Non-contrast liver MRI as an alternative to gadoxetic acid-enhanced MRI for liver metastasis from colorectal cancer

Abstract

Background

Liver magnetic resonance imaging (MRI) provides reliable diagnostic performance for detecting liver metastasis but is costly and time-consuming.

Purpose

To compare the diagnostic performance of non-contrast liver MRI to whole MRI using gadoxetic acid for detecting liver metastasis in patients with colorectal cancer (CRC).

Material and Methods

We included 175 patients with histologically confirmed 401 liver metastases and 73 benign liver lesions. A non-contrast MRI (T1-weighted, T2-weighted, and diffusion-weighted images) with or without multidetector computed tomography (MDCT) and a whole MRI (gadoxetic acid-enhanced and non-contrast MRI) were analyzed independently by two observers to detect liver metastasis using receiver operating characteristic analysis.

Results

We found no significant differences in Az value (range = 0.914–0.997), sensitivity (range = 95.2–99.6%), specificity (range = 77.3–100%), or positive (range = 92.9–100%) or negative predictive value (range = 87.5–95.7%) between the non-contrast MRI with or without MDCT and the whole MRI for both observers for all lesions as well as lesions ≤1.0 cm and lesions >1.0 cm in size (P = 0.203–1.000). Combined MDCT and non-contrast MRI led to similar numbers of false-positive diagnosis to the whole MRI (eight for Observers 1 and 4 vs. 3 for Observer 2).

Conclusion

Non-contrast liver MRI may serve as an alternative to gadoxetic acid-enhanced MRI for detecting and characterizing liver metastasis from CRC, at least in patients with relatively high risk of liver metastasis who underwent MDCT. Non-contrast liver MRI could be beneficial especially for patients with lesions that are already documented as benign but require additional follow-up MRIs.

Keywords

Liver metastases colorectal cancer MRI contrast agents intravenous

Introduction

Colorectal cancer (CRC) is one of the most common cancers and leading causes of death worldwide (1). About 30% of patients diagnosed with CRC will develop liver metastases in up to 19% of patients synchronously and 14–20% of patients metachronously during the subsequent five years after initial diagnosis (2 –5). Therefore, accurate detection and characterization of hepatic metastasis are essential for planning an overall surgical strategy for patients with CRC (6). Magnetic resonance imaging (MRI) is generally accepted to be better than computed tomography (CT) for diagnosis of liver metastasis (7). Nevertheless, the American College of Radiology Appropriateness Criteria for the pretreatment staging of CRC recommends chest, abdomen, and pelvis CT for initial evaluation of disease. Throughput and high cost hinder the introduction of liver MRI into routine protocols, particularly where rectal MRI is routinely performed.

Diffusion-weighted imaging (DWI) has become a routine protocol for liver MRI (8 –10). Prior studies have shown the feasibility of liver DWI as a routine part of a combined MR evaluation of patients with rectal cancers by demonstrating the superiority of DWI over CT for detection of liver metastasis (9,10). DWI is limited for lesion characterization due to substantial overlap in ADC values for a variety of hepatic tumors (11). Given that differentiation between metastasis and cysts or hemangiomas is a major concern (10), T1-weighted (T1W) and T2-weighted (T2W) imaging could complement DWI for lesion characterization (12,13).

Gadoxetic acid (Gd-EOB-DTPA; Primovist, Bayer-Schering HealthCare AG, Berlin, Germany) is widely employed hepatocyte-specific MR contrast agent due to its high sensitivity for identifying hepatic malignancies with hepatobiliary phase (HBP) (14,15). We conducted this study to compare the diagnostic performance of non-contrast liver MRI including T1W imaging, T2W imaging, and DWI to whole MRI using gadoxetic acid for detecting liver metastasis in patients with CRC. Thereby, we wanted to investigate whether non-contrast MRI could be an alternative to whole MRI for detecting liver metastasis.

