G Model EURR-7169; No. of Pages 9
ARTICLE IN PRESS European Journal of Radiology xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Focal liver lesions detection: Comparison of respiratory-triggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging Said El Bouchaibi a , Kenneth Coenegrachts b , Maria Antonietta Bali a , Julie Absil a , Thierry Metens a,∗ , Celso Matos a a b
Department of Radiology, Clinics of MRI, Hôpital Erasme Université Libre de Bruxelles, 808 Route de Lennik B, 1070 Bruxelles, Belgium Department of Radiology AZ Sint-Jan Brugge – Oostende AV, Campus Brugge, Ruddershove 10, B-8000 Brugge, Belgium
a r t i c l e
i n f o
Article history: Received 10 November 2014 Received in revised form 14 June 2015 Accepted 16 June 2015 Keywords: Liver Focal liver lesions Diffusion-weighted MRI Respiratory synchronization Navigators
a b s t r a c t Purpose: To compare low b value (10 s/mm2 ) spin-echo echo-planar (SE-EP) diffusion-weighted imaging (DWI) acquired with respiratory-triggering (RT), triggering and tracking navigator (TT), tracking only navigator (TRON) techniques for image quality and focal liver lesions (FLL) detection in non-cirrhotic patients. Material and methods: This bi-centric study was approved by the institutional review boards; informed consent was obtained. Eighty-three patients were prospectively included and SE-EP-DWI with RT, TT and TRON techniques were performed. DWI sequences were randomized and independently analyzed by two readers. The qualitative evaluation was based on a 3-point score for axial artifacts (motion, ghost, susceptibility artifacts and distortion) and stair-step artifacts. Sensitivity of FLL detection was calculated for all lesions together and after lesion size stratification (≤10 mm, >10–20 mm and >20 mm). The standard of reference consisted of a retrospective reading of the conventional MRI, the three DWI sequences and by follow-up (12 months): a total of 409 FLL were detected. Data between sequences was compared with non-parametric tests. Cohen’s kappa coefficient was used for inter-observer agreement. Results: Image quality was comparable for RT and TT. TRON showed statistically significantly more axial artifacts for the two readers (p < 0.05). Stair-step artifacts were not statistically significantly different between DWI sequences. Overall sensitivities for RT, TT, TRON were 85%, 86%, 82% and 86%, 89% 83%, respectively, for readers 1 and 2. The inter-observer agreement was very good. Conclusion: Image quality was better for RT and TT compared to TRON. Overall sensitivities for FLL detection were comparable between techniques and readers. © 2015 Elsevier Ireland Ltd. All rights reserved.
1. Introduction The detection and characterization of focal liver lesions (FLL), especially malignant primary and secondary tumors, are mandatory for treatment planification and patient prognostic assessment [1,2]. In this setting, magnetic resonance imaging (MRI) is the imaging tool of choice since it has shown higher sensitivity and
Abbreviations: BH, breath-hold; DWI, diffusion-weighted imaging; FLL, focal liver lesion; MRI, magnetic resonance imaging; RT, respiratory-triggering; SE-EP, spin-echo echo-planar; TRON, tracking only navigator; TT, trigerring and tracking. ∗ Corresponding author. Fax: +32 25553994. E-mail address: [email protected] (T. Metens).
specificity compared to multidetector CT and PET-CT, especially for the detection of liver metastases of gastro-intestinal tract tumors [3,4]. Diffusion-weighted imaging (DWI) is a MRI acquisition technique sensitive to the Brownian motion of water molecules. In the detection and characterisation of FLL, DWI has shown added value to conventional MRI sequences [4–8]. When spin-echo echo-planar (SE-EP) DWI is performed in the liver with low b values, i.e., equal or inferior to 50 s/mm2 , the signal of flowing spins is suppressed and therefore, the signal of liver vasculature is suppressed. This phenomena known as black-blood (BB) effect, may improve lesion conspicuity when compared to conventional T2-weighted images [9–14]. Thereby, in the setting of FLL detection, black-blood SE-
http://dx.doi.org/10.1016/j.ejrad.2015.06.018 0720-048X/© 2015 Elsevier Ireland Ltd. All rights reserved.
