- Email: [email protected]
Diffusion-weighted MRI for the detection of colorectal polyps: feasibility study
Available online at www.sciencedirect.com
Magnetic Resonance Imaging 31 (2013) 28 – 35
Diffusion-weighted MRI for the detection of colorectal polyps: feasibility study☆,☆☆,★,★★ Anke M. Leufkensa , Thomas C. Kwee b,⁎, Maurice A.A.J. van den Boschb , Willem P.Th.M. Malib , Taro Takaharab , Peter D. Siersema a a
Department of Gastroenterology and Hepatology, University Medical Center Utrecht, Utrecht, The Netherlands b Department of Radiology, University Medical Center Utrecht, Utrecht, The Netherlands Received 30 April 2012; accepted 21 June 2012
Abstract The purpose of this study was to determine the feasibility of diffusion-weighted magnetic resonance imaging (DWI) for detecting colorectal polyps. DWI (high b-value of 1000 s/mm 2) was prospectively performed in 26 symptomatic patients who were scheduled to undergo colonoscopy. DWI and colonoscopic findings were interpreted in a blinded manner. The sensitivity and positive predictive value (PPV) of DWI for the detection of clinically relevant polyps (≥ 6 mm) and colorectal cancer (CRC) were calculated on a per-lesion basis, using colonoscopy results as the standard of reference. Sensitivity, specificity, PPV and negative predictive value (NPV) on a per-patient basis were also calculated. Sensitivity and PPV on a per-lesion basis were 80.0% [95% confidence interval (CI): 49.0%–94.3%] and 72.7% (95% CI: 43.4%–90.3%) for polyps ≥ 6 mm and CRC. Sensitivity, specificity, PPV and NPV on a per-patient basis were 85.7% (95% CI: 48.7%–97.4%), 84.2% (95% CI: 62.4%–94.5%), 66.7% (95% CI: 35.4%–87.9%) and 94.1% (95% CI: 73.0%–99.0%) for polyps ≥ 6mm and CRC. In conclusion, DWI cannot yet be recommended in a clinical setting in which DWI is performed first and subsequent colonoscopy is only performed in patients with positive findings at DWI. Further (technical) developments are required to increase its diagnostic yield. © 2013 Elsevier Inc. All rights reserved. Keywords: Colorectal; Polyps; Carcinoma; Magnetic resonance imaging; MRI; Diffusion-weighted
1. Introduction Colorectal cancer (CRC) is the third most common cancer in men and women and the third leading cause of cancer death in the Western world [1]. It has been shown that colonoscopic polypectomy results in a decreased incidence and mortality of CRC [2–4]. Magnetic resonance imaging (MRI) does not use any ionizing radiation and provides a high soft tissue contrast, making it a potentially useful method for the detection or exclusion of colorectal polyps and CRC. Of particular interest is the introduction of diffusion-weighted MRI (DWI) for tumor detection and staging in the body [5,6]. DWI is a functional MRI technique ☆ All authors contributed to study concept/design, data acquisition, data analysis/interpretation and manuscript drafting/editing. ☆☆ All authors approved the final version of the submitted manuscript. ★ All authors guarantee the integrity of the entire study. ★★ None of the authors has any (potential) conflict of interest to disclose. ⁎ Corresponding author. Tel.: +31 88 7556687; fax: +31 30 2581098. E-mail address: [email protected] (T.C. Kwee).
0730-725X/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.mri.2012.06.029
that allows visualization of the random (Brownian) motion of water molecules. As many malignant tumors, including CRC, exhibit an impeded diffusion (most likely due to increased cellularity), they can be highlighted (i.e., high signal intensity) at DWI, while most normal background tissue can be suppressed (i.e., low signal intensity). The high lesion-to-background contrast that can be provided by DWI is an advantage compared to conventional (T1-weighted or T2-weighted) sequences and may even allow for the detection of lesions with a size around or even below that of the voxel size. Furthermore, DWI does not require the administration of any contrast agents and can be performed without any bowel preparation. Several previous studies have indicated the utility of DWI for the detection of CRC [7–9]. However, the evidence provided by these studies is limited due to retrospective inclusion of patients (and controls) and patient selection bias. Furthermore, and most importantly, no previous study has investigated whether DWI allows detection of clinically relevant colorectal polyps (i.e., ≥ 6 mm). If DWI proves to be accurate in detecting
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35
clinically relevant colorectal polyps, it may be used in a clinical setting in which DWI is performed first and subsequent colonoscopy is only performed in those patients with positive DWI findings to confirm the presence of lesions and perform polypectomy. Patients with a negative DWI may be spared a (more invasive, more time-consuming and more expensive) colonoscopic investigation, and this strategy may reduce the increasing pressure on the utilization rate of colonoscopy. The aim of this study was therefore to prospectively determine the feasibility of DWI for the detection of colorectal polyps. 2. Materials and methods 2.1. Patients This study was approval by the local Institutional Review Board, and each patient provided written informed consent prior to inclusion. Included patients prospectively underwent MRI before colonoscopy. Patients were eligible for inclusion if they were aged ≥ 50 years and scheduled to undergo colonoscopy as part of routine clinical care because of symptoms such as rectal bleeding, altered bowel habits, unexplained weight loss, abdominal pain or anemia. Exclusion criteria were incomplete colonoscopy, general contraindications for MRI (including metal objects in the body like brain aneurysm clips or implanted cardiac pacemakers, and claustrophobia), a body size that did not allow for MRI scanning, patients who had undergone bowel preparation, previous abdominal malignancies or melanoma, pregnancy or lactation, previous CRC-directed treatment, and time interval of more than 1 month between MRI and colonoscopy. 2.2. MRI MRI was performed using a 1.5-T system (Achieva, Philips Healthcare, Best, the Netherlands) with a 16-element phased-array surface coil for signal reception (SENSE Torso XL). Applied parameters for DWI were as follows: singleshot spin-echo echo-planar imaging, repetition time (TR)/ echo time (TE)/inversion time of 8551/74/180 ms, axial slice orientation, number of stations of 2, slice thickness/gap of 4/0mm, number of slices of 60 for each of the two stations, overlap of 8 mm between the two stations, field of view of 400×325 mm 2, acquisition matrix of 128×83, receiver bandwidth of 2168.7 Hz/pixel, motion probing gradients in three orthogonal axes, b-values of 0 and 1000 s/mm 2, number of excitations of 4, partial Fourier (half scan) factor of 0.6, parallel acquisition (SENSitivity Encoding) factor of 2, echo train length of 45, acquired voxel size of 3.12×3.92×4.00 mm 3 , reconstructed voxel size of 1.56×1.56×4.00 mm 3, image acquisition under free breathing, and scan time of 11 min and 42 s (5 min and 51 s for each station). In all patients, axial breath-hold T2-weighted (TR/TE of 1000/80 ms, slice thickness/gap of 8/0mm), coronal breath-hold T2-weighted and fat-suppressed T2-
29