Material and Methods

Patients

Our institutional database was reviewed retrospectively for liver MRI of patients with CRC at our tertiary referral hospital between January 2012 and April 2016. During this period, 381 consecutively registered patients suspected of having liver metastases from CRC underwent liver MRI. Inclusion criteria for the patient group were: (i) histologically proven focal liver lesions; (ii) liver metastasis with <10 lesions; and (iii) no local ablation treatment for liver metastasis. We excluded 161 patients: 133 with no histology proof; 21 with too many nodules (>10) to be analyzed; and seven with previous local ablation treatment. The remaining 175 patients formed the final study group. The detailed selection process is shown in Fig. 1.

Fig. 1.

Flow chart of the study population.

Lesion confirmation-reference standard

The reference standard for liver metastases was based on histopathological examination of surgical specimens from all 170 patients. Of the 170, 36 also had 65 benign liver lesions including 45 that were surgically proven and the other 20 hepatic cysts that were assigned by reviewers. In total, 155 patients simultaneously underwent hepatic resection and primary CRC resection within 18 days after MRI examination. Fifteen patients received chemotherapy before liver resection (range of time interval between MRI and surgery = 6–18 months). Of five patients with no liver metastasis, one with peliosis underwent hepatic resection, and four with a total of seven inflammatory lesions (one with four lesions and three with one lesion each) underwent image-guided biopsy. The seven inflammatory lesions disappeared on follow up CT or MRI without treatment. All patients underwent follow-up CT or MRI for at least three months (range = 3–76 months).

Imaging techniques

Two phases of CT, unenhanced and portal venous phase, were conducted with a 40-multidetector CT (MDCT) scanner (Brilliance 40; Philips Healthcare, Best, The Netherlands) or a 64-MDCT scanner (Aquilion 64; Toshiba Medical Systems, Japan and LightSpeed VCT 64 GE Healthcare, Milwaukee, WI, USA). Portal venous phase scanning began 70 s after injection of 110 or 120 mL non-ionic iodinated contrast agent iopamidol (Iopamiro 300, Bracco, Milano, Italy) at 3–4 mL/s using a bolus-triggered technique (Suppl. Table 1).

All MRIs were acquired using a 3.0-T MR system (InteraAchieva 3.0-T, Philips Healthcare, Best, The Netherlands) with a 16-channel phased-array coil. Baseline MRI included T1W turbo field-echo in-phase and opposed sequence and breath-hold multishot T2W imaging. DWIs were acquired using respiratory-triggered single-shot echo planar imaging with a b-value of 0, 100, or 800 s/mm². The apparent diffusion coefficient (ADC) was calculated by a monoexponential function using b-values of 100 and 800 s/mm².

For gadoxetic acid-enhanced imaging, unenhanced, arterial-phase (20–35 s), portal-phase (60 s), late-phase (3 min), and 20-minute HBP were obtained using a T1W 3D turbo-field-echo sequence (T1 high-resolution isotropic volume examination, eTHRIVE, Philips Healthcare). Using a power injector, contrast agent was administered intravenously at 2 mL/s at a dose of 0.025 mmol/kg body weight, followed by a 20-mL saline flush. The detailed MR sequence parameters used are shown in Suppl. Table 2.

Image analysis

Two gastrointestinal radiologists (YKK and JAH, with 16 and two years of experience, respectively) reviewed images on a picture archiving and communication system (Centricity Radiology RA 1000; GE Healthcare, Chicago, IL, USA). The reviewers were aware of the overall study goal, but were unaware of information for liver lesions

Image review consisted of two sessions with a four-week interval. The non-contrast MRI set (T1W and T2W imaging plus DWI) was presented randomly and blinded at the first reading session. After reviewing the non-contrast MRI set, reviewers were also asked to read MDCTs. In the second session, the whole MRI set (gadoxetic acid-enhanced imaging plus non-contrast MR set) was read in a non-selected order.

