3-phase dual-energy CT scan as a feasible salvage imaging modality for the identification of non-localizing parathyroid adenomas: a prospective study

Accurate pre-operative localization of parathyroid adenomas (PAs) is essential for successful minimally invasive surgery. At many institutions, including ours, this is done based on two concordant studies. Typically, the most common approach for PA localization is by a combination of sestamibi and ultrasound [1, 2]. However, despite their popularity, these techniques have certain pitfalls. Limitations of sestamibi studies for parathyroid adenoma identification include absent radiotracer retention in some adenomas, diminishing sensitivity with decreasing adenomatous tissue, and potential confounding by concurrent thyroid disease or the occasional hot thyroid nodule [2]. Sestamibi also has low sensitivity for multiglandular disease. Ultrasound is operator dependent and, in addition, locations where ectopic PAs typically are located such as deep within the neck, the retropharyngeal space, and the mediastinum, tend to be areas that are blind-spots for ultrasound [2]. Reported sensitivities for the ability to lateralize (localize) PAs to the correct side of the neck are approximately 57 to 88 % for ultrasound and 65 to 86 % for sestamibi [3]. The addition of SPECT or SPECT/CT may further increase sensitivity for PAs to approximately 90 % or more according to some studies [2, 4], but that has not been the experience in our institution.

4-dimensional CT (4D-CT) is increasingly used for localization of PAs [3, 5–10]. 4D-CT enables characterization of perfusion characteristics of candidate PAs. The main principle behind 4D-CT is that PAs have different perfusion characteristics compared to lymph nodes and normal thyroid gland [3, 5]. In its original form, 4D-CT included a non-contrast acquisition followed by three post-contrast acquisitions that include an arterial phase (usually at 25 s) with two additional scans obtained after variable delays [2, 3, 5]. In general, PAs have more rapid and greater arterial phase enhancement and a more rapid rate of contrast wash-out compared to the normal thyroid gland [5]. Lymph nodes are typically hypoenhancing compared to PAs on arterial phase images, but demonstrate slow progressive enhancement on more delayed images, also a pattern different from typical PAs [5]. The combination of perfusion characteristics and high spatial resolution of CT technique accounts for the success of 4D-CT [5], with some studies reporting accuracy for lateralization of 94 % [5]. As a result, there is increasing interest and use of 4D-CT for PA identification and localization. However, one of the concerns about 4D-CT is radiation exposure because of multiple acquisitions. To this end, there are reports demonstrating that not all of the phases described in the original 4D-CT protocol may be necessary for accurate PA localization [11, 12]. Whereas one approach is to simply eliminate one or more phases from conventional multiphasic CT, another approach is to use more advanced techniques such as dual-energy CT for increasing the diagnostic yield and therefore potentially reducing the number of acquisition needed for a diagnostic exam.

Dual energy CT (DECT) is an advanced CT technique that evaluates tissues at different X-ray energies, enabling spectral evaluation and material tissue characterization beyond what is possible with conventional CT [13–16]. Normally, the attenuation of different tissues and materials varies when scanned at high and low tube voltages, depending on their specific elemental properties. With DECT, projection data are typically obtained simultaneously or near-simultaneously at 80 and 140 kVp (kilovolt peak) [14]. Using sophisticated computer algorithms, the data at different acquisition energies can then be normalized to specific combinations of two reference materials, such as iodine, water or calcium. Furthermore, the spectral data can be used to generate image sets at different predicted energy levels (keV; kiloelectron volts), referred to as virtual monochromatic images (VMI). As such, DECT enables generation of virtual unenhanced images as well as other advanced tissue characterization not possible with conventional CT, all done by post-processing and without the need for any additional scan acquisitions. There are emerging applications of DECT in all of the major subspecialties in radiology [16–22]. In the head and neck, there is increasing evidence that DECT can improve visualization of head and neck squamous cell carcinoma and increase accuracy for evaluation of thyroid cartilage invasion, among other applications [13, 22–28].

Currently, there are only isolated reports of DECT for localization of PAs [29] but no systematic evaluation of this technique. In this study, we prospectively evaluated the utility of multiphasic DECT for PA localization in a group of patients having discordant or unidentified PAs on a workup consisting at a minimum of ultrasound and sestamibi. A 3-phase DECT, without an unenhanced phase, was performed with the ability to create virtual unenhanced images as needed if needed for diagnostic evaluation. This was followed by a post-hoc spectral density evaluation of PAs and lymph nodes.

Primary hyperparathyroidism (PHPT) is most commonly caused by a solitary benign parathyroid adenoma and the treatment is surgical excision [30]. In order to limit the extent of dissection in searching for the pathologic gland, preoperative localization studies are used [31]. These studies currently include sestamibi scanning, ultrasonography, computed tomography (CT), magnetic resonance imaging (MRI), positron emission tomography (PET) and angiography [2, 3, 32–34]. Pre-operative concordant images have a dramatic impact on success and associated morbidity of minimally invasive surgery [35].