weighted (TR/TE of 1258/80 ms, slice thickness/gap of 10/0mm), and coronal in-and-opposed phase T1-weighted [TR/TE of 132/2.3 (opposed-phase) and 4.6 (in-phase) ms, slice thickness/gap of 9/1 mm] were obtained in addition to DWI. All sequences covered the area from the diaphragm to the pubic bone. Total examination time was approximately 20–25 min. No bowel preparation or antiperistaltic agents were administered, and no dietary restrictions were recommended in any of the patients. Using software implemented in the standard operating console, additional coronally reformatted images with a slice thickness/gap of 5/0 mm and coronal maximum intensity projection diffusion-weighted images were created of the axial DWI data set. Diffusion-weighted images were displayed using grayscale inversion. The MRI data set was then transferred to a Picture Archiving and Communications System which allows manual window level setting and systematic evaluation by a board-certified radiologist who was blinded to colonoscopy findings. To that end, high-b-value (1000 s/mm 2) diffusionweighted images were assessed for the probability of the absence or presence of lesions (colorectal polyps or CRC) using the following 5-point grading scale: 1=definitely absent (no signal); 2=probably absent (nonlocalized, mild to Table 1 Characteristics of included patients Overall No. of patients Male/female Mean age (range) Mean no. of days between MRI and colonoscopy (range) Symptoms Rectal blood loss Altered bowel movements Weight loss Abdominal pain Anemia
26 11/15 (42%/58%) 66 (50–83) 10.8 (0–30)
9 (35%) 17 (65%) 1 (4%) 14 (54%) 4 (15%)
Colonoscopy results Mean number of lesions (mean, range) Locations of lesions Cecum Ascending colon Hepatic flexure Transverse colon Splenic flexure Descending colon Sigmoid Rectum Histology Hyperplastic Adenomatous Villous Tubulovillous Tubular Serrated Adenocarcinoma Not available
0.77 (0–5) 1 (5%) 4 (20%) 0 (0%) 3 (15%) 0 (0%) 1 (5%) 8 (40%) 3 (15%) 2 (10%) 8 (40%) 0 (0%) 3 (38%) 5 (62%) 0 (0%) 3 (15%) 7 (35%)
30
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35
moderate signal); 3=undetermined (localized, mild to moderate signal); 4=probably present (localized, strong signal with no definite margins); 5=definitely present (localized, strong signal with definite margins). Scores of 1 and 2 were considered to indicate the absence of a lesion (negative DWI), whereas scores of 3, 4 or 5 were considered to indicate the presence of a lesion (positive DWI). Note that a score of 3 (undetermined) was given the benefit of the doubt (i.e., considered to indicate a positive test result) because of the importance not to miss any (potential) lesion. (Fat-suppressed) T2-weighted and in-and-opposed phase T1weighted images were used to facilitate segmental localization of lesions, if detected at DWI. The number and likely nature of extracolonic findings were also recorded. 2.3. Colonoscopy Colonoscopy was performed by board-certified gastroenterologists who used a video colonoscope (CF-Q160AL, CFQ165L, CF-Q180AL, CF-H180AL, CF-H180DL — Olympus, Inc). The endoscopist was blinded to the MRI results. Bowel preparation complied with the standard protocol in our institution. If sedation was required, it was personalized for each individual and consisted of 25–50 mg of pethidine or 0.02–0.075 mg of fentanyl, and 2.5–7.5 mg of midazolam. Segmental location of detected lesions, lesion size as
estimated by comparing it to the size of biopsy forceps, and morphologic type were recorded. Biopsied lesions were analyzed by board-certified pathologists at our institution. 2.4. Statistical analyses The sensitivity and positive predictive value (PPV) of DWI for the detection of clinically relevant polyps (≥ 6 mm) and CRC were calculated on a per-lesion basis using colonoscopy results as the standard of reference, along with binomial exact 95% confidence intervals (CIs). Sensitivity, specificity, PPV and negative predictive value (NPV) on a per-patient basis, along with binomial exact 95% CIs, were also calculated. To these ends, a true positive was defined as a polyp or CRC detected by both DWI and colonoscopy in the same segmental location, a false positive was defined as a polyps or CRC detected by DWI in a particular segment location but not by colonoscopy, and a false negative was defined as a polyp or CRC detected in a particular segmental location with colonoscopy but not with DWI. For both the per-lesion and the per-patient assessments, separate analyses were made for all lesions (polyps of any size and CRC), polyps ≥ 6mm and CRC, and polyps ≥ 10mm and CRC. Statistical analyses were done using Stata version 10 software (StataCorp LP, College Station, TX, USA).
Fig. 1. A 71-year-old female with a 30-mm large adenocarcinoma in the sigmoid (Table 2, patient 20) and true-positive DWI. Coronal maximum intensity projection grayscale inverted DWI (A) and axial grayscale inverted DWI (B) show a high-signal-intensity lesion (arrows). Also note normal high signal intensity of nerve roots (A, arrowheads) and lymph nodes (A and B, encircled). Axial T2-weighted image (C) localizes the lesion in the sigmoid (arrow). A score of 5 [lesion definitely present (localized, strong signal with definite margins at DWI)] was assigned. Digital photograph from optical colonoscopy (D) shows a stenotic tumor in the sigmoid (arrowheads). Biopsies of the tumor and subsequent histopathological examination revealed a moderately differentiated adenocarcinoma.
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35
3. Results A total of 33 patients were potentially eligible for inclusion. Of these 33 patients, 7 patients were excluded because of undesired bowel preparation before the MRI examination (n= 2), claustrophobia (n= 2), time interval between MRI and colonoscopy of more than 1 month (n= 1), total hip prosthesis (n= 1) and incomplete colonoscopy due to a fixated sigmoid (n= 1). Thus, a total of 26 patients were finally included, and their characteristics are shown in Table 1. Altered bowel movements and abdominal pain were the predominant symptoms for patients to undergo colonoscopy. Mean time between MRI and colonoscopy was 10.8 days, with a range of 0–30 days. No complications or adverse events occurred during the MRI examinations or colonoscopies. Lesions found at colonoscopy were most often located in the sigmoid (40%), and these were most frequently adenomatous polyps (40%). In total, three adenocarcinomas were colonoscopically diagnosed in three patients, which were all detected by DWI (Fig. 1). Of 17 colonoscopically diagnosed polyps, 6 were detected by DWI and had a mean size of 9.8 mm (range,
31
5–10 mm) (Fig. 2). Missed polyps (n= 11) were located in the sigmoid (n= 3), ascending colon (n= 3), rectum (n= 2), transverse colon (n= 2) and cecum (n= 1), and had a mean size of 5.2 mm (range, 2–20 mm) (Fig. 3). In three patients, DWI suggested the presence of lesions, which, however, were not detected by colonoscopy (i.e., false positives) (Fig. 4). Table 2 shows the detailed characteristics of all lesions detected by DWI and colonoscopy. Tables 3 and 4 show the diagnostic performance of DWI on a per-lesion basis and on a per-patient basis, respectively. Sensitivity and PPV on a per-lesion basis were 45.0% and 75.0% for all lesions, 80.0% and 72.7% for polyps ≥ 6 mm and CRC (i.e., clinically relevant lesions), and 85.7% and 66.7% for polyps ≥ 10 mm and CRC. Sensitivity, specificity, PPV and NPV on a per-patient basis were 63.6%, 80.0%, 70.0% and 75.0% for all lesions and 85.7%, 84.2%, 66.7% and 94.1% for both polyps ≥ 6mm and CRC (i.e., clinically relevant lesions), and polyps ≥ 10mm and CRC. In eight patients, clinically irrelevant extracolonic findings were detected: simple liver cyst(s) in four patients, adnexal cysts in three patients (Fig. 2) and a prostate lesion (probably benign prostate hyperplasia) in one patient.