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model EURR-7169; No. of Pages 9
ARTICLE IN PRESS S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
2
EP-DWI may replace the T2-weighted sequence as suggested by Hussain et al. [10]. SE-EP-DWI sequences, with low b value, used in several published studies have been acquired either with respiratory triggered (RT) [10,11,14] or breath-hold (BH) techniques [9,12,13]. The RT approach synchronizes the image acquisition with the patient’s breathing cycle and acquires the imaging data during the end expiratory phase to avoid motion artifacts [15]. The BH technique has the advantage of being faster, however it has poorer signal-to-noise ratio, greater sensitivity to distortions and more ghosting artifacts compared to RT [15,16]. Technical advances have been recently implemented to improve image quality and to reduce acquisition time. The navigator based triggering is a respiratory-triggered prospective acquisition technique in which the position of the diaphragm is measured by the navigator-echo and triggered on the expiratory phase [14,17]. This approach has been compared to BH and free-breathing (FB) techniques for image quality and FLL detection [17,18]. The triggering and tracking (TT) technique is a navigator based triggering technique with an additional real time correction of the selection gradients in order to realign the slice position according to the diaphragm position at the end expiratory phase. Tracking only navigator (TRON) is a modified navigator-echo technique that allows real-time slice tracking and position correction during the entire respiratory cycle, without using any gating window. Few reports have been published evaluating the TRON, limited by a low number of subjects and assessing only image quality [19–21]. The aim of the present study was therefore to compare the image quality of SE-EP-DWI sequences acquired with a b value of 10 s/mm2 and performed with RT, TT and TRON techniques and to assess the sensitivity of these three SE-EP-DWI sequences in the detection of FLL.
Gadovist© , Bayer Healthcare, Leverkusen, Germany). The acquisition sequences included axial and coronal respiratory triggered T2-weighted turbo spin echo, axial T2-weighted turbo spin echo with spectral fat suppression, axial T1-weighted gradient-echo in and out of phase and axial three-dimensional T1-weighted gradient-echo with spectral fat suppression sequences (before and after administration of intravenous contrast agent). The three SE-EP-DWI sequences were acquired in the axial plane with a spectral fat-suppressed SE-EP sequence, b value of 10 s/mm2 in three orthogonal diffusion-sensitized directions and with four repetitions using the following parameters: TE 56 ms, EPI factor 63, half Fourier factor 0.60, Sense factor 2, anterior–posterior phase encoding direction, 30 slices with no interslice gap, voxel size 1.9 mm × 2.4 mm × 5 mm, and a field of view of 290 mm × 305 mm. SE-EP-DWI sequences were consecutively acquired just before administration of intravenous contrast agent following the same order for each patient: (1) RT sequence used an air-filled sensor placed on the hypocondrial region and fixed with an elastic belt. Data were acquired at the end- expiratory phase; (2) TT sequence using the navigator-echo to detect and to trigger the acquisition at the end expiratory phase and additionally tracking for the actual diaphragm position. For RT and TT data were acquired during 1200 ms. During these 1200 ms, 8 slices were consecutively acquired meaning that 4 packages were required to acquire the 30 slices. The mean acquisition time for RT and TT (considering a 4 s respiratory cycle duration) was estimated to 3 min 12 s; (3) TRON technique was used to track and correct for liver displacement, using a navigator-echo to measure the displacement of the diaphragm, between the liver and the lung and to update the gradients in order to align the slice position to the actual anatomical position (Tracking). Data were acquired during the entire respiratory cycle. A fixed TR of 4200 ms was used. The acquisition time was 50 s. All the acquired axial images were then reformatted in the coronal plane.
