Gastrointestinal Applications of Iodine Quantification Using Dual-Energy CT: A Systematic Review

Dual-energy computed tomography (DECT) can estimate tissue vascularity and perfusion via iodine quantification. The aim of this systematic review was to outline current and emerging clinical applications of iodine quantification within the gastrointestinal tract using DECT. The search was conducted with three databases: EMBASE, Pubmed and The Cochrane Library. This identified 449 studies after duplicate removal. From a total of 570 selected studies, 30 studies were enrolled for the systematic review. The studies were categorized into four main topics: gastric tumors (12 studies), colorectal tumors (8 studies), Crohn’s disease (4 studies) and miscellaneous applications (6 studies). Findings included a significant difference in iodine concentration (IC) measurements in perigastric fat between T1–3 vs. T4 stage gastric cancer, poorly and well differentiated gastric and colorectal cancer, responders vs. non-responders following chemo- or chemoradiotherapy treatment among cancer patients, and a positive correlation between IC and Crohn’s disease activity. In conclusion, iodine quantification with DECT may be used preoperatively in cancer imaging as well as for monitoring treatment response. Future studies are warranted to evaluate the capabilities and limitations of DECT in splanchnic flow.


Introduction
Multi-detector computed tomography (MDCT) is the first-line imaging modality for various conditions ranging from acute bleedings [1] and acute intestinal ischemia [2] to Crohn's disease [3] and gastrointestinal cancer [4,5]. While MDCT is based on X-ray emissions of one energy level, dual-energy computed tomography (DECT) acquires datasets at two different energy levels, either through emission or through separation at the detector level, providing new imaging and reconstruction possibilities [6]. Current DECT scanners produce images comparable to conventional single-energy computed tomography (CT) without increased radiation exposure or decreased image quality [7,8].

Materials and Methods
This systematic review was performed using the preferred reporting items for systemic reviews and meta-analyses (PRISMA) guidelines [33]. The review protocol was not published in advance.

Search Strategy
The literature search was conducted on the 13th of May 2020 using three databases: Pubmed, EMBASE and Cochrane Library. The search was restricted to peer-reviewed publications of original research using the population, intervention, comparison and outcome approach (PICO) model [34]: the patient group had gastrointestinal-related conditions; the intervention consisted of DECT examination from which IQ was measured; the comparison was to other verified methods of evaluation for the given condition, such as conventional CT, surgical findings, pathology, biochemistry, etc.; and the main outcome was establishing whether DECT examination and IQ measurements correlate with verified measures of evaluation.
The search thread in PubMed contained two aspects, consisting of MeSH (Medical subject headings) terms as well as text words (in the title and/or abstract), which were combined using "OR" and "AND".
The first aspect focused on DECT, including the MeSH term "Radiography, Dual-Energy Scanned Projection" as well as a combination of the MeSH term "Tomography, X-Ray Computed/methods" with the text word "dual energy". Additional text words were added including "DECT", "Dual-energy CT", "Spectral CT" and "Dual-energy Computed Tomography".
The Cochrane Library search was conducted with the Pubmed search string using two aspects combining MeSH terms as well as free text words. The function "explode all trees" was applied to all MeSH terms as well as "include word variations" to all free text words.
The EMBASE search was conducted in a similar fashion using the EMTREE terms "dual energy computed tomography", "gastrointestinal disease" and "gastrointestinal tract" in combination with identical text words.

Study Selection
Study selection was conducted on the online platform covidence.org. The initial selection was based on study titles and abstracts by two independent assessors (J.J.X. and M.T.). The first assessor was a PhD fellow with one year of clinical radiology experience, and the second assessor was a radiologist specialized in interventional radiology with 10 years of clinical radiology experience. Selection of studies was carried out separately. Studies were selected based on the presence of search terms in abstract and title, and the full text was retrieved for studies that were eligible or possibly eligible, and then independently screened by the assessors. Discrepancies regarding potential eligibility and inclusion were resolved by consensus.
Eligibility for this systematic review included two inclusion criteria: DECT of the gastrointestinal tract and DECT including or focusing on IQ as an outcome measure. Exclusion criteria were as follows: study population <10 patients, phantom-only or animal-only studies, language other than English, review articles, case reports or editor's letters.

Quality Assessment
Potential risk of bias was assessed using the QUADAS-2 tool [35]. The four domains-patient selection, index test, reference test, and patient flow-were assessed for potential risks of bias and applicability concerns ( Figure S1). The rating score was either low ( ), high ( ), or unclear (?).