Fig. 2. A 69-year-old female with a pedunculated polyp (20 mm×12 mm×15 mm) in the sigmoid (Table 2, patient 7) and true-positive DWI. Coronal maximum intensity projection grayscale inverted DWI (A) and axial grayscale inverted DWI (B) show a high-signal-intensity lesion (arrows). Also note normal high signal intensity of lymph nodes (A, encircled). Axial T2-weighted image (C) localizes the lesion in the sigmoid (arrow) and also shows a left adnexal cyst (asterisk). A score of 5 [lesion definitely present (localized, strong signal with definite margins at DWI)] was assigned. Digital photograph from optical colonoscopy (D) shows a pedunculated polyp in the sigmoid (arrowhead). The lesion was excised, and subsequent histopathological examination indicated this lesion to be a tubulovillous adenoma with low-grade dysplasia. Colonoscopy also detected two more tubular adenomas with low-grade dysplasia more distal to the anus (not shown).
32
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35
Fig. 3. A 70-year-old female with a flat, 20-mm large tubulovillous adenoma in the cecum (Table 2, patient 17) and false-negative DWI. Coronal maximum intensity projection grayscale inverted DWI (A) does not show any (high signal intensity) lesions. Only note normal high-signal-intensity nerve roots (arrowheads) and lymph nodes (encircled). A score of 1 [lesion definitely absent (no signal at DWI)] was assigned. Digital photograph from optical colonoscopy (B) shows a flat polyp in the sigmoid (arrowhead). The lesion was excised, and subsequent histopathological examination indicated this lesion to be a tubulovillous adenoma with low-grade dysplasia.
4. Discussion The results of the present study show that DWI can detect colorectal polyps (Fig. 2), and confirm results of previous studies that DWI allows detecting CRC (Fig. 1) [7–9]. Nevertheless, DWI performed poorly in detecting polyps with a size less than 6 mm (lesion-based sensitivity of 45.0% for all lesions) and achieved a lesion-based sensitivity of only 80.0% for clinically relevant lesions (i.e.,
polyps ≥ 6 mm and CRC). Two clinically relevant lesions were missed (not recognized at all) by DWI: one 8-mm polyp in the ascending colon and one 20-mm tubulovillous adenoma with low-grade dysplasia in the cecum (Fig. 3). The first of these lesions was found in a patient who had three other clinically relevant lesions that were detected by DWI. Therefore, missing this lesion would not have any negative influence on clinical management because this patient would have undergone subsequent colonoscopy
Fig. 4. A 70-year-old female with diverticulosis but no other lesions at colonoscopy (Table 2, patient 22) and false-positive DWI. Axial grayscale inverted DWI (A) shows a high-signal-intensity structure (arrow). Axial T2-weighted image (B) localizes this high signal intensity in the sigmoid (arrow). A score of 4 [lesion probably present (localized, strong signal with no definite margins at DWI)] was assigned. Colonoscopy, however, did not detect any lesions. Retrospective review of the T2-weighted image (B) shows an air bubble in the sigmoid (arrow), which may have led to a susceptibility artifact and high signal intensity at DWI. Also note air in the adjacent rectum (asterisk).
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35
33
Table 2 Detailed characteristics of all lesions detected by DWI and colonoscopy Patient no.
Lesion no. 1 (location, size, PA, score MRI a)
Lesions detected with both DWI and colonoscopy (true positives) 7 Sigmoid, 20 mm, tubulovillous adenoma with LGD, 5 15 Sigmoid, 10 mm, tubular adenoma with LGD, 4 20 Sigmoid, 30 mm, adenocarcinoma, 5 24 Ascending colon, 3 mm, tubular adenoma with LGD, 4 25 Rectum, 30 mm, adenocarcinoma, 5 27 Transverse colon, 10 mm, tubulovillous adenoma with LGD, 3 30 Sigmoid, 10 mm Adenocarcinoma, 3 Lesions detected with colonoscopy only (false negatives) 7 Sigmoid, 3 mm, tubular adenoma with LGD, x 9 15 17 24 25 27
Rectum, 3 mm, hyperplastic, 1 Ascending colon, 8 mm, no PA, x Cecum, 20 mm Tubulovillous adenoma with LGD, 1 Transverse colon, 4 mm, no PA, x Rectum, 3 mm, hyperplastic, x Ascending colon, 2 mm, no PA, x
Lesion no. 2 (location, size, PA, score MRI a)
Lesion no. 3 (location, size, PA, score MRI a)
Sigmoid, 8 mm, no PA, 4
Descending colon, 6 mm, no PA, 3
Sigmoid, 3 mm, tubular adenoma with LGD, x Sigmoid, 5 mm, no PA, x
Ascending colon, 3 mm, no PA, x
Transverse colon, 3 mm, tubular adenoma with LGD, x
Lesions detected with DWI only (false positives) 4 Descending colon, NA, NA, 4 12 Sigmoid, NA, NA, 4 22 Cecum, NA, NA, 3 Notes: Clinically relevant lesions (polyps ≥ 6mm and CRC) are marked italic. Abbreviations: LGD: low-grade dysplasia; NA: not applicable; PA: pathological result. a Probability score MRI: 1=definitely no lesion (no signal); 2=probably no lesion (nonlocalized, mild signal to moderate signal); 3=undetermined (localized, mild to moderate signal); 4=probably a lesion (localized, strong signal with no definite margins); 5=very likely a lesion (localized, strong signal with definite margins); x=no score assigned.
anyway because of DWI findings. The second lesion was found in a patient without any other colorectal lesions. Therefore, missing this lesion would have a negative influence on patient management because subsequent colonoscopy would have been denied to this patient based on DWI findings. Consequently, patient-based sensitivity and NPV for the detection of clinically relevant lesions were imperfect, with values of 85.7% and 94.1%, respectively. Furthermore, DWI incorrectly classified three lesions as positive in three patients, as a result of which these patients would have gone unnecessary subsequent colonoscopy, and patient-based PPV was only 66.7%. The false-positive results may be explained by residual high bowel signal intensity (probably in case of viscous liquid-like stool) and susceptibility artifacts due to intraluminal air (Fig. 4). Based on the results of the present study in which (lesionbased and patient-based) sensitivity and (patient-based) NPV were imperfect, DWI cannot yet be recommended in a clinical setting in which DWI is performed first and subsequent colonoscopy is only performed in patients with positive findings at DWI. Several previous studies have investigated the utility of DWI for the detection of (colo)rectal cancer [7–9]. Hosonuma et al. [7] conducted a retrospective study in 15 patients with rectal cancer and 20 patients with colon cancer and no other lesions as controls to determine the detec-
tability of rectal cancer at DWI. DWI detected all rectal cancers. In the control group, 13 cases were classified as negative and 7 cases as positive for rectal cancer. Thus, sensitivity for the detection of rectal cancer was 100% (15/15), and specificity was 65% (13/20). Although Hosonuma et al. [7] concluded that DWI may be used to screen for rectal cancer, their study did not include any rectal polyps, as a result of which its value in this setting actually remained unclear. In another study, Ichikawa et al. [8] retrospectively assessed the diagnostic value of DWI for the detection of CRC in 33 patients with CRC and 15 controls, and reported a sensitivity of 90% and a specificity 100%. Although Ichikawa et al. [8] concluded that DWI Table 3 Sensitivity and PPV of DWI on a per-lesion basis for all lesions (polyps of any size and CRC), polyps ≥ 6 mm and CRC, and polyps ≥ 10 mm and CRC)
All lesions Polyps ≥ 6 mm and CRC Polyps ≥ 10 mm and CRC
Sensitivity
PPV
45.0% (6/7) (25.8%–65.8%) 80.0% (8/10) (49.0%–94.3%) 85.7% (6/7) (48.7%–97.4%)
75.0% (9/12) (46.8%–91.1%) 72.7% (8/11) (43.4%–90.3%) 66.7% (6/9) (35.4%–87.9%)
Absolute numbers and binomial 95% CIs are given between parentheses.