2. Material and methods This bi-centric study was approved by the institutional review boards of each institution. Written informed consent was obtained from all participants. 2.1. Study population During a 15-month period between March 2010 and June 2011, a total of 103 consecutive liver MRI investigations were prospectively conducted in the two institutions. The inclusion criteria for enrolment were the suspicion or presence of a FLL based on other imaging modalities (US, CT, PET/CT) and/or on the basis of laboratory findings (tumor markers, hepatic enzymes). Exclusion criteria were cirrhosis and general contraindication to MRI. Twenty patients were later excluded: seven patients because of an incomplete examination (at least one diffusion sequence and/or one conventional sequence was lacking) and thirteen patients because of cirrhosis. Finally 83 patients were included in this study, 47 men and 36 women (mean age: 60.6 years, age range: 27–83 years). 2.2. MRI protocol In the two institutions, MRI examinations were performed using a 1.5-T Unit (Achieva; Philips Medical System, Best, The Netherlands) equipped with a 16-channel phased-array surface coil and with parallel imaging capabilities (sensitivity encoding). All patients were placed in the magnet in a supine position. The conventional MRI protocol consisted of unenhanced and contrast-enhanced MRI-sequences after intravenous administration of gadolinium chelates (0.1 ml/kg) (Gadobutrol 1 mmol/ml,
2.3. Image analysis The three SE-EP-DWI sequences were randomized using a table of randomization (randomizer.org) and then transferred for analysis to an independent diagnostic workstation (Viewforum, Philips Medical System, Best, The Netherlands). The randomized SE-EPDWI sequences were analyzed by two independents readers: a senior radiologist, reader 1, and a radiologist in training, reader 2, with respectively, 13 years and 1 year of experience in liver MRI interpretation. Both readers were blind to the type of SE-EP-DWI sequence under evaluation. They were unaware of any information regarding patient history, clinical examination, laboratory and other imaging findings and the final diagnosis. One month before the beginning of the independent reading sessions, a learning consensus session took place. This session was performed using images from patients that were not included in the study and served to define axial artifacts (motion, ghost, susceptibility artifacts and distortions) and stair-step artifacts (i.e., stair-step-like deformation of organ contours, such as those of the liver and gallbladder, due to a mismatch between two sequential image acquisition) (Figs. 1 and 2). A 3-point score, grading axial and stair-step artifacts, was also defined: (1) good, when no artifacts at all were observed; (2) moderate, when minimal artifacts were present with little perturbations on interpretation; (3) poor, when serious artifacts with difficulties on interpretation were present. 2.4. Image quality analysis Image quality was assessed in the axial plane for the presence of motion, ghosts, susceptibility artifacts and distortions; and in the
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model EURR-7169; No. of Pages 9
ARTICLE IN PRESS S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
3
Fig. 1. Spin-echo echo-planar diffusion-weighted imaging (b value: 10 s/mm2 ), in the axial plane performed with (a) respiratory-triggering (RT), (b) triggering and tracking (TT), (c) tracking only navigator echo (TRON) techniques. No artifacts are visualized on the RT image, score 1. On the TT image, moderate artifacts are present, score 2. On the TRON image, severe artifacts are present, score 3. The focal liver lesion localised in the segment IV is better depicted and delimited on the RT image (arrow).
coronal reformatted images for stair-step artifacts using the 3-point score grading defined in the consensus lecture.
2.5. Lesion detection analysis On DW images FLL was defined as an intrahepatic hyperintense signal spot, after exclusion of residual signal from vessels, bile ducts or artifacts. The major diameter of each lesion and its location, according to the Couinaud classification, were recorded. When multiple lesions were present, only ten, those with the greatest diameter, were selected for analysis. A lesion detected on conventional MRI and SE-EP-DWI sequences was considered as true positive. A lesion seen on conventional MRI sequences and missed on a SE-EP-DWI sequence was considered as false negative. A lesion seen on a SE-EP-DWI sequence but not identified with the standard of reference was considered as false positive. Finally we provided the detection analysis when it is limited to malignant lesions only.