Results
The primary search thread identified 570 studies for inclusion in the methodological review. After duplicates were removed, 449 studies were screened based on set inclusion and exclusion criteria resulting in the inclusion of 30 studies (Figure 1). The 30 included studies involved 1778 patients with a mean population size of 59 patients (range: . Twelve studies involved gastric tumors, eight studies on colorectal tumors, four studies on Crohn's disease, and six on miscellaneous applications. The latter included peristalsis-related streak artifact reduction, acute bowel ischemia, esophageal cancer and gastrointestinal stromal tumor (GIST) risk stratification.

Gastric Tumors
Four studies demonstrated a significant difference in IQ when comparing poorly differentiated and moderately/well-differentiated adenocarcinoma with region of interest (ROI) placement on a solid tumor mass. Outcome measures of nIC in arterial phase (nIC-A) and nIC in venous phase (nIC-V) were used in all studies, with one study also including IC in arterial phase (IC-A) and IC in venous phase (IC-V) [36][37][38][39]. Among these four studies, there seemed to be no correlation with TNM classification of malignant tumors. However, studies investigating IC in perigastric adipose tissue among gastric cancer patients [40][41][42][43] found a significant difference between patients with (T4) and without serosal invasion (T1-T3). Additionally, a study by Cheng et al. [44] noted a significant difference between early (confined to mucosa/submucosa) vs. advanced gastric cancer (invasion of the submucosa) in nIC-V and nIC in delayed phase (nIC-D) with ROI placement on the gastric mass. One study by Tang et al. [45] demonstrated that the relative reduction in IC (defined by [ICafter − ICbefore]/ICbefore × 100%) following neoadjuvant chemotherapy correlated well with the histopathological regression. Furthermore, two studies by Liu et al. [46] and Meng et al. [47] found a significant difference in IC between GIST and gastric schwannomas, as well as gastric cancer and normal gastric mucosa or gastric inflammation (Studies shown in Table 1).

Gastric Tumors
Four studies demonstrated a significant difference in IQ when comparing poorly differentiated and moderately/well-differentiated adenocarcinoma with region of interest (ROI) placement on a solid tumor mass. Outcome measures of nIC in arterial phase (nIC-A) and nIC in venous phase (nIC-V) were used in all studies, with one study also including IC in arterial phase (IC-A) and IC in venous phase (IC-V) [36][37][38][39]. Among these four studies, there seemed to be no correlation with TNM classification of malignant tumors. However, studies investigating IC in perigastric adipose tissue among gastric cancer patients [40][41][42][43] found a significant difference between patients with (T4) and without serosal invasion (T1-T3). Additionally, a study by Cheng et al. [44] noted a significant difference between early (confined to mucosa/submucosa) vs. advanced gastric cancer (invasion of the submucosa) in nIC-V and nIC in delayed phase (nIC-D) with ROI placement on the gastric mass. One study by Tang et al. [45] demonstrated that the relative reduction in IC (defined by [IC after − IC before ]/IC before × 100%) following neoadjuvant chemotherapy correlated well with the histopathological regression. Furthermore, two studies by Liu et al. [46] and Meng et al. [47] found a significant difference in IC between GIST and gastric schwannomas, as well as gastric cancer and normal gastric mucosa or gastric inflammation (Studies shown in Table 1).

Colorectal Tumors
Two studies demonstrated a significant difference in IC between poorly and moderate/well-differentiated colorectal cancer [48,49]. Chuang-bo et al. [49] only found a significant difference during the arterial phase, while Gong et al. [48] reported a significant difference in both the arterial and the venous phase. Similar to the gastric cancer studies [42,45], a positive correlation between IC and pathological grading of rectal cancer prior and following chemoradiotherapy treatment was reported [50]. However, there is a discrepancy of whether IC may be used in differentiating between malignant and benign colorectal tumors. While Al-Najami et al. [51] found no significant differences in IC when comparing malignant with benign rectal tumors, Sun et al. [52] demonstrated significant differences in IC between colonic adenomas and adenocarcinomas. Additionally, three studies determined positive correlations with other paraclinical measures such as perfusion computed tomography (CT) parameters (blood flow, blood volume, permeability, mean transit time), immunohistochemical evaluation of Ki-67 and HIF-1α levels as well as microsatellite stability and instability [53][54][55] (Studies shown in Table 2).

Crohn's Disease
Two of four DECT enterography studies concerning Crohn's disease used Crohn's disease activity index (CDAI) as reference [56,57], and two studies used endoscopy, biochemistry and clinical symptoms as reference [30,58]. The outcome measure included nIC-V for all studies except one study by Kim et al. [56], which only included IC-V. ROI placements were set on either iodine maps or conventional images covering the most enhanced areas. All studies found a strong correlation between IC measurements and CDAI or endoscopy findings (Studies shown in Table 3).