The observers recorded the possibility of liver metastasis for each lesion by assigning following confidence level: 1 = probably not a metastasis; 2 = possibly not a metastasis; 3 = probably a metastasis; and 4 = definitely a metastasis. Lesions not assigned on review were rated 0. Reviewers were asked to not assign definitive cysts because inclusion of too many cysts would be difficult to analyze and influence statistical power. Thus, 31 hepatic cysts assigned by reviewers were included into data analysis. The diagnostic criteria for liver metastasis in the non-contrast MRI were defined as nodule with T1W imaging hypointensity, moderate hyperintensity on T2W imaging, and/or diffusion restriction on DWI (16). The diagnostic criteria for metastases in the whole MRI were nodule showing hypointensity on the HBP with poor enhancement centrally and increased peripheral rim enhancement during the early dynamic phase (also for MDCT imaging) plus the criteria used for the non-contrast MRI. As an ancillary finding of metastasis, observers also determined the presence of a target appearance of a central hypointense area and a peripheral hyperintense rim on DWI (17) or central enhancement and peripheral hypointense rim on HBP (18). Readers also characterized hepatic lesions using imaging criteria (19 –21).

Statistical analysis

Statistical analysis was executed using SAS version 9.4 (SAS Institute, Cary, NC, USA) and R 3.0.3 (Vienna, Austria; http://www.R-project.org/). Alternative-free response receiver operating characteristic (ROC) curve analysis was performed on a lesion-by-lesion basis (22). Diagnostic accuracy of each imaging set was assessed by calculating the area under the ROC curve (Az), with pairwise comparisons performed between imaging sets for each observer using the variance z-test. Non-parametric analysis of clustered ROC curve data was performed, considering the possible influence of lesion clustering on diagnostic accuracy and using the method of Obuchowski (23). Using 401 liver metastases and 73 benign lesions, sensitivity for detecting metastasis was evaluated according to number of metastases assigned a confidence level of 3 or 4.

Specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for each image set. We calculated 95% confidence intervals using Wilson’s method for addressing reliability of the results obtained from a sample (24). Bennett’s test was used to compare sensitivity, specificity, PPV, and NPV of image sets, and Chi-square test was used to compare the accuracy of two image sets. Bonferroni correction was used to adjust P values for multiple comparisons. P < 0.05 was considered significant. Inter-reviewer agreement for assessing MRI was analyzed using kappa statistics, with 0.8–1.0 considered almost perfect agreement; 0.6–0.79, substantial; 0.40–0.59, moderate; 0.2–0.39, fair; 0–0.19, slight; and 0–1.0, poor.

Results

Patient characteristics

A total of 401 metastases (size range = 0.4–12.2 cm; mean = 1.8 cm) were identified in the 170 patients: 74 patients had a solitary lesion and the remaining 96 had multiple lesions (Table 1). Of 170 patients, 36 had 65 benign lesions; thus, 41 patients, including five with no liver metastasis had 73 benign liver lesions (size range = 0.4–3.8 cm; mean = 1.1 cm).

Table 1.

Characteristics of study groups and lesions.

Characteristics	Patients with metastases (n = 170)	Patients with only benign lesions (n = 5)
Age, years (mean ± SD)	59.2 ± 9.6	61.8 ± 9.7
Male/female	110/60	5 /0
Lesion size (cm)	1.8 ± 1.3 (0.4–12.2)	1.0 ± 0.3 (0.5–15)
Size subgroup (cm)
≤1.0	170 [124/46]	5
>1.0	296 [277/19]	3
Lesions per patient (n)	Range 1–8 /mean 2.4	Range 1–4/mean 1.6
Category of lesions	n = 466	n = 8
Metastasis	401*	0
Hemangioma	22	0
Inflammatory nodule	8	3
Eosinophilic abscess	0	4
Focal nodular hyperplasia	3	0
Regenerative nodule	1	0
Peliosis	0	1
Cyst	31	0
Diagnosis
Operation	401 metastases 45 benign lesions	Peliosis (n = 1)
Percutaneous biopsy	0	Inflammatory nodules (n = 3) Eosinophilic abscesses (n = 1)
Imaging follow-up	20 cysts	3 Eosinophilic abscesses

Numbers in brackets represent number of metastases and benign lesions.

Five patients had mucinous adenocarcinoma.