2.6. Standard of reference The standard of reference consisted of a lecture session performed by a third reader, radiologist with 25 years of experience in liver MRI interpretation, who independently from the other two readers identified the FLL on the basis of findings obtained from all the MRI sequences, i.e., unenhanced and contrast-enhanced conventional MRI and the three SE-EP-DWI sequences. Imaging follow-up at 12 months was further considered to confirm or exclude the presence of FLL. For each lesion, location and size were recorded. Lesions were further stratified according to their size in three sets: ≤10 mm, >10–20 mm, >20 mm. 2.7. Statistical analysis Two different software programs were used for statistical analysis: SPSS (Version 20, SPSS, Chicago, Illinois, USA) and MEDCALC (Version 12.5.0, Medcalc software, Ostend, Belgium). To assess image quality, a patient-based analysis was performed for axial and stair-step artifacts. The scores for each sequence were reported as mean and standard deviation for each reader. The
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model EURR-7169; No. of Pages 9
ARTICLE IN PRESS S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
4
Fig. 2. Coronal reformatted images from axial spin-echo echo-planar diffusion-weighted imaging (b value: 10 s/mm2 ) with (a) respiratory-triggering (RT), (b) triggering and tracking (TT), (c) tracking only navigator echo (TRON) techniques. On the RT image, severe stair- step artifacts are present, score 3. On the TT image, no artifacts are present, score 1. On the TRON image presence of moderate artifacts, score 2. The focal liver lesion is better depicted on the TT image (arrow).
non-parametric Wilcoxon test was used to identify significant differences between sequences for each reader. A linear Cohen’s kappa coefficient was computed to compare scores between the two readers for the same SE-EP-DWI sequence. The level of agreement k was considered poor if 20 mm; the mean size of the lesions was 14.1 mm (range of size: 3–100 mm). Sensitivities for FLL detection calculated for each reader and the inter-observer agreement for each sequence are shown in Table 3(a). Table 4(a) shows no statistically significant differences between the three SE-EP-DWI sequences sensitivities calculated for each reader, except between TRON and TT for both readers. When restricting the analysis to malignant lesions only, no statistically significant differences between the three SE-EP-DWI sequences sensitivities were demonstrated (Tables 3 and 4(b)) and sensitivities for detection remain high. Table 5 reports the sensitivity of the three SE-EP-DWI sequences calculated by the two readers after stratification according to the
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model
ARTICLE IN PRESS
EURR-7169; No. of Pages 9
S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
5
Table 1 Image quality assessment with comparisons for each reader. SE-EP-DWI Artifacts Axial artifacts Reader 1 Reader 2 Stair-step artifacts Reader 1 Reader 2
p of pairwise comparisons
RT
TT
TRON
RT/TT
RT/TRON
TT/TRON
1.30 ± 0.56 1.32 ± 0.56
1.37 ± 0.55 1.38 ± 0.56
1.75 ± 0.5 1.72 ± 0.5
0.25 0.35
10–20 mm >20 mm
Reader 1 77.4% (48/62) 16.1% (10/62) 6.5% (4/62)
Reader 2 77.6% (45/58) 15.5% (9/58) 6.9% (4/58)
Reader 1 78.9% (45/57) 15.8% (9/57) 5.3% (3/57)
Reader 2 73.9% (34/46) 19.6% (9/46) 6.5% (3/46)
Reader 1 74.7% (56/75) 18.7% (14/75) 6.7% (5/75)
Reader 2 76.8% (53/69) 20.3% (14/69) 2.9% (2/69)
SE-EP-DWI: spin-echo echo-planar diffusion-weighted imaging; RT: respiratory-triggering; TT: triggering and tracking; TRON: tracking only navigator.
