Dual-Energy Computed Tomography in Stroke Imaging

Stroke is the second leading cause of mortality worldwide and an important contributor to disability [1]. Numerous studies have shown the benefit of endovascular mechanical thrombectomy (EVT) in patients with large vessel stroke [2, 3]. Follow-up imaging after the intervention can be performed to assess the extension of the ischemic brain tissue, to rule out intracranial hemorrhage, to evaluate the vessel status and to exclude further potential complications. As this might impact the ongoing treatment as well as the prognosis for the patient, an affordable and easily accessible one stop imaging technique is one of the focuses in stroke imaging and assessment. Dual-energy computed tomography (DECT) can be a helpful diagnostic imaging modality in order to optimize these imaging tasks. Currently there are a number of available DECT techniques of the main vendors, such as: (1) dual-source X‑ray tubes performing simultaneous acquisitions using different tube voltages (e.g., 80–100 kV and 140–150 kV), single-source rapid voltage switching between two tube voltages, a single-source split beam into two different energy spectra and a dual-layer detector with simultaneous data acquisition of the low-energy and high-energy dataset [4‐7]. Various DECT applications have been implemented in stroke assessment, for example, in differentiating acute intracranial hemorrhage from contrast staining or calcification [8, 9], visualization of ischemic changes and brain edema [5, 10, 11], prediction of infarction development or hemorrhagic transformation in contrast-enhancing areas [12], and detecting a hyperdense artery sign [13]. Recently, a new technique has been developed, TwinSpiral DECT, a variant of dual spiral DECT, which uses a tin filter and an improved detector enabling low-dose dual-energy scans for optimal spectral separation with a better soft tissue contrast [14]. It acquires two sequential scans in a very short time, one with high and one with low energy immediately after each other.

A recent publication has shown the feasibility of TwinSpiral DECT to differentiate between intracranial hemorrhage and iodine staining with a high accuracy [15]. The aim of the current study was to assess whether TwinSpiral DECT allows an improved visualization of ischemic brain tissue after mechanical thrombectomy.

Cohen’s κ‑coefficients were calculated to evaluate the inter-rater agreement regarding the assessment of contrast and image noise in VNC and mixed images. The κ values below 0.4 were interpreted as poor and between 0.41–0.75 as fair to good according to criteria originally proposed by Landis and Koch [19].

Comparison of qualitative scores (contrast and image noise) between VNC and mixed images was conducted using the pairwise Wilcoxon signed rank test to assess for significant differences.

For quantitative analysis, a statistical significance of mean in differences of HU values for each reconstruction modality was determine using the paired Student’s t‑test. A p-value < 0.05 was considered statistically significant. Statistical analyses were conducted using commercially available software (IBM SPSS Statistics, Version 26.0, IBM Corp. Armonk, NY, USA and Microsoft Excel, Microsoft Corporation. (2019). Microsoft Excel. Retrieved from https://​office.​microsoft.​com/​excel).

This study investigated the value and feasibility of a new DECT technique: TwinSpiral DECT, for an improved ischemic stroke detection and visualization. Our results suggest that TwinSpiral DECT is a feasible method to be performed in patients after mechanical thrombectomy with an improved visualization of the ischemic tissue in VNC compared to the standard mixed images; however, despite our paper focusing on patients post-EVT, as DECT is primarily performed in patients after EVT, the better contrast visibility between healthy and ischemic tissue could be implied in initial stroke assessment as well. Thus, the next step would be a study in which this DECT technique would be applied for the initial imaging in the acute ischemic stroke setting (i.e. before EVT) to assess the benefit of VNC images for the detection of ischemic brain tissue and therefore for a potentially more accurate clinical prognosis and optimal therapeutic management.

By comparing the qualitative values of each individual reader, the VNC images from the tested DECT demonstrated the affected ischemic brain tissue to be significantly more hypodense than in standard mixed images, although with a slight increase of noise. The infarcted areas were better visible in VNC images compared to standard mixed images. Measurements of HU and comparing the values of VNC and standard mixed images showed a markedly increased difference in radiodensity between ischemic and healthy brain tissue, thus allowing an improved delineation and assessment of ischemic stroke.