ROC analysis

ROC curves for comparison of diagnostic accuracy of each imaging set are shown in Suppl. Fig. 1. Non-parametric analysis of clustered ROC data showed that the Az for non-contrast MRI (0.956 for Observer 1 and 0.975 for Observer 2) was slightly lower than those of whole MRI (0.975 and 0.985) (Table 2). The difference was not significant (P = 0.665 and 1.000) (Suppl. Table 3). Results were also not significant for 175 lesions ≤1.0 cm (P = 1.000) and 299 lesions >1.0 cm (P = 1.000) (Fig. 2). Adding the MDCT reading to non-contrast MRI did not significantly increase Az compared with non-contrast MRI (P = 1.000).

Table 2.

Accuracies (Az), sensitivities, specificities, and positive (PPV) and negative predictive values (NPV) for three image sets for detection of 401 liver metastases.

	Accuracy	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Reviewer 1
Non-contrast MR set	0.956/0.933–0.971	98.3 (394)/96.4–99.2	80.8/70.3–88.2	96.6 [14]/94.3–97.6	89.4/79.7–94.8
Combined set	0.968/0.948–0.981	98.3 (394)/96.4–99.2	89.0/79.8–94.3	98.0 [8]/96.1–99.0	90.3/81.2–95.2
Whole MR set	0.975/0.956–0.986	99.0 (397)/97.5–99.6	89.0/79.8–94.3	98.0 [8]/96.2–99.0	94.2/86.0–97.7
Reviewer 2
Non-contrast MR set	0.975/0.956–0.986	98.8 (396)/97.1–99.5	90.4/81.5–95.3	98.3 [7]/96.5–99.2	93.0/94.6–97.0
Combined set	0.981/0.964–0.990	98.8 (396)/97.1–99.5	94.5/86.7–97.9	99.0 [4]/97.5–99.6	93.2/85.1–97.1
Whole MR set	0.985/0.970–0.993	99.0 (397)/97.5–99.6	95.9/88.6–98.6	99.3 [3]/97.8–99.7	94.6/86.9–97.9

Data are presented as value/95% confidence interval. Numbers in parentheses are number of true-positive lesions. Numbers in brackets are false-positive lesions. Combined set, non-contrast MR set plus MDCT images.

Fig. 2.

Two surgically confirmed liver metastases from a rectal adenocarcinoma from a 59-year-old man. (a) Axial breath-hold T2W image and (b) single-shot echo-planar DWI, b = 800 s/mm² with hyperintense metastasis in segment 8 (arrows). Larger metastasis (black arrows) partly seen in segment 6. (c) Corresponding ADC map with lesion signal intensity lower than adjacent liver (arrow). (d) Axial portal venous phase, multidetector row CT image, with single liver mass in segment 6 (black arrow). (e) Axial arterial phase image after gadoxetic acid with rim enhancement of small liver metastasis in segment 8 (arrow).

Sensitivity and negative predictive value

A trend toward increased sensitivity was observed with whole MRI (99.0% for both observers) compared with either non-contrast MRI or combined MDCT and non-contrast MRI (98.3% for Observer 1 and 98.8% for Observer 2 in both images). However, no significant differences were observed between image sets (P = 0.499 and 1.000) (Table 2). We found the same tendency when analyzing sensitivity for lesions ≤1.0 cm (P = 0.499 and 1.000) (Table 3) or lesions >1.0 cm (P = 1.000) (Table 4). NPVs showed no significant differences among the three image sets for all lesions or in analysis according to lesion size (P = 0.374 or 0.504 or 1.000). Five metastases were not detected by the reviewers with non-contrast MRI; one was not depicted on retrospective review and four (size range = 5–11 mm) were not verified by either observer on non-contrast MRI with or without MDCT or whole MRI. Three metastases were ≤10 mm in diameter and were missed (score 0) or misclassified as benign (score 2). The remaining metastasis (11 mm) originated from mucinous adenocarcinoma (Fig. 3). There were 18 metastases (<1.5 cm) depicted on non-contrast MRI but not clearly delineated on MDCT on review (Fig. 2). No metastases were depicted on MDCT but not on non-contrast MRI.

Table 3.

Accuracies (Az), sensitivities, specificities, and positive (PPV) and negative predictive values (NPV) for three image sets for detection of 124 liver metastases ≤1.0 cm in diameter.

