications can be difficult on axial slices, whereas reformatted images facilitate the assessment of the contrast media continuity along the vessel path.

Interestingly, below-the-knee arteries, in particular their limited number of orthogonal viewing pairs observed in an initial analysis of mpCPRs in 2007, were the reason for the recommendation to use mpCPRs not alone, but rather in conjunction with spCPRs, MIPs, and axial images [22]. Looking at the results of the less-experienced reader only, this recommendation might seem to be confirmed, as he reached a significantly lower sensitivity with mpCPRs for the infra-popliteal arteries, but not for the iliac or femoro-popliteal arteries. However, taking a closer look at Table 2, the less-experienced reader excluded more vessel segments on axial images than on mpCPRs due to low contrast or prosthesis-related artifacts, whereas the absolute number of stenosed segments was almost the same on mpCPRs (n = 137) and on axial images (n = 138). The expert reader, on the other hand, showed no significant difference between mpCPRs and axial images in any statistical test, for the infra-popliteal arteries. Therefore, we hypothesized that the higher number of false-negative findings below the knee for the less-experienced reader might be attributed to some cases with occluded infra-popliteal arteries and early enhancement of the accompanying veins. In these cases, the automatic vessel tracking algorithm can misinterpret the venous overlay as the correct vessel path. Figure 4 shows an example in which the more-experienced reader correctly identified the faulty reformation and rated the vessel as occluded, whereas the less-experienced reader rated the vessel as patent. This faulty reformation should have been identified and corrected during the image post-processing, which clearly shows that this task must be diligently performed to leverage the full strength of this reformation technique.

Although we showed that mpCPRs of the infra-popliteal arteries can comprise pitfalls for less-experienced readers, this reformation technique was advantageous, especially for the less-experienced reader. He reached a higher level of specificity and accuracy than that reached with axial images, overall, and, in particular, in the femoro-popliteal arteries, while maintaining a similar level of sensitivity.

Furthermore, the reading of mpCPRs was statistically significantly faster compared to that of axial images, although the absolute time saving was smaller than anticipated. In particular, for the less-experienced reader, a more pronounced effect was expected. An explanation for the smaller difference might be the sub-segments that were defined for long vessels in this study. For each sub-segment, a separate rating was necessary and the allocation of stenoses to the correct segment required additional time for both axial images and mpCPRs. Of note, technologists routinely perform the image post-processing at our institution and the mpCPRs are stored in the PACS. In other institutions, where radiologists are performing image post-processing themselves, the required post-processing time would exceed the time savings of mpCPRs during image analysis. However, in addition to being faster with the mpCPR images, both readers reported that the reading of axial images was more tiring. In theory, this could lead to an increased number of careless mistakes during the stressful clinical routine [13]. Besides time savings during the diagnostic workup of PAD, the angiogram-like overview of the peripheral vascular tree provided by mpCPRs facilitates the assessment of all treatment-relevant parameters [14, 23]. Thus, they have become the preferred imaging modality for treatment planning in our interdisciplinary vascular board.

There are certain limitations to this study that need to be addressed. First, the DSA procedure was not specifically modified for study reasons due to ethical considerations about radiation exposure. Therefore, DSA was not obtained for all segments. However, this was considered in the initial power analysis. In addition, most vessel segments were visualized by a single plane of DSA; a second plane was only acquired if clinically indicated during the procedure. Second, none of our patients had a femoro-femoral crossover bypass, which causes severe artifacts in mpCPRs due to their path running parallel to the horizontal axis [22]. We recommend CTA for the follow-up of such bypasses, to be assessed on axial images or coronal reformations. Third, the prototype software used requires dedicated hardware and needs to be recompiled for each workstation; thus, it is not commercially available. However, a web-based application with a server processing the datasets in the background is currently being developed and will be made available once completed.

In conclusion, mpCPRs are a promising reformation technique that facilitates a fast assessment of PAD with high diagnostic accuracy. However, radiologists need to be aware of possible reformation-related pitfalls in case of occluded infra-popliteal arteries with venous overlay.