Miscellaneous Applications
Similar to gastric and colorectal cancer, Ge et al. [59] demonstrated a positive correlation among esophageal cancer patients between nIC and response following chemoradiotherapy based on response criteria in solid tumors (RECIST) criteria. Unlike the studies in the gastric and colorectal group [36,48,49], two studies [31,60] suggested that IC/nIC can differentiate between specific cancer subtypes such as squamous cell carcinoma and adenocarcinoma in the gastroesophageal junction, as well as discriminate between small bowel adenocarcinoma and primary small intestine lymphoma. Additionally, Zhang et al. [32] found a significant difference in IC between high-and moderate/low-risk GIST patients based on GIST recurrence risk stratification criteria [61].
There were two outlying studies in this group. The first study by Lourenco et al. [62] demonstrated a significant reduction in IC in bowel segments suffering from acute bowel ischemia. The second study by Winklhofer et al. [63] demonstrated that peristalsis-related streak artifacts can be reduced using iodine maps (Studies shown in Table 4).

Discussion
This review represents a heterogenous group of studies with a major focus on gastrointestinal cancer evaluation and diagnoses. The main findings include positive correlations between IC and degree of cell differentiation in adenocarcinomas, treatment response following chemo-or chemoradiotherapy, Crohn's disease activity and differentiation of T1-3 vs. T4 stage gastric cancer (Tables 1-3).
One of the most convincing DECT applications for gastrointestinal imaging is probably related to the differentiation of T1-3 vs. T4 stage gastric cancer based on IC measurements in the perigastric adipose tissue as seen in Table 1 [40][41][42][43]. An additional finding was the correlation between IC and varying degrees of cell differentiation in adenocarcinoma. Two colorectal cancer and five gastric adenocarcinoma studies [36][37][38][39]41,48,49] demonstrated an overall positive correlation between IC/nIC and the degree of differentiation, when the ROI was placed within the tumor mass as shown in Tables 1 and 2. These studies had an overall low risk of bias ( Figure S1). Only one study by Xie et al. [41] found no significant difference in the degree of differentiation, with p-values of 0.06, 0.07 and 0.09 in nIC-A, nIC-V and nIV-D, respectively. Aside from the IC measurements in perigastric adipose tissue, several studies have also suggested that IC measurements in lymph nodes may be used to discriminate metastatic from non-metastatic lymph nodes relating to gastric as well as colorectal cancers [36,64]. For gastric cancers, the evidence suggests that IQ with ROI placement on solid tumor correlates well with degree of differentiation, while IC measurements with ROI placement in the perigastric adipose tissue correlates with serosal invasion [40,42].
Four studies [42,45,50,59] investigated the correlation between IC measurements with pre-and post-chemo or chemoradiotherapy treatment (three studies = neoadjuvant, one study = curative). IC was correlated to pathological findings, and all studies found a positive correlation between IC measurements and pathological response or non-response, suggesting that IQ may be used in monitoring treatment response. These findings stand in contrast to a study by Mazzei et al. [65], which identified no significant differences between tumor HU attenuation and different tumor regression grades among gastric cancer patients prior to and following neoadjuvant chemotherapy. Similarly, IQ may also aid in the monitoring of disease activity in the relatively homogenous group of Crohn's disease patients. Three studies [30,56,57], with a relatively low risk of bias ( Figure S1), found significant correlation between IC measurements and various validated methods of evaluating disease activity such as CDAI, endoscopy and biochemistry as shown in Table 3.
An inconsistency, aside from the varying study focuses, seems to be the chosen outcome measures. Of the 30 studies, 12 studies included only IC measurements, 12 studies included only nIC measurements, and eight studies included IC as well as nIC measurements. None of the included studies detailed the rationale behind including or excluding nIC as an outcome measure. The applications and benefits of IC normalization remain unclear. Logically, nIC may reduce variability in cases of varying contrast administration times, flow rates and varying iodine contrast medium concentrations, as the measurement in tissue is referenced to a highly iodinated structure (e.g., the aorta). One study by Patel et al. [66] focused on vascular vs. non-vascular renal lesions and whether normalization of IC to the aorta could reduce inter-manufacturer threshold variability. The study found that nIC reduces the inter-manufacturer variability in IC measurements with no significant inter-manufacturer difference in nIC (p > 0.05), but a significant difference in absolute IC (p < 0.05). However, the study did not assess the effects of nIC on minimizing patient (e.g., reduced cardiac output) or technical variabilities (i.e., timing of contrast medium administration, total iodine, flow rate).
Additionally, IC normalization reference points varied among included studies, with the most common reference being the aorta. However, multiple studies have used other arteries such as the external iliac artery [53,55] as well as the psoas muscle as reference [48], making comparison of nIC between studies a challenge.
