Background
Dual-energy CT (DECT) has gained much attention in recent years, and it is frequently used in clinical practice [
1,
2]. Previous reports have been published to demonstrate its clinical utility for evaluation of various abdominal diseases compared with conventional CT, including radiation dose reduction, iodine extraction, increased lesion conspicuity by increasing iodine contrast, reduced image artifacts such as beam-hardening artifacts, and improved tissue and material characterization [
1,
3‐
10]. These advantages of DECT are attributed to the fact that spectral decomposition of DECT data can differentiate intrinsic attenuation related to different atomic numbers and tissue density; whereas conventional polychromatic images from single energy CT cannot [
11‐
18].
Until now, source-based DECT techniques have had disadvantages that include additional radiation exposure and cross beam scattering [
1]. Recently, spectral detector CT (SDCT) with a dual-layer based detector has been developed. Using this technique, the superficial detector layer absorbs the lower-energy photons, and the deeper detector layer absorbs the higher-energy photons [
6]. The technique may provide several advantages compared to previous DECT approaches [
1,
19]. First, it maintains the capability to produce conventional polychromatic images, as well as a variety of spectral post-processed images. Second, since the energy separation is performed by a dual-layer based detector, it is not necessary to adjust the sampling number and dose modulation, ultimately resulting in reduced radiation dose [
19‐
23] and probably not increased radiation dose in obese patients. Third, advanced planning to use the dual energy mode scanning prior to CT examination is not required, which may also provide the advantages of improved workflow. Finally, virtual monochromatic images can be created in the projection (raw data) domain, which have a theoretic quality benefit compared to image-based methods regarding beam-hardening artifact correction [
10].
Indeed, several studies have been reported about clinical benefits of SDCT. T. Seller et al. addressed that SDCT deliver more accurate iodine concentration values with higher image contrast than source-based DECT [
24]. S. Ehn et al. revealed that SDCT present only small variation (3%) of iodine concentration with increasing phantom size [
25]. In addition, considering the higher contrast resolution of virtual monochromatic images compared to conventional polychromatic images, we surmised that image quality and lesion conspicuity in obese patient could increase without additional radiation exposure using low keV images from dual-layer spectral detector CT. The purpose of this study was to compare subjective and objective image quality of virtual monochromatic spectral images and polychromatic images reconstructed from dual-layer spectral detector CT (SDCT) with different body size and radiation dose using anthropomorphic body phantom.
Discussion
Compared to polychromatic images, we found that subjective image noise score and diagnostic acceptability were improved in both small and large phantoms using VMI. In addition, when the phantom size increased, VMI had more gradual CNR decrease and noise increase than conventional polychromatic images. We can attribute this improvement in image quality to effective noise reduction and the superior CNR of dual-layer spectral detector CT, especially in low energy level. Indeed, lesion conspicuity of low contrast FLLs was significantly increased in VMI, compared with polychromatic images.
Our study results were in good agreement with previous studies of DECT using a new dual-source CT scanner [
32‐
36]. Accumulating evidence suggests that optimized virtual monochromatic images (60~70 keV) with the lowest noise in the reconstructed monochromatic dataset can improve image quality compared with a conventional single energy CT technique with the same radiation dose [
1,
3,
10,
33,
37]. One thing different from previous studies is optimal keV range in VMI. Our study resulted demonstrated that VMI had lower image noise, and higher CNR values at low energy ranges (40 to 75~95 keV for hyper-attenuating FLL; 40 to 80~90 keV for hypo-attenuating FLL) compared with polychromatic images. However, previous studies using the earlier dual-source DECT technique or fast-switching voltage DECT techniques demonstrated that the gain in CNR seen in low energy level virtual monochromatic images is counterbalanced by a substantial increase in image noise, compared with a single energy scan. Therefore, a range of 60 to 70 keV (similar to 120 kVp) was suggested as the optimal energy level [
15,
32,
33,
38]. The improved CNR values of the VMI taken at low energy ranges could be attributed to the increased attenuation of iodinated contrast agents and to vendor-specific noise reduction algorithms, including mitigation of anti-correlative noise after decomposition. In our study, VMI of dual-layer spectral detector CT system allowed the use of lower energy to benefit from the increased iodine CNR, without increasing the image noise at the same radiation dose level used in conventional single-energy CT. The elevated CNR would be beneficial for lesion detection in the abdominal solid organs, including hepatocellular carcinoma or pancreatic tumors; also, low-energy VMI from dual-layer spectral detector CT may provide additional benefit such as radiation dose reduction compared with polychromatic images.
In terms of radiation dose reduction, VMI technique may be used to reduce radiation dose considering the higher CNR values observed in images collected at low energy ranges. The CNR values of 40, 50 and 60 keV VMIs with a DRI 16 was higher than a polychromatic image with a DRI 19. The results were due to the property of iodine accentuation, and they suggest that images with lower DRI values with VMI reconstruction are not inferior to conventional CT images with higher DRI values taken at a low energy range, which offers sufficient diagnostic image quality. In addition, new technologic for noise reduction such as discriminative feature representation (DFR) would be a promising tool for further noise reduction especially in low keV range [
39‐
41].
Phantom size notably affected the CNR and noise in both polychromatic and virtual monochromatic images collected from dual-layer spectral detector CT. As body size increased, VMI had more gradual CNR decrease and noise increase than conventional polychromatic images. The CNR and noise value differences in polychromatic and virtual monochromatic images were greater in large phantom compared with small phantoms (max. Noise gaps in small phantom: 1.4 HU, large phantom: 2.4 HU; higher CNR < 78 keV in small phantom, higher CNR < 82 keV in large phantom). These results may be due to the fact that a polychromatic beam possess a wide energy spectrum; the energy spectrum is more hardened when passing through a larger phantom as the attenuation of lower energy X-rays is higher than that of high-energy X-rays [
4,
7,
42]. Although the use of low-keV virtual monochromatic images may be also limited in patients with large body size [
10], the VMI was able to provide virtual high keV photon images from spectral data as well as low keV photon images to optimize image quality.
In addition, the lesion conspicuity of hypo- or hyper-attenuating FLLs in large phantoms were improved in VMI compared with polychromatic images, especially in low-contrast FLLs. Our study results were also well matched with the results of recent studies which demonstrated improved visualization of hypoattenuating liver lesions [
35,
43,
44] or hyperattenuating liver lesions [
36] using advanced image-based virtual monochromatic images with a recent dual source DECT system compared with SECT scan. In addition, the lesion conspicuity of low-contrast FLLs was significantly improved in VMI compared to polychromatic images (DRI 16,
P = 0.01; DRI 19,
P = 0.02; DRI 22,
P = 0.05), although there was no significant difference among high-contrast FLL images. The improvement of lesion conspicuity in low-contrast FLLs was attributed to effective noise reduction with the use of VMI, suggesting VMI may be an option to differentiate less visible FLLs.
Our study has several limitations we should acknowledge. First, there is a difference between a study phantom and a real human. However, before the clinical application of SDCT, we tried to find out the value of SDCT in obese simulate phantom. Therefore, further studies in real human is strongly warranted. Secondly, we only evaluated the performance of SDCT regarding CNR, image noise, subjective image quality, and FLLs conspicuity. We did not compare data with other DECT vendors. Therefore, the comparison with other vendors should be explored in a future study. Third, we used single parameter set in CT acquisition to focus on comparison of conventional and VMI. Thus, further study regarding the influence of CT parameters are strongly warranted.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.