34
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35
Table 4 Sensitivity, specificity, PPV and NPV on a per-patient basis for all lesions (polyps of any size and CRC, polyps ≥ 6 mm and CRC, and polyps ≥ 10 mm and CRC) Sensitivity
Specificity
PPV
NPV
All lesions
63.6% (7/11) (35.4%–84.8%)
80.0% (12/15) (54.8%–93.0%)
70.0% (7/10) (39.7%–89.2%)
Polyps ≥ 6 mm and CRC
85.7% (6/7) (48.7%–97.4%)
84.2% (16/19) (62.4%–94.5%)
66.7% (6/9) (35.4%–87.9%)
Polyps ≥ 10 mm and CRC
85.7% (6/7) (48.7%–97.4%)
84.2% (16/19) (62.4%–94.5%)
66.7% (6/9) (35.4%–87.9%)
75.0% (12/16) (50.5%–89.8%) 94.1% (16/17) (73.0%–99.0%) 94.1% (16/17) (73.0%–99.0%)
Absolute numbers and binomial 95% CIs are given between parentheses.
allows detection of colorectal adenocarcinoma with a high sensitivity and specificity, no patients with polyps were included. Therefore, specificity could have been overrated, and the authors were in fact not able to make a statement on the potential of DWI to detect colorectal polyps. In yet another study, Rao et al [9] retrospectively evaluated 45 patients with rectal cancer and 20 controls to assess the value of DWI in combination with T2-weighted imaging for the detection of rectal cancer as compared with T2-weighted imaging alone. Areas under the receiver operating characteristic curve of DWI combined with T2-weighted imaging (0.991 and 0.997 for readers 1 and 2, respectively) were significantly higher (P=.0494 and P=.0475 for readers 1 and 2, respectively) than those of T2-weighted imaging alone (0.918 and 0.934 for readers 1 and 2, respectively). Thus, Rao et al [9] concluded that the addition of DWI to conventional T2-weighted imaging provides better detection of rectal cancer. Again, however, no rectal polyps were included in this study [9]. Further developments are required to increase the diagnostic yield of DWI in the detection of colorectal polyps. More specifically, signal of the tumor should be enhanced, surrounding normal background tissue should be further suppressed, and disturbing artifacts should be eliminated. Increasing the field strength (i.e., moving from 1.5 T to 3.0 T), performing the MRI examinations with systems that allow for parallel radiofrequency transmission, sequence optimization through use of improved fat/background suppression techniques (such as single-axis diffusion encoding and short inversion time inversion recovery fat suppression combined with slice-selective gradient reversal fat suppression) and optimized b-values may allow us to achieve these goals [10–12]. Another consideration is to use negative oral contrast agents to reduce bowel signal intensity. The present study had several limitations. First, sample size was relatively small. Second, the included patients were not representative of those in a screening setting because they were all symptomatic. Nevertheless, regardless of the patient population, the results demonstrated that DWI is still insufficiently reliable to be used as a selection method for subsequent colonoscopy. Third, only one observer evaluated the MRI data sets; future studies should determine the intraand interobserver variability of DWI in the detection of colorectal polyps and CRC. Fourth, the diagnostic perfor-
mance of DWI was compared to that of colonoscopy, which is regarded as the gold standard for the detection of colorectal polyps and CRC [13]. Although direct observation is assumed to be most accurate, lesions could have been missed by colonoscopy because of blind spots [13,14]. Fifth, in the three patients in whom a lesion was detected with DWI but not with subsequent colonoscopy (note that the endoscopist was blinded to the MRI results), no second colonoscopy was performed. As a result, it could not be completely excluded that, indeed, no lesions were present in these patients. Sixth, in the present study, DWI was not compared to other diagnostic tests such as stool-based tests and computed tomography colonography. In conclusion, DWI cannot yet be recommended in a clinical setting in which DWI is performed first and subsequent colonoscopy is only performed in patients with positive findings at DWI. Further (technical) developments are required to increase its diagnostic yield. References [1] Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA Cancer J Clin 2011;61:69–90. [2] Winawer SJ, Zauber AG, Ho MN, et al. Prevention of colorectal cancer by colonoscopic polypectomy. The National Polyp Study Workgroup. N Engl J Med 1993;329:1977–81. [3] Steinwachs D, Allen JD, Barlow WE, et al. National Institutes of Health state-of-the-science conference statement: enhancing use and quality of colorectal cancer screening. Ann Intern Med 2010;152:663–7. [4] U.S. Preventive Services Task Force. Screening for colorectal cancer: U.S. Preventive Services Task Force recommendation statement. Ann Intern Med 2008;149:627–37. [5] Takahara T, Imai Y, Yamashita T, Yasuda S, Nasu S, Van Cauteren M. Diffusion weighted whole body imaging with background body signal suppression (DWIBS): technical improvement using free breathing, STIR and high resolution 3D display. Radiat Med 2004;22:275–82. [6] Kwee TC, Takahara T, Ochiai R, Nievelstein RA, Luijten PR. Diffusion-weighted whole-body imaging with background body signal suppression (DWIBS): features and potential applications in oncology. Eur Radiol 2008;18:1937–52. [7] Hosonuma T, Tozaki M, Ichiba N, et al. Clinical usefulness of diffusion-weighted imaging using low and high b-values to detect rectal cancer. Magn Reson Med Sci 2006;5:173–7. [8] Ichikawa T, Erturk SM, Motosugi U, et al. High-B-value diffusionweighted MRI in colorectal cancer. AJR Am J Roentgenol 2006;187: 181–4. [9] Rao SX, Zeng MS, Chen CZ, et al. The value of diffusion-weighted imaging in combination with T2-weighted imaging for rectal cancer detection. Eur J Radiol 2008;65:299–303.
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35 [10] Kwee TC, Takahara T, Niwa T, et al. Improving background suppression in diffusion-weighted imaging of the abdomen and pelvis using STIR with single-axis diffusion encoding. Magn Reson Imaging 2011;29:877–80. [11] Mürtz P, Kaschner M, Träber F, et al. Diffusion-weighted whole-body MRI with background body signal suppression: technical improvements at 3.0 T. J Magn Reson Imaging 2012;35:456–61. [12] Mürtz P, Kaschner M, Träber F, et al. Evaluation of dual-source parallel RF excitation for diffusion-weighted whole-body MR imaging with
35
background body signal suppression at 3.0T. Eur J Radiol 2011. Dec 15 [Epub ahead of print], http://dx.doi.org/10.1016/j.ejrad.2011.11.024. [13] Citarda F, Tomaselli G, Capocaccia R, Barcherini S, Crespi M, Italian Multicentre Study Group. Efficacy in standard clinical practice of colonoscopic polypectomy in reducing colorectal cancer incidence. Gut 2001;48:812–5. [14] Van Rijn JC, Reitsma JB, Stoker J, Bossuyt PM, van Deventer SJ, Dekker E. Polyp miss rate determined by tandem colonoscopy: a systematic review. Am J Gastroenterol 2006;101:343–50.