	Accuracy	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Reviewer 1
Non-contrast MR set	0.914/0.863–0.947	95.2 (118)/89.8–97.8	82.4/69.8–90.4	92.9 [9]/87.1–96.2	87.5/75.3–94.1
Combined set	0.926/0.877–0.956	95.2 (118)/89.8–97.8	86.3/74.3–93.2	94.4 [7]/88.9–97.3	88.0/76.2–94.4
Whole MR set	0.943/0.898–0.969	97.6 (121)/93.1–99.2	86.3/74.3–93.2	94.5 [7]/89.1–97.3	93.6/82.8–97.8
Reviewer 2
Non-contrast MR set	0.943/0.898–0.969	96.8 (120)/92.0–98.7	88.2/76.6–94.5	95.2 [6]/90.0–97.8	91.8/80.8–96.8
Combined set	0.954/0.912–0.977	96.8 (120)/92.0–98.7	92.2/91.5–96.9	96.8 [4]/92.0–98.7	92.2/81.5–96.9
Whole MR set	0.966/0.927–0.984	97.6 (121)/93.1–99.2	94.1/84.1–98.0	97.6 [3]/93.1–99.2	94.1/84.1–98.0

Data are presented as value/95% confidence interval. Combined set, non-contrast MR set plus MDCT images. Numbers in parentheses are number of true-positive lesions. Numbers in brackets are false-positive lesions.

Table 4.

Accuracies (Az), sensitivities, specificities, and positive (PPV) and negative predictive values (NPV) for three image sets for detection of 277 liver metastases >1.0 cm in diameter.

	Accuracy	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
Reviewer 1
Non-contrast MR set	0.980/0.957–0.991	99.6 (276)/98.0–99.9	77.3/56.6–89.9	98.2 [5]/95.9–99.2	94.4/74.2–99.0
Combined set	0.993/0.976–0.998	99.6 (276)/98.0–99.9	95.5/78.2–99.2	99.6 [1]/98.0–99.9	95.5/78.2–99.2
Whole MR set	0.993/0.976–0.998	99.6 (276)/98.0–99.9	95.5/78.2–99.2	99.6 [1]/98.0–99.9	95.5/78.2–99.2
Reviewer 2
Non-contrast MR set	0.993/0.976–0.998	99.6 (276)/98.0–99.9	95.5/78.2–99.2	99.6 [1]/98.0–99.9	95.5/78.2–99.2
Combined set	0.997/0.981–0.999	99.6 (276)/98.0–99.9	100.0/85.1–100.0	100.0 [0]/98.6–100.0	95.7/79.0–99.2
Whole MR set	0.997/0.981–0.999	99.6 (276)/98.0–99.9	100.0/85.1–100.0	100.0 [0]/98.6–100.0	95.7/79.0–99.2

Fig. 3.

Single surgically confirmed liver metastasis from mucinous adenocarcinoma of the sigmoid colon in a 44-year-old woman. (a) Axial breath-hold T2W image of a small, bright, hyperintense nodule (arrow). (b) ADC map with signal intensity of a lesion higher than that of the adjacent liver (arrow). (c) Axial portal venous phase, multidetector row CT image with a small hypoattenuated nodule with no enhancement (arrow). (d) Axial portal venous phase images after gadoxetic acid with no lesion enhancement. Lesion was scored category 1 or 2 by both observers in three image sets.

Specificity and positive predictive value

We found no significant differences in specificity or PPV between the two MRIs (P = 0.499 and 0.485 for Observer 1; P = 0.273 and 0.277 for Observer 2). Adding MDCT to non-contrast MRI led to values of specificity and PPV similar to those of the whole MRI (P = 1.000). Seven and three benign lesions misclassified by each reviewer with the non-contrast MRI, respectively, were accurately characterized by adding MDCT (Fig. 4). We found the same tendency when analyzing lesions ≤ 1.0 cm.

Fig. 4.

Hepatic hemangioma in a 61-year-old man with sigmoid colon cancer. (a) Axial breath-hold T2W image and (b) single-shot echo-planar DWI, b = 800 s/mm² with hyperintense lesion (arrows). This lesion was categorized as metastasis by both observers in the non-contrast MR set. (c) Axial portal venous phase, multidetector row CT image with homogeneous lesion enhancement, indicating hemangioma. Adding CT to the non-contrast MR set led to correct diagnosis of hemangioma by both observers. (d) Axial arterial phase MR image after gadoxetic acid of a liver lesion with homogeneous enhancement, indicating hemangioma (arrow).