Detecting ischemic changes post-EVT is just as important as pre-EVT. The distant embolization of the affected vessel and infarcts in new vascular territories are both associated with worse prognosis due to newly developed ischemic regions [20, 21] and impairment of collateral blood flow to already affected areas [22]. Prompt treatment, either by EVT or intra-arterial thrombolysis, is important for favorable clinical outcome and prevention of further complications [23].

The results of our study are comparable to results from Gariani et al. [10]. They have shown to be able to separate ischemic from non-ischemic tissue with major advantages in DECT by using VNC compared to non-contrast CT weighted sum images (NCCT). Another study showed superiority of VNC images compared to conventional images (using a dual-layer spectral CT) in ischemic infarct delineation [24], as well as in detection of ischemic changes in posterior fossa [25]; however, both studies were performed on a DECT acquisition technology other than TwinSpiral DECT. In addition, previous studies have also reported a higher noise in both VNC [26] and virtual monoenergetic imaging [27, 28], which was also observed in our study. Nevertheless, the increased image noise seems not to have a major impact on the detection of hypodense ischemic brain tissue.

The first concept of DECT was mentioned by Hounsfield in 1973, when he described acquiring two separate images (at 100 kV and the other at 140 kV) at the same slice, in order to differentiate iodine from calcium based on different atomic numbers [29]. This required two scans for a single series of images, thus approximately doubling the radiation dose. Recently, a new DECT image technique was developed, TwinSpiral DECT, enabling two scans at low-kV and high-kV data sets integrated into one single acquisition, without an increase in radiation dose. This method does not require dual source, dual layer, or rapid energy switching ability. Instead, a tin filter provides powerful spectral separation [30, 31], while an improved detector used herein enables high-quality imaging at a high temporal resolution in low dose or low signal imaging [14], with utilization of advanced (semi-)automated postprocessing algorithms enabling facilitated routine-ready solutions. As such, it may be applied by a broader pool of health institutions, with less costs than other high-end DECT scanners.

One of the drawbacks of the TwinSpiral DECT technique could be its susceptibility to motion artefacts, such as in agitated patients, where the movement of the patient between the low kV and the high kV scan could potentially compromise the quality of DECT images. In our series, all of the included studies demonstrated diagnostic image quality. No major motion artifacts were seen, which also reflects our personal clinical experience from daily routine with this new dual energy technique; however, even in the case of patient movement in one of the two scans, the second scan could still be used for diagnostic purposes (even though the dual-energy information might get lost). Furthermore, it is important to note that the mean radiation dose was within the range of a normal single-energy head CT scan and in the range of the diagnostic reference levels (DRLs) and achievable doses (ADs) in the USA [32].

In addition to the above investigated application, the TwinSpiral DECT technique might also provide further already established benefits of a DECT, not only in detecting ischemic changes, but differentiating between hemorrhage [15], iodine, calcifications and metal artifacts, without the need for multiple scans [33, 34]. Furthermore, this study might open new frontiers in stroke research, as to evaluate the technique’s ability for ischemic stroke in the posterior circulation or for an improved stroke detection in initial patient presentation.

Although there was somewhat of a discrepancy between inter-reader agreement in qualitative image analysis in our study, each reader represented a separate control, and what is most important, both readers were able to detect significant changes in contrast between VNC and standard mixed images, a key finding in ischemic brain tissue.

Limitations of our study were the relatively small retrospective study size and blinding, which was not completely possible, as experienced readers could easily recognize the type of DECT imaging reconstruction. Furthermore, qualitative image analysis could be done by ASPECTS, which, in future studies including more patients, might be used to assess prognosis in pre-EVT in post-EVT patients.

Our study demonstrated that TwinSpiral as a new DECT technique allows an improved qualitative and quantitative visualization of ischemic brain tissue. This indicates that TwinSpiral DECT might be a valuable tool for imaging of ischemic stroke patients after endovascular treatment.