Magnetic Resonance Imaging 31 (2013) 28 – 35
Diffusion-weighted MRI for the detection of colorectal polyps: feasibility study☆,☆☆,★,★★ Anke M. Leufkensa , Thomas C. Kwee b,⁎, Maurice A.A.J. van den Boschb , Willem P.Th.M. Malib , Taro Takaharab , Peter D. Siersema a a
Department of Gastroenterology and Hepatology, University Medical Center Utrecht, Utrecht, The Netherlands b Department of Radiology, University Medical Center Utrecht, Utrecht, The Netherlands Received 30 April 2012; accepted 21 June 2012
Abstract The purpose of this study was to determine the feasibility of diffusion-weighted magnetic resonance imaging (DWI) for detecting colorectal polyps. DWI (high b-value of 1000 s/mm 2) was prospectively performed in 26 symptomatic patients who were scheduled to undergo colonoscopy. DWI and colonoscopic findings were interpreted in a blinded manner. The sensitivity and positive predictive value (PPV) of DWI for the detection of clinically relevant polyps (≥ 6 mm) and colorectal cancer (CRC) were calculated on a per-lesion basis, using colonoscopy results as the standard of reference. Sensitivity, specificity, PPV and negative predictive value (NPV) on a per-patient basis were also calculated. Sensitivity and PPV on a per-lesion basis were 80.0% [95% confidence interval (CI): 49.0%–94.3%] and 72.7% (95% CI: 43.4%–90.3%) for polyps ≥ 6 mm and CRC. Sensitivity, specificity, PPV and NPV on a per-patient basis were 85.7% (95% CI: 48.7%–97.4%), 84.2% (95% CI: 62.4%–94.5%), 66.7% (95% CI: 35.4%–87.9%) and 94.1% (95% CI: 73.0%–99.0%) for polyps ≥ 6mm and CRC. In conclusion, DWI cannot yet be recommended in a clinical setting in which DWI is performed first and subsequent colonoscopy is only performed in patients with positive findings at DWI. Further (technical) developments are required to increase its diagnostic yield. © 2013 Elsevier Inc. All rights reserved. Keywords: Colorectal; Polyps; Carcinoma; Magnetic resonance imaging; MRI; Diffusion-weighted
1. Introduction Colorectal cancer (CRC) is the third most common cancer in men and women and the third leading cause of cancer death in the Western world [1]. It has been shown that colonoscopic polypectomy results in a decreased incidence and mortality of CRC [2–4]. Magnetic resonance imaging (MRI) does not use any ionizing radiation and provides a high soft tissue contrast, making it a potentially useful method for the detection or exclusion of colorectal polyps and CRC. Of particular interest is the introduction of diffusion-weighted MRI (DWI) for tumor detection and staging in the body [5,6]. DWI is a functional MRI technique ☆ All authors contributed to study concept/design, data acquisition, data analysis/interpretation and manuscript drafting/editing. ☆☆ All authors approved the final version of the submitted manuscript. ★ All authors guarantee the integrity of the entire study. ★★ None of the authors has any (potential) conflict of interest to disclose. ⁎ Corresponding author. Tel.: +31 88 7556687; fax: +31 30 2581098. E-mail address: [email protected] (T.C. Kwee).
0730-725X/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.mri.2012.06.029
that allows visualization of the random (Brownian) motion of water molecules. As many malignant tumors, including CRC, exhibit an impeded diffusion (most likely due to increased cellularity), they can be highlighted (i.e., high signal intensity) at DWI, while most normal background tissue can be suppressed (i.e., low signal intensity). The high lesion-to-background contrast that can be provided by DWI is an advantage compared to conventional (T1-weighted or T2-weighted) sequences and may even allow for the detection of lesions with a size around or even below that of the voxel size. Furthermore, DWI does not require the administration of any contrast agents and can be performed without any bowel preparation. Several previous studies have indicated the utility of DWI for the detection of CRC [7–9]. However, the evidence provided by these studies is limited due to retrospective inclusion of patients (and controls) and patient selection bias. Furthermore, and most importantly, no previous study has investigated whether DWI allows detection of clinically relevant colorectal polyps (i.e., ≥ 6 mm). If DWI proves to be accurate in detecting
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35
clinically relevant colorectal polyps, it may be used in a clinical setting in which DWI is performed first and subsequent colonoscopy is only performed in those patients with positive DWI findings to confirm the presence of lesions and perform polypectomy. Patients with a negative DWI may be spared a (more invasive, more time-consuming and more expensive) colonoscopic investigation, and this strategy may reduce the increasing pressure on the utilization rate of colonoscopy. The aim of this study was therefore to prospectively determine the feasibility of DWI for the detection of colorectal polyps. 2. Materials and methods 2.1. Patients This study was approval by the local Institutional Review Board, and each patient provided written informed consent prior to inclusion. Included patients prospectively underwent MRI before colonoscopy. Patients were eligible for inclusion if they were aged ≥ 50 years and scheduled to undergo colonoscopy as part of routine clinical care because of symptoms such as rectal bleeding, altered bowel habits, unexplained weight loss, abdominal pain or anemia. Exclusion criteria were incomplete colonoscopy, general contraindications for MRI (including metal objects in the body like brain aneurysm clips or implanted cardiac pacemakers, and claustrophobia), a body size that did not allow for MRI scanning, patients who had undergone bowel preparation, previous abdominal malignancies or melanoma, pregnancy or lactation, previous CRC-directed treatment, and time interval of more than 1 month between MRI and colonoscopy. 2.2. MRI MRI was performed using a 1.5-T system (Achieva, Philips Healthcare, Best, the Netherlands) with a 16-element phased-array surface coil for signal reception (SENSE Torso XL). Applied parameters for DWI were as follows: singleshot spin-echo echo-planar imaging, repetition time (TR)/ echo time (TE)/inversion time of 8551/74/180 ms, axial slice orientation, number of stations of 2, slice thickness/gap of 4/0mm, number of slices of 60 for each of the two stations, overlap of 8 mm between the two stations, field of view of 400×325 mm 2, acquisition matrix of 128×83, receiver bandwidth of 2168.7 Hz/pixel, motion probing gradients in three orthogonal axes, b-values of 0 and 1000 s/mm 2, number of excitations of 4, partial Fourier (half scan) factor of 0.6, parallel acquisition (SENSitivity Encoding) factor of 2, echo train length of 45, acquired voxel size of 3.12×3.92×4.00 mm 3 , reconstructed voxel size of 1.56×1.56×4.00 mm 3, image acquisition under free breathing, and scan time of 11 min and 42 s (5 min and 51 s for each station). In all patients, axial breath-hold T2-weighted (TR/TE of 1000/80 ms, slice thickness/gap of 8/0mm), coronal breath-hold T2-weighted and fat-suppressed T2-
29