of a biliary duct and, in the last case, was due to a non-saturated intrahepatic vessel. 4. Discussion The results of the present study demonstrate that between the three SE-EP-DWI sequences, the image quality with TRON, based on the axial artifacts, is statistically significantly lower compared to RT and TT for the two readers. TRON allows continuous real-time slice tracking and position correction due to the action of a navigator-echo that measures the displacement of the diaphragm between the liver and the lung, without using any gating window. Compared to RT and TT, the TRON technique samples continuously during the respiratory cycle
including the inspiration phase and is therefore much more prone to axial artifacts, explaining the lower image quality observed in our study. Our results are in disagreement with a previous study that reported an equal image quality between the TRON and RT sequences [20]. Takahara et al. evaluated the TRON sequence compared to RT and FB techniques, using a b value of 50 s/mm2 in young healthy volunteers. The main difference in our work compared to that of Takahara et al. is the study population. In their population, mean age was 31.7 years, however in our patient population it was 60.6 years. When compared to volunteers, our population may present an irregular respiratory cycle, thus generating more artifacts.
Fig. 3. (a) Axial contrast-enhanced 3D T1-weighted image acquired during portal phase showing two liver metastases in the segment VII (arrows). Axial spin-echo echo-planar diffusion-weighted imaging (SE-EP-DWI) (b value: 10 s/mm2 ) acquired with (b) respiratory triggering (RT), (c) triggering and tracking (TT) and (d) tracking only navigator echo (TRON) techniques. Susceptibility artifacts in the sub-phrenic areas are present on the three SE-EP DWI images (dotted arrows) but more importantly on TRON image. Both readers missed the more medial liver metastasis on TRON (arrowheads).
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model EURR-7169; No. of Pages 9
ARTICLE IN PRESS S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
7
Fig. 4. (a) Axial contrast-enhanced 3D T1-weighted image acquired during the arterial phase showing a nodular focal hyperplasia (FNH) in the segment VIII (arrow). Axial spin- echo echo-planar diffusion-weighted imaging (SE-EP-DWI) (b value: 10 s/mm2 ) acquired with (b) respiratory triggering (RT), (c) triggering and tracking (TT) and (d) tracking only navigator echo (TRON) techniques. The FNH was missed by the two readers on the three SE-EP DWI due to the low lesion-to-liver contrast.
In a study performed in healthy volunteers (mean age 33 years), Ivancevic et al. demonstrated, in agreement with our study, a lower image quality with TRON sequence compared to RT [21]. The image quality based on the axial artifacts is equivalent between RT and TT with, respectively, a score of 1.30 ± 0.56 and 1.37 ± 0.55 for reader 1 and 1.32 ± 0.56 and 1.38 ± 0.56 for reader 2. These sequences were acquired using, respectively, an air-filled sensor and a navigator-echo to trigger the end expiratory phase. These two SE-EP-DWI sequences were acquired using a sampling duration fixed at 1200 ms underlying the hypothesis that, with a regular respiratory cycle, the inspiration phase would not start during the sampling. The image quality assessment evaluated on the coronal reformatted images, based on the stair-step-like deformation, reported better values for TT compared to TRON and RT, even though not statistically significant. The combined effect in the TT technique of the sampling in end expiratory phase and the tracking due to the navigator-echo, may explain these observations. The absence of difference concerning the stair-step artifacts between the TRON and RT sequences in our study is in agreement with the results of the work of Takahara et al. [20]. However, our results are in dis-
agreement with the study of Ivancevic et al. reporting less stair-step artifacts with TRON in comparison with RT and FB. In their study, Ivancevic et al. have modified the original TRON sequence [21]. They established a maximum threshold of displacement of liver motion in the superior-inferior direction and, hence, all exceeding displacement detected by the navigator-echo was rejected. This modification could explain the lower level of stair-step artifacts of the TRON compared to the other SE-EP-DWI sequences. Furthermore, this modification in the TRON technique increases the acquisition time especially in patients with irregular breathing cycle. Our results for the stair-step artifacts demonstrate the effectiveness of the navigator-echo in the motion correction of the liver in the craniocaudal axis including the phases of inspiration and early expiration of the respiratory cycle. The absence of motion correction of liver in the anterior-posterior axis especially in the phase of inspiration and early expiration can explain the lower image quality in the axial plane with TRON compared to RT and TT. Our standard of reference consisted of a combined lecture of the conventional MRI and the three SE-EP-DWI sequences. Fiftysix lesions were retrospectively detected on conventional MRI