There is increasing popularity of 4D-CT for localization of PAs although there are concerns about the radiation exposure associated with the classic 4 phase 4D-CT [3, 5–10]. While the effective dose of the typical 4D-CT protocol is greater than that of scintigraphy, studies have shown the lifetime incidence of cancer compared to baseline cancer risk for this population is negligible for either study [36]. Despite this and in order to minimize radiation exposure from the procedure to the extent possible, some groups are decreasing the number of acquisitions [11, 12]. DECT scans can be used to create virtual unenhanced images or iodine overlay maps that can estimate the iodine content of a structure based on a single acquisition [14, 15]. In addition, virtual monochromatic images can be created at different energy levels, and these can be evaluated quantitatively, potentially increasing the analytic capabilities of CT technique [13–15]. This could potentially improve accuracy and in turn enable a reduced number of acquisitions.

In this investigation, we used a 3 phase CT technique, with DECT acquisitions, to localize unidentified or discordant PAs prospectively. Consistent with other studies [3, 5–12], multiphasic CT was effective in localizing PAs in a significant percentage of cases, including cases of multi-gland disease and intrathyroidal PA. Although the radiologists had the ability to generate and use virtual unenhanced images, this was overall deemed not necessary and after the use for two cases in the early part of the study, without benefit, these were not created or used for the other cases. The problem of using virtual unenhanced images for PA identification is that both the iodinated contrast in the enhancing PA and the intrinsic iodine within the thyroid gland are suppressed (Fig. 4), defeating the purpose of the reconstructions for distinction of PA from thyroid tissue. On the other hand, we also demonstrate successful identification of PAs prospectively without the need for an unenhanced scan. During the early part of the study, two patients were called back for an unenhanced study. However, in retrospect these were deemed not necessary and were not performed in any of the patients later on. This is consistent with more recent reports demonstrating successful “4D-CT” with a reduced number of phases [11, 12].

Although our investigation did not reveal a role for the DECT virtual unenhanced images for PA localization, post-hoc spectral Hounsfield unit attenuation curve analysis demonstrated a difference in the characteristics of PAs compared to lymph nodes on arterial phase images (Fig. 5). Arterial phase images are one of the most important acquisitions in 4D-CT performed for PA localization and these preliminary observations suggest that DECT can further increase accuracy during this phase of the exam, which may in turn enable further reduction of the number of acquisitions without decreasing diagnostic accuracy in the future. Other potential applications of DECT could be improved PA visualization on low energy virtual monochromatic images and the use of estimated iodine content for PA identification (Fig. 2). These are topics of great interest for future research.

Although the perfusion characteristics are central in identifying PAs, other features such as location, presence of feeding or polar artery, and other characteristics that help distinguish PAs from normal lymph nodes such as cystic internal change were also important for accurate identification of PAs (Table 2). Furthermore, not all of the PAs demonstrated a typical robust arterial phase enhancement with rapid wash-out (Fig. 2). Therefore, rather than focusing on absolute thresholds, it may be more important to identify combinations of features that help distinguish PAs from potentially mimicking normal structures. Of course, it is possible that we observed a higher frequency of atypical appearing PAs because the study was used to evaluate unidentified or discordant PAs, resulting in a selection bias.

In this study, we were able to localize previously unidentified parathyroid adenomas in 26 of 29 patients. Twenty three of these patients have undergone surgical exploration at this time, and DECT correctly identified 21 PAs in 20 of those patients. We also demonstrate a high success-rate of minimally invasive parathyroidectomy on primary cases and even some secondary cases. Among the three false positives, two were in patients with prior surgery. Therefore, one must at least consider the possibility that these PAs may have not been found because of extensive scarring from the patient’s prior surgery. One of the strengths of this study is that all PA identification was done prospectively. The limitation is that the numbers are relatively small. As many of these were outside referrals, another limitation could be that not all of the standard imaging was done at the institution were DECT was performed, potentially introducing a bias. However, among the 20 patients who successfully underwent surgery, 13 had sestamibi and 12 US at the same institution and therefore the proposed bias could not account for the success of DECT in these cases. In addition, a small number of patients could not be analyzed because they have not undergone surgery yet (either because of loss to follow-up or surgical wait-list time). Nonetheless, our results demonstrate the feasibility of a multiphasic study without an unenhanced phase and promising results for DECT spectral analysis for improving the diagnostic evaluation of PAs. The impact of more advanced DECT analysis will have to be tested in larger and ideally prospective use of these characteristics in future studies.

In this prospective study, we demonstrate that a 3 phase CT technique, with DECT acquisitions and without an unenhanced phase, has high accuracy in identifying previously unidentified or discordant PAs. Furthermore, our post-hoc analysis demonstrates significant differences in the spectral characteristics of PAs compared to lymph nodes on arterial phase images. This suggests that advanced DECT analysis has the potential to further increase accuracy for PA identification, which could potentially enable a reduction in the number of CT acquisitions and associated radiation exposure. This is an interesting topic for future research.