weighted (TR/TE of 1258/80 ms, slice thickness/gap of 10/0mm), and coronal in-and-opposed phase T1-weighted [TR/TE of 132/2.3 (opposed-phase) and 4.6 (in-phase) ms, slice thickness/gap of 9/1 mm] were obtained in addition to DWI. All sequences covered the area from the diaphragm to the pubic bone. Total examination time was approximately 20–25 min. No bowel preparation or antiperistaltic agents were administered, and no dietary restrictions were recommended in any of the patients. Using software implemented in the standard operating console, additional coronally reformatted images with a slice thickness/gap of 5/0 mm and coronal maximum intensity projection diffusion-weighted images were created of the axial DWI data set. Diffusion-weighted images were displayed using grayscale inversion. The MRI data set was then transferred to a Picture Archiving and Communications System which allows manual window level setting and systematic evaluation by a board-certified radiologist who was blinded to colonoscopy findings. To that end, high-b-value (1000 s/mm 2) diffusionweighted images were assessed for the probability of the absence or presence of lesions (colorectal polyps or CRC) using the following 5-point grading scale: 1=definitely absent (no signal); 2=probably absent (nonlocalized, mild to Table 1 Characteristics of included patients Overall No. of patients Male/female Mean age (range) Mean no. of days between MRI and colonoscopy (range) Symptoms Rectal blood loss Altered bowel movements Weight loss Abdominal pain Anemia
26 11/15 (42%/58%) 66 (50–83) 10.8 (0–30)
9 (35%) 17 (65%) 1 (4%) 14 (54%) 4 (15%)
Colonoscopy results Mean number of lesions (mean, range) Locations of lesions Cecum Ascending colon Hepatic flexure Transverse colon Splenic flexure Descending colon Sigmoid Rectum Histology Hyperplastic Adenomatous Villous Tubulovillous Tubular Serrated Adenocarcinoma Not available
0.77 (0–5) 1 (5%) 4 (20%) 0 (0%) 3 (15%) 0 (0%) 1 (5%) 8 (40%) 3 (15%) 2 (10%) 8 (40%) 0 (0%) 3 (38%) 5 (62%) 0 (0%) 3 (15%) 7 (35%)
30
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35
moderate signal); 3=undetermined (localized, mild to moderate signal); 4=probably present (localized, strong signal with no definite margins); 5=definitely present (localized, strong signal with definite margins). Scores of 1 and 2 were considered to indicate the absence of a lesion (negative DWI), whereas scores of 3, 4 or 5 were considered to indicate the presence of a lesion (positive DWI). Note that a score of 3 (undetermined) was given the benefit of the doubt (i.e., considered to indicate a positive test result) because of the importance not to miss any (potential) lesion. (Fat-suppressed) T2-weighted and in-and-opposed phase T1weighted images were used to facilitate segmental localization of lesions, if detected at DWI. The number and likely nature of extracolonic findings were also recorded. 2.3. Colonoscopy Colonoscopy was performed by board-certified gastroenterologists who used a video colonoscope (CF-Q160AL, CFQ165L, CF-Q180AL, CF-H180AL, CF-H180DL — Olympus, Inc). The endoscopist was blinded to the MRI results. Bowel preparation complied with the standard protocol in our institution. If sedation was required, it was personalized for each individual and consisted of 25–50 mg of pethidine or 0.02–0.075 mg of fentanyl, and 2.5–7.5 mg of midazolam. Segmental location of detected lesions, lesion size as
estimated by comparing it to the size of biopsy forceps, and morphologic type were recorded. Biopsied lesions were analyzed by board-certified pathologists at our institution. 2.4. Statistical analyses The sensitivity and positive predictive value (PPV) of DWI for the detection of clinically relevant polyps (≥ 6 mm) and CRC were calculated on a per-lesion basis using colonoscopy results as the standard of reference, along with binomial exact 95% confidence intervals (CIs). Sensitivity, specificity, PPV and negative predictive value (NPV) on a per-patient basis, along with binomial exact 95% CIs, were also calculated. To these ends, a true positive was defined as a polyp or CRC detected by both DWI and colonoscopy in the same segmental location, a false positive was defined as a polyps or CRC detected by DWI in a particular segment location but not by colonoscopy, and a false negative was defined as a polyp or CRC detected in a particular segmental location with colonoscopy but not with DWI. For both the per-lesion and the per-patient assessments, separate analyses were made for all lesions (polyps of any size and CRC), polyps ≥ 6mm and CRC, and polyps ≥ 10mm and CRC. Statistical analyses were done using Stata version 10 software (StataCorp LP, College Station, TX, USA).
Fig. 1. A 71-year-old female with a 30-mm large adenocarcinoma in the sigmoid (Table 2, patient 20) and true-positive DWI. Coronal maximum intensity projection grayscale inverted DWI (A) and axial grayscale inverted DWI (B) show a high-signal-intensity lesion (arrows). Also note normal high signal intensity of nerve roots (A, arrowheads) and lymph nodes (A and B, encircled). Axial T2-weighted image (C) localizes the lesion in the sigmoid (arrow). A score of 5 [lesion definitely present (localized, strong signal with definite margins at DWI)] was assigned. Digital photograph from optical colonoscopy (D) shows a stenotic tumor in the sigmoid (arrowheads). Biopsies of the tumor and subsequent histopathological examination revealed a moderately differentiated adenocarcinoma.
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35
3. Results A total of 33 patients were potentially eligible for inclusion. Of these 33 patients, 7 patients were excluded because of undesired bowel preparation before the MRI examination (n= 2), claustrophobia (n= 2), time interval between MRI and colonoscopy of more than 1 month (n= 1), total hip prosthesis (n= 1) and incomplete colonoscopy due to a fixated sigmoid (n= 1). Thus, a total of 26 patients were finally included, and their characteristics are shown in Table 1. Altered bowel movements and abdominal pain were the predominant symptoms for patients to undergo colonoscopy. Mean time between MRI and colonoscopy was 10.8 days, with a range of 0–30 days. No complications or adverse events occurred during the MRI examinations or colonoscopies. Lesions found at colonoscopy were most often located in the sigmoid (40%), and these were most frequently adenomatous polyps (40%). In total, three adenocarcinomas were colonoscopically diagnosed in three patients, which were all detected by DWI (Fig. 1). Of 17 colonoscopically diagnosed polyps, 6 were detected by DWI and had a mean size of 9.8 mm (range,
31
5–10 mm) (Fig. 2). Missed polyps (n= 11) were located in the sigmoid (n= 3), ascending colon (n= 3), rectum (n= 2), transverse colon (n= 2) and cecum (n= 1), and had a mean size of 5.2 mm (range, 2–20 mm) (Fig. 3). In three patients, DWI suggested the presence of lesions, which, however, were not detected by colonoscopy (i.e., false positives) (Fig. 4). Table 2 shows the detailed characteristics of all lesions detected by DWI and colonoscopy. Tables 3 and 4 show the diagnostic performance of DWI on a per-lesion basis and on a per-patient basis, respectively. Sensitivity and PPV on a per-lesion basis were 45.0% and 75.0% for all lesions, 80.0% and 72.7% for polyps ≥ 6 mm and CRC (i.e., clinically relevant lesions), and 85.7% and 66.7% for polyps ≥ 10 mm and CRC. Sensitivity, specificity, PPV and NPV on a per-patient basis were 63.6%, 80.0%, 70.0% and 75.0% for all lesions and 85.7%, 84.2%, 66.7% and 94.1% for both polyps ≥ 6mm and CRC (i.e., clinically relevant lesions), and polyps ≥ 10mm and CRC. In eight patients, clinically irrelevant extracolonic findings were detected: simple liver cyst(s) in four patients, adnexal cysts in three patients (Fig. 2) and a prostate lesion (probably benign prostate hyperplasia) in one patient.