The false positives were highest in the non-contrast MRI (14 and seven lesions for the observers), followed by the combined set (eight and four) and then the whole MRI (eight and three). Of 19 false-positive lesions assigned by both observers, eight were considered to be inflammatory nodules (Suppl. Table 4): One was an eosinophilic abscess, eight were hemangiomas (size range = 0.6–2.0 cm) including one sclerosing hemangioma, and one each was focal nodular hyperplasia and regenerative nodule.

Kappa values for the two observers were 0.805 for the non-contrast MRI, 0.857 for the combined MDCT and non-contrast MRI, and 0.755 for whole MRI, indicating almost perfect or substantial inter-observer agreement. Target appearance on DWI was observed in 346 metastases (86.3%) (Suppl. Fig. 2). Target appearance on HBP was observed in 207 metastases (51.6%).

Discussion

In this study, no statistically significant differences were observed among all five diagnostic values measured for three image sets by two observers (P > 0.05), although a trend toward higher values was observed with the whole MRI compared to the non-contrast MRI. Thus, our findings indicate that non-contrast liver MRI could be an alternative to whole MRI using gadoxetic acid for detecting and characterizing liver metastasis, at least in a preselected cohort of patients with a comparably high fraction of liver metastasis.

Although studies have shown mixed results about DWI performance relative to contrast-enhanced MRI for metastasis detection (21,25 –27), combining both is generally accepted to be better than each modality alone (14,17,28). In view of this assumption, our results are somewhat surprising. Since lesion detection with DWI is easy, recommendations are to first interpret DWI to pinpoint lesions and then use contrast-enhanced images for characterization (17), implying that correctly characterizing detected lesions is more challenging than detecting lesions with a non-contrast MRI set. Accordingly, of five metastases not detected by both reviewers with the non-contrast MRI set, only one was not depicted on retrospective review and the other four were misclassified. We found no benefit for tumor detection when we added MDCT to a non-contrast MRI. Eighteen small metastases depicted on the non-contrast MRI were not clearly delineated on MDCT on review.

On retrospective review, two mucinous metastases misclassified by both reviewers in the non-contrast MRI set were bright on T2W imaging and ADC maps, suggesting hemangiomas or cysts. The observers misclassified 14 and seven benign lesions in the non-contrast MRI set as metastases. Most were hemangioma (n = 8) or non-specific inflammation (n = 5). Thus, although 346 metastases (86.3%) showed a target appearance on DWI, indicating metastasis, we acknowledge the limitations of T2W imaging in differentiating between solid and non-solid lesions with the naked eye and DWI for tumor characterization. Seven and three benign lesions misclassified by both reviewers in the non-contrast MRI set were accurately characterized by adding MDCT. These results led to similar specificities for the combined set and the whole MRI set, supporting the use of non-contrast MRI.

We are aware that not all clinicians agree with our strategy because liver MRI application varies by institution. In a recent randomized trial comparing diagnostic performance and efficacy of liver imaging as a first-line imaging method in patients with suspected CRC liver metastasis, gadoxetic acid-enhanced MRI was superior to MRI with extracellular contrast medium and contrast-enhanced CT (29). We agree that contrast enhanced MRI especially using gadoxetic acid is more recommendable as a first-line imaging method, especially if highly exact hepatic staging and anatomical information are needed. Setting a range of patients that can be applied to non-contrast MRI is a major concern. Considering the high cost and lengthy examination time of MRI using hepatocyte agents, our strategy seems acceptable, particularly in situations where rectal MRI is routinely performed. Given that the NPV for CT for detecting hepatic metastasis from CRC by patient- or lesion-based analysis is up to 98.4% (10,28,30,31), active implementation of non-contrast MRI is recommendable when CT findings are inconclusive or suspicious of liver metastasis as our study showed the best performance of non-contrast MRI especially in small liver metastasis. In addition, in the clinical setting with follow-up after treatment of CRC, in which newly developed liver lesions are more likely to be metastasis than benignity such as hemangioma, non-contrast MRI is recommendable as characterization of hemangioma would be no more challenging in that situation. One concern is variability in DWI quality across MR systems with different equipment or field strengths. The 3.0-T platform with applying dual-source RF transmission, as used in our study, offers high signal-to-noise ratios with improved B1 uniformity over high-field MR systems.