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model EURR-7169; No. of Pages 9 8
ARTICLE IN PRESS S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
sequences after integrating SE-EP-DWI sequences in the interpretation. This added value of SE-EP-DWI sequences in lesion detection has already been reported, especially when dealing with lesions of small size, i.e., inferior to 10 mm [8]. The overall sensitivity among the three SE-EP-DWI techniques is equivalent, though a statistically significant difference was observed for both readers between TRON and TT, with TT reporting a higher percentage. The per-lesion analysis performed after sizebased stratification also showed a statistically significant lower sensitivity with TRON compared to TT for reader 2 for the detection of lesions measuring ≤10 mm. For focal liver lesion detection, it is known that high image quality is mandatory. The lower image quality of TRON compared to RT and TT can explain the lower overall and after size-based stratification sensitivity of this technique. Globally, our results are in agreement with previous published data [12,22]. The majority of false negative lesions for both readers have a diameter ≤10 mm. Furthermore, most of those lesions are localized in regions of the liver that are more susceptible to susceptibility artifacts and distortions, such as close to the lung and stomach, as has been previously reported [11,23–25]. The missed FNH may be explained by the fact that the typical FNH has a cellular structure similar to that of normal hepatic parenchyma, and as a consequence there is a low lesion-to-liver contrast resolution [26,27]. Interestingly when the detection is limited to malignant lesions, the sensitivity remains elevated with no significant differences between techniques. However as high b value DW images were not included in the present study we cannot characterize the malignancy and therefore specificities could not be compared between techniques. It is known that DWI sequences artifacts can simulate suspicious lesions and result in false-positive findings as we experienced in our study [12,21,24]. In fifty percent of our cases, the error of interpretation can be explained by susceptibility artifacts that are especially present in sub-phrenic areas [12,24]. For extra-hepatic lesions, the most likely explanation is the lower signal-to-noise ratio and lower spatial resolution in the SE-EP-DWI sequences compared to the conventional MRI sequences [12,24]. Finally, we also reported a very high inter-observer agreement, suggesting that SE-EP-DWI sequences with low b value interpretation can be considered straightforward, with no possible intermediate solution. This simplicity in interpretation may represent an advantage for non-experienced radiologists [6]. There are several limitations to our study. We only have assessed the image quality on the presence or absence of qualitative criteria evaluated by the two readers. Even though the inter-observer agreement is very high, this approach is based on a subjective evaluation. A more objective approach could have been the estimation of the signal-to-noise ratio. We qualitatively evaluated the three SE-EP-DWI sequences and their sensitivity. This approach is not consistent with a clinical practice in which DWI is used not only for lesion detection but also for lesion characterization by performing DWI sequences with higher b values (>500 s/mm2 ) and apparent diffusion coefficient calculation. More generally, in non-cirrhotic liver the detection of liver lesions and possibly their characterization using DWI with both low and high b values should be compared to the use of Gd-EOB-DTPA, maybe even combined with it [28]. Finally, RT and TT techniques provided image quality advantages compared to the TRON approach, especially in older patients as in our study. In patients with an irregular respiratory cycle this superior image quality might be at the cost of 2–3 min longer acquisition time, which is worth, considering that robust lesion detection on DW images is an important pre-condition to the interpretation of all other MR sequences.
In conclusion, our results show that among the three SE-EP-DWI sequences with a b value of 10 s/mm2 , while RT and TT techniques are equivalent, the use of TRON induces lower image quality. Considering the overall sensitivity for focal