Fig. 2. A 69-year-old female with a pedunculated polyp (20 mm×12 mm×15 mm) in the sigmoid (Table 2, patient 7) and true-positive DWI. Coronal maximum intensity projection grayscale inverted DWI (A) and axial grayscale inverted DWI (B) show a high-signal-intensity lesion (arrows). Also note normal high signal intensity of lymph nodes (A, encircled). Axial T2-weighted image (C) localizes the lesion in the sigmoid (arrow) and also shows a left adnexal cyst (asterisk). A score of 5 [lesion definitely present (localized, strong signal with definite margins at DWI)] was assigned. Digital photograph from optical colonoscopy (D) shows a pedunculated polyp in the sigmoid (arrowhead). The lesion was excised, and subsequent histopathological examination indicated this lesion to be a tubulovillous adenoma with low-grade dysplasia. Colonoscopy also detected two more tubular adenomas with low-grade dysplasia more distal to the anus (not shown).
32
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35
Fig. 3. A 70-year-old female with a flat, 20-mm large tubulovillous adenoma in the cecum (Table 2, patient 17) and false-negative DWI. Coronal maximum intensity projection grayscale inverted DWI (A) does not show any (high signal intensity) lesions. Only note normal high-signal-intensity nerve roots (arrowheads) and lymph nodes (encircled). A score of 1 [lesion definitely absent (no signal at DWI)] was assigned. Digital photograph from optical colonoscopy (B) shows a flat polyp in the sigmoid (arrowhead). The lesion was excised, and subsequent histopathological examination indicated this lesion to be a tubulovillous adenoma with low-grade dysplasia.
4. Discussion The results of the present study show that DWI can detect colorectal polyps (Fig. 2), and confirm results of previous studies that DWI allows detecting CRC (Fig. 1) [7–9]. Nevertheless, DWI performed poorly in detecting polyps with a size less than 6 mm (lesion-based sensitivity of 45.0% for all lesions) and achieved a lesion-based sensitivity of only 80.0% for clinically relevant lesions (i.e.,
polyps ≥ 6 mm and CRC). Two clinically relevant lesions were missed (not recognized at all) by DWI: one 8-mm polyp in the ascending colon and one 20-mm tubulovillous adenoma with low-grade dysplasia in the cecum (Fig. 3). The first of these lesions was found in a patient who had three other clinically relevant lesions that were detected by DWI. Therefore, missing this lesion would not have any negative influence on clinical management because this patient would have undergone subsequent colonoscopy
Fig. 4. A 70-year-old female with diverticulosis but no other lesions at colonoscopy (Table 2, patient 22) and false-positive DWI. Axial grayscale inverted DWI (A) shows a high-signal-intensity structure (arrow). Axial T2-weighted image (B) localizes this high signal intensity in the sigmoid (arrow). A score of 4 [lesion probably present (localized, strong signal with no definite margins at DWI)] was assigned. Colonoscopy, however, did not detect any lesions. Retrospective review of the T2-weighted image (B) shows an air bubble in the sigmoid (arrow), which may have led to a susceptibility artifact and high signal intensity at DWI. Also note air in the adjacent rectum (asterisk).
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35
33
Table 2 Detailed characteristics of all lesions detected by DWI and colonoscopy Patient no.
Lesion no. 1 (location, size, PA, score MRI a)
Lesions detected with both DWI and colonoscopy (true positives) 7 Sigmoid, 20 mm, tubulovillous adenoma with LGD, 5 15 Sigmoid, 10 mm, tubular adenoma with LGD, 4 20 Sigmoid, 30 mm, adenocarcinoma, 5 24 Ascending colon, 3 mm, tubular adenoma with LGD, 4 25 Rectum, 30 mm, adenocarcinoma, 5 27 Transverse colon, 10 mm, tubulovillous adenoma with LGD, 3 30 Sigmoid, 10 mm Adenocarcinoma, 3 Lesions detected with colonoscopy only (false negatives) 7 Sigmoid, 3 mm, tubular adenoma with LGD, x 9 15 17 24 25 27
Rectum, 3 mm, hyperplastic, 1 Ascending colon, 8 mm, no PA, x Cecum, 20 mm Tubulovillous adenoma with LGD, 1 Transverse colon, 4 mm, no PA, x Rectum, 3 mm, hyperplastic, x Ascending colon, 2 mm, no PA, x
Lesion no. 2 (location, size, PA, score MRI a)
Lesion no. 3 (location, size, PA, score MRI a)
Sigmoid, 8 mm, no PA, 4
Descending colon, 6 mm, no PA, 3
Sigmoid, 3 mm, tubular adenoma with LGD, x Sigmoid, 5 mm, no PA, x
Ascending colon, 3 mm, no PA, x
Transverse colon, 3 mm, tubular adenoma with LGD, x
Lesions detected with DWI only (false positives) 4 Descending colon, NA, NA, 4 12 Sigmoid, NA, NA, 4 22 Cecum, NA, NA, 3 Notes: Clinically relevant lesions (polyps ≥ 6mm and CRC) are marked italic. Abbreviations: LGD: low-grade dysplasia; NA: not applicable; PA: pathological result. a Probability score MRI: 1=definitely no lesion (no signal); 2=probably no lesion (nonlocalized, mild signal to moderate signal); 3=undetermined (localized, mild to moderate signal); 4=probably a lesion (localized, strong signal with no definite margins); 5=very likely a lesion (localized, strong signal with definite margins); x=no score assigned.
anyway because of DWI findings. The second lesion was found in a patient without any other colorectal lesions. Therefore, missing this lesion would have a negative influence on patient management because subsequent colonoscopy would have been denied to this patient based on DWI findings. Consequently, patient-based sensitivity and NPV for the detection of clinically relevant lesions were imperfect, with values of 85.7% and 94.1%, respectively. Furthermore, DWI incorrectly classified three lesions as positive in three patients, as a result of which these patients would have gone unnecessary subsequent colonoscopy, and patient-based PPV was only 66.7%. The false-positive results may be explained by residual high bowel signal intensity (probably in case of viscous liquid-like stool) and susceptibility artifacts due to intraluminal air (Fig. 4). Based on the results of the present study in which (lesionbased and patient-based) sensitivity and (patient-based) NPV were imperfect, DWI cannot yet be recommended in a clinical setting in which DWI is performed first and subsequent colonoscopy is only performed in patients with positive findings at DWI. Several previous studies have investigated the utility of DWI for the detection of (colo)rectal cancer [7–9]. Hosonuma et al. [7] conducted a retrospective study in 15 patients with rectal cancer and 20 patients with colon cancer and no other lesions as controls to determine the detec-
tability of rectal cancer at DWI. DWI detected all rectal cancers. In the control group, 13 cases were classified as negative and 7 cases as positive for rectal cancer. Thus, sensitivity for the detection of rectal cancer was 100% (15/15), and specificity was 65% (13/20). Although Hosonuma et al. [7] concluded that DWI may be used to screen for rectal cancer, their study did not include any rectal polyps, as a result of which its value in this setting actually remained unclear. In another study, Ichikawa et al. [8] retrospectively assessed the diagnostic value of DWI for the detection of CRC in 33 patients with CRC and 15 controls, and reported a sensitivity of 90% and a specificity 100%. Although Ichikawa et al. [8] concluded that DWI Table 3 Sensitivity and PPV of DWI on a per-lesion basis for all lesions (polyps of any size and CRC), polyps ≥ 6 mm and CRC, and polyps ≥ 10 mm and CRC)
All lesions Polyps ≥ 6 mm and CRC Polyps ≥ 10 mm and CRC
Sensitivity
PPV
45.0% (6/7) (25.8%–65.8%) 80.0% (8/10) (49.0%–94.3%) 85.7% (6/7) (48.7%–97.4%)
75.0% (9/12) (46.8%–91.1%) 72.7% (8/11) (43.4%–90.3%) 66.7% (6/9) (35.4%–87.9%)
Absolute numbers and binomial 95% CIs are given between parentheses.