This study had several limitations. First, the retrospective nature may have introduced inherent selection bias. Our patient cohort may have been biased by the large number of metastases and the relatively small number of benign cases. Thus, standard statistic parameters might have been influenced by a low number of benign cases. However, given that MRI is usually used as a problem-solving tool after using CT, our study group might be representative of clinical practice. Second, since part of the study populations received chemotherapy before liver surgery, small occult metastases might have not been included in study samples. Third, our data came from a single institution and might not be exportable to other centers with different MR imagers. Fourth, although black blood imaging of DWI provides hepatic vascular anatomy at a certain degree, non-contrast MRI protocol has limitations in assessment of the topography and anatomy for surgical planning.

In conclusion, non-contrast MRI including DWI in combination with MDCT is a reasonable strategy with high sensitivity for detecting colorectal liver metastases. Thus, non-contrast liver MRI may serve as an alternative to gadoxetic acid-enhanced MRI for detecting and characterizing liver metastasis from CRC, at least in patients with relatively high risk of liver metastasis who underwent MDCT. Non-contrast liver MRI could be beneficial especially for patients with lesions that are already documented as benign but require additional follow-up MRIs. Gadoxetic acid-enhanced MRI achieved slightly higher sensitivity and specificity although the difference was not statistically significant; however, it may still be necessary under circumstances that each and every lesion must be precisely identified (such as preoperative assessment).

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Jemal

Bray

Center

et al.

Global cancer statistics. CA Cancer J Clin 2011; 61: 69–90.

Manfredi

Lepage

Hatem

et al.

Epidemiology and management of liver metastases from colorectal cancer. Ann Surg 2006; 244: 254–259.

Ghiringhelli

Hennequin

Drouillard

et al.

Epidemiology and prognosis of synchronous and metachronous colon cancer metastases: a French population-based study. Dig Liver Dis 2014; 46: 854–858.

Landreau

Drouillard

Launoy

et al.

Incidence and survival in late liver metastases of colorectal cancer. J Gastroenterol Hepatol 2015; 30: 82–85.

Leporrier

Maurel

Chiche

et al.

A population-based study of the incidence, management and prognosis of hepatic metastases from colorectal cancer. Br J Surg 2006; 93: 465–474.

Harmon

Ryan

Jr Biehl

et al.

Benefits and safety of hepatic resection for colorectal metastases. Am J Surg 1999; 177: 402–404.

Tsurusaki

Sofue

Murakami

. Current evidence for the diagnostic value of gadoxetic acid-enhanced magnetic resonance imaging for liver metastasis. Hepatol Res 2016; 46: 853–861.

Parikh

Drew

Lee

et al.

Focal liver lesion detection and characterization with diffusion-weighted MR imaging: comparison with standard breath-hold T2-weighted imaging. Radiology 2008; 246: 812–822.

Achiam

Logager

Skjoldbye

et al.

Preoperative CT versus diffusion weighted magnetic resonance imaging of the liver in patients with rectal cancer; a prospective randomized trial. PeerJ 2016; 4: e1532–e1532.

10.

Kwon

Kim

Hwang

et al.

Efficiency of non-contrast-enhanced liver imaging sequences added to initial rectal MRI in rectal cancer patients. PLoS One 2015; 10: e0137320–e0137320.

11.

Taouli

. Diffusion-weighted MR imaging for liver lesion characterization: a critical look. Radiology 2012; 262: 378–380.

12.

Kondo

Kanematsu

Itoh

et al.

Does T2-weighted MR imaging improve preoperative detection of malignant hepatic tumors? Observer performance study in 49 surgically proven cases. Magn Reson Imaging 2005; 23: 89–95.

13.

Lim

Kim

Lee

et al.