IQ is only one of several possible reconstructive measures in DECT image acquisition. Alongside IQ generated through the ROI placement, several other quantitative parameters may be measured at the same time, such as Z-effective number and the slope of the spectral HU curve. In the scope of this review, these parameters should in theory not differ significantly from each other, as they are all directly or indirectly an expression of the IC within the ROI. This assumption aligned well with the outcomes of all studies including Z-effective and/or slope of the HU curve (n = 10). Of these, only one study, by Al-Najami et al. [51] with a sample size of 16 patients, found a significant difference in the effective Z number, but not in the IC for the differentiation of malignant vs. benign rectal tumors.
Several potential clinical implications of IQ have been suggested. In the case of gastric and colon cancers, studies have suggested that IQ may improve the preoperative diagnosis and evaluation [34,38]. Chuang-bo et al. [49] reported increased sensitivity and specificity for well-differentiated vs. poorly differentiated carcinoma using IQ, when compared with conventional CT images at 70 KeV (p < 0.05). Additionally, several studies have suggested that IQ may be an alternative method to evaluate chemoor chemoradiotherapy treatment response [50]. Ge et al. [59] reported a significant difference in nIC (p < 0.05) when comparing the effective treatment group with the ineffective group using RECIST as reference. A study by Uhrig et al. [67] suggested that IQ using DECT may be a complementary method to RECIST, as RECIST only accounts for size reduction of the tumor due to the cytotoxic effects of chemotherapy. Targeted therapies interfere with various biological pathways and may cause tumor necrosis or hemorrhage [68,69], rendering an underestimation of the treatment response according to RECIST criteria. Compared with conventional CT, IQ using DECT overcomes these drawbacks, as IC measurements are not affected by tissue modifications, such as necrosis or hemorrhage, but are purely a representation of vascularized tumor tissue. Other modalities such as magnetic resonance imaging (MRI) and fluorodeoxyglucose positron emission tomography (FDG-PET) have proven to be beneficial in monitoring treatment response [70][71][72]; however, there is no established golden standard, and the examination costs associated with DECT are substantially lower compared to MRI and FDG-PET.
CT colonography also seems to benefit from IQ and DECT image acquisition improving diagnostic accuracy for colorectal cancer screening [52] when comparing conventional CT colonography with DECT colonography. Lastly, IQ may be a convenient and reproducible measure in the evaluation of Crohn's disease [30], with a study by Kim et al. [54] reporting significant correlation (r = 0.74) between IC and CDAI. However, in the case of Crohn's disease, the patient age and radiation dosage should be considered.
IQ by means of ROI placement has its limitations. In the case of acute bowel ischemia, ROI placement may be challenging in times of bowel wall thinning due to, e.g., arterial occlusion [73], and mucosal enhancements in arterial occlusive and non-occlusive ischemia are different in the presence or absence of a reperfusion event [74]. In certain cases, qualitative reconstructions such as virtual monoenergetic or iodine mapping have shown to increase conspicuity and confidence in the diagnosis of acute bowel ischemia [62,75].
There are several limitations to this study. First, there is an obvious heterogeneity in the studies, preventing meta-analysis in comparing IQ with specific correlation measures. Among the different subgroups, the objectives of the studies varied vastly from differentiating between cancer subtypes to evaluating the response following neoadjuvant chemotherapy. Second, the median patient sample size is quite small considering the heterogeneity of the studies. Several studies included no more than 20 patients, making statistical significance questionable.
Third, there are considerable inconsistencies in terms of DECT image acquisition parameters across studies. Variables include different concentrations of iodinated contrast agents, contrast flow rates, total iodine, phase times for all three phases as well as different CT scanners with different postprocessing software, which has shown to have varying accuracies in regard to iodine measurements [76]. These inconsistencies may also be troublesome for future meta-analyses, as the outcome measure relies on an input measure, which for various reasons will vary.
Future studies in this field should include larger sample sizes to decrease the margin of error. In addition, the potential applications of DECT and IQ relating to small bowel pathologies such as bowel ischemia are poorly elucidated and warrant further investigation. Additionally, the use of nIC vs. IC should be investigated further to assess whether nIC provides benefits in terms of reducing variability among patients with decreased cardiac output, varying technical factors such as flow rate, total iodine and varying phase times.

Conclusions
Despite the heterogeneity of this systematic review, certain applications within the GI tract are better elucidated than others. Some of the promising applications of IQ include differentiating between gastric cancers with and without serosal invasion, identifying the degree of differentiation in adenocarcinomas, monitoring of chemo-or chemoradiotherapy treatment response and Crohn's disease activity.

Conflicts of Interest:
The authors declare no conflict of interest.