34
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35
Table 4 Sensitivity, specificity, PPV and NPV on a per-patient basis for all lesions (polyps of any size and CRC, polyps ≥ 6 mm and CRC, and polyps ≥ 10 mm and CRC) Sensitivity
Specificity
PPV
NPV
All lesions
63.6% (7/11) (35.4%–84.8%)
80.0% (12/15) (54.8%–93.0%)
70.0% (7/10) (39.7%–89.2%)
Polyps ≥ 6 mm and CRC
85.7% (6/7) (48.7%–97.4%)
84.2% (16/19) (62.4%–94.5%)
66.7% (6/9) (35.4%–87.9%)
Polyps ≥ 10 mm and CRC
85.7% (6/7) (48.7%–97.4%)
84.2% (16/19) (62.4%–94.5%)
66.7% (6/9) (35.4%–87.9%)
75.0% (12/16) (50.5%–89.8%) 94.1% (16/17) (73.0%–99.0%) 94.1% (16/17) (73.0%–99.0%)
Absolute numbers and binomial 95% CIs are given between parentheses.
allows detection of colorectal adenocarcinoma with a high sensitivity and specificity, no patients with polyps were included. Therefore, specificity could have been overrated, and the authors were in fact not able to make a statement on the potential of DWI to detect colorectal polyps. In yet another study, Rao et al [9] retrospectively evaluated 45 patients with rectal cancer and 20 controls to assess the value of DWI in combination with T2-weighted imaging for the detection of rectal cancer as compared with T2-weighted imaging alone. Areas under the receiver operating characteristic curve of DWI combined with T2-weighted imaging (0.991 and 0.997 for readers 1 and 2, respectively) were significantly higher (P=.0494 and P=.0475 for readers 1 and 2, respectively) than those of T2-weighted imaging alone (0.918 and 0.934 for readers 1 and 2, respectively). Thus, Rao et al [9] concluded that the addition of DWI to conventional T2-weighted imaging provides better detection of rectal cancer. Again, however, no rectal polyps were included in this study [9]. Further developments are required to increase the diagnostic yield of DWI in the detection of colorectal polyps. More specifically, signal of the tumor should be enhanced, surrounding normal background tissue should be further suppressed, and disturbing artifacts should be eliminated. Increasing the field strength (i.e., moving from 1.5 T to 3.0 T), performing the MRI examinations with systems that allow for parallel radiofrequency transmission, sequence optimization through use of improved fat/background suppression techniques (such as single-axis diffusion encoding and short inversion time inversion recovery fat suppression combined with slice-selective gradient reversal fat suppression) and optimized b-values may allow us to achieve these goals [10–12]. Another consideration is to use negative oral contrast agents to reduce bowel signal intensity. The present study had several limitations. First, sample size was relatively small. Second, the included patients were not representative of those in a screening setting because they were all symptomatic. Nevertheless, regardless of the patient population, the results demonstrated that DWI is still insufficiently reliable to be used as a selection method for subsequent colonoscopy. Third, only one observer evaluated the MRI data sets; future studies should determine the intraand interobserver variability of DWI in the detection of colorectal polyps and CRC. Fourth, the diagnostic perfor-
mance of DWI was compared to that of colonoscopy, which is regarded as the gold standard for the detection of colorectal polyps and CRC [13]. Although direct observation is assumed to be most accurate, lesions could have been missed by colonoscopy because of blind spots [13,14]. Fifth, in the three patients in whom a lesion was detected with DWI but not with subsequent colonoscopy (note that the endoscopist was blinded to the MRI results), no second colonoscopy was performed. As a result, it could not be completely excluded that, indeed, no lesions were present in these patients. Sixth, in the present study, DWI was not compared to other diagnostic tests such as stool-based tests and computed tomography colonography. In conclusion, DWI cannot yet be recommended in a clinical setting in which DWI is performed first and subsequent colonoscopy is only performed in patients with positive findings at DWI. Further (technical) developments are required to increase its diagnostic yield. References [1] Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA Cancer J Clin 2011;61:69–90. [2] Winawer SJ, Zauber AG, Ho MN, et al. Prevention of colorectal cancer by colonoscopic polypectomy. The National Polyp Study Workgroup. N Engl J Med 1993;329:1977–81. [3] Steinwachs D, Allen JD, Barlow WE, et al. National Institutes of Health state-of-the-science conference statement: enhancing use and quality of colorectal cancer screening. Ann Intern Med 2010;152:663–7. [4] U.S. Preventive Services Task Force. Screening for colorectal cancer: U.S. Preventive Services Task Force recommendation statement. Ann Intern Med 2008;149:627–37. [5] Takahara T, Imai Y, Yamashita T, Yasuda S, Nasu S, Van Cauteren M. Diffusion weighted whole body imaging with background body signal suppression (DWIBS): technical improvement using free breathing, STIR and high resolution 3D display. Radiat Med 2004;22:275–82. [6] Kwee TC, Takahara T, Ochiai R, Nievelstein RA, Luijten PR. Diffusion-weighted whole-body imaging with background body signal suppression (DWIBS): features and potential applications in oncology. Eur Radiol 2008;18:1937–52. [7] Hosonuma T, Tozaki M, Ichiba N, et al. Clinical usefulness of diffusion-weighted imaging using low and high b-values to detect rectal cancer. Magn Reson Med Sci 2006;5:173–7. [8] Ichikawa T, Erturk SM, Motosugi U, et al. High-B-value diffusionweighted MRI in colorectal cancer. AJR Am J Roentgenol 2006;187: 181–4. [9] Rao SX, Zeng MS, Chen CZ, et al. The value of diffusion-weighted imaging in combination with T2-weighted imaging for rectal cancer detection. Eur J Radiol 2008;65:299–303.
A.M. Leufkens et al. / Magnetic Resonance Imaging 31 (2013) 28–35 [10] Kwee TC, Takahara T, Niwa T, et al. Improving background suppression in diffusion-weighted imaging of the abdomen and pelvis using STIR with single-axis diffusion encoding. Magn Reson Imaging 2011;29:877–80. [11] Mürtz P, Kaschner M, Träber F, et al. Diffusion-weighted whole-body MRI with background body signal suppression: technical improvements at 3.0 T. J Magn Reson Imaging 2012;35:456–61. [12] Mürtz P, Kaschner M, Träber F, et al. Evaluation of dual-source parallel RF excitation for diffusion-weighted whole-body MR imaging with
35
background body signal suppression at 3.0T. Eur J Radiol 2011. Dec 15 [Epub ahead of print], http://dx.doi.org/10.1016/j.ejrad.2011.11.024. [13] Citarda F, Tomaselli G, Capocaccia R, Barcherini S, Crespi M, Italian Multicentre Study Group. Efficacy in standard clinical practice of colonoscopic polypectomy in reducing colorectal cancer incidence. Gut 2001;48:812–5. [14] Van Rijn JC, Reitsma JB, Stoker J, Bossuyt PM, van Deventer SJ, Dekker E. Polyp miss rate determined by tandem colonoscopy: a systematic review. Am J Gastroenterol 2006;101:343–50.