Follow-up result afters negative findings on unenhanced hepatic MR imaging for hepatic metastasis from rectal cancer. Korean J Radiol 2004; 5: 225–230.

14.

Kim

Lee

Byun

et al.

Incremental value of liver MR imaging in patients with potentially curable colorectal hepatic metastasis detected at CT: a prospective comparison of diffusion-weighted imaging, gadoxetic acid-enhanced MR imaging, and a combination of both MR techniques. Radiology 2015; 274: 712–722.

15.

Lee

Park

et al.

MR imaging in patients with suspected liver metastases: value of liver-specific contrast agent gadoxetic acid. Korean J Radiol 2013; 14: 894–904.

16.

Taouli

Koh

. Diffusion-weighted MR imaging of the liver. Radiology 2010; 254: 47–66.

17.

Kim

Lee

et al.

Diagnostic accuracy and sensitivity of diffusion-weighted and of gadoxetic acid-enhanced 3-T MR imaging alone or in combination in the detection of small liver metastasis (</= 1.5 cm in diameter). Invest Radiol 2012; 47: 159–166.

18.

Kim

Lee

Kim

et al.

Gadoxetic acid-enhanced 3.0T MRI for the evaluation of hepatic metastasis from colorectal cancer: metastasis is not always seen as a “defect” on the hepatobiliary phase. Eur J Radiol 2012; 81: 3998–4004.

19.

Zech

Herrmann

Reiser

et al.

MR imaging in patients with suspected liver metastases: value of liver-specific contrast agent Gd-EOB-DTPA. Magn Reson Med Sci 2007; 6: 43–52.

20.

Goshima

Kanematsu

Watanabe

et al.

Hepatic hemangioma and metastasis: differentiation with gadoxetate disodium-enhanced 3-T MRI. Am J Roentgenol 2010; 195: 941–946.

21.

Lowenthal

Zeile

Lim

et al.

Detection and characterisation of focal liver lesions in colorectal carcinoma patients: comparison of diffusion-weighted and Gd-EOB-DTPA enhanced MR imaging. Eur Radiol 2011; 21: 832–840.

22.

Chakraborty

Winter

. Free-response methodology: alternate analysis and a new observer-performance experiment. Radiology 1990; 174: 873–881.

23.

Obuchowski

. Nonparametric analysis of clustered ROC curve data. Biometrics 1997; 53: 567–578.

24.

Newcombe

. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998; 17: 857–872.

25.

Chung

Kim

Chung

et al.

Comparison of gadoxetic acid-enhanced dynamic imaging and diffusion-weighted imaging for the preoperative evaluation of colorectal liver metastases. J Magn Reson Imaging 2011; 34: 345–353.

26.

Macera

Lario

Petracchini

et al.

Staging of colorectal liver metastases after preoperative chemotherapy. Diffusion-weighted imaging in combination with Gd-EOB-DTPA MRI sequences increases sensitivity and diagnostic accuracy. Eur Radiol 2013; 23: 739–747.

27.

Shimada

Isoda

Hirokawa

et al.

Comparison of gadolinium-EOB-DTPA-enhanced and diffusion-weighted liver MRI for detection of small hepatic metastases. Eur Radiol 2010; 20: 2690–2698.

28.

Larsen

Rosenkilde

Christensen

et al.

The value of contrast enhanced ultrasonography in detection of liver metastases from colorectal cancer: a prospective double-blinded study. Eur J Radiol 2007; 62: 302–307.

29.

Zech

Korpraphong

Huppertz

et al.

Randomized multicentre trial of gadoxetic acid-enhanced MRI versus conventional MRI or CT in the staging of colorectal cancer liver metastases. Br J Surg 2014; 101: 613–621.

30.

Kukuk

Gieseke

Weber

et al.

Focal liver lesions at 3.0 T: lesion detectability and image quality with T2-weighted imaging by using conventional and dual-source parallel radiofrequency transmission. Radiology 2011; 259: 421–428.

31.

Skrovina

Bartos

Cech

et al.

Intra-operative liver ultrasound–a contribution to colorectal carcinoma staging. Acta Chir Belg 2008; 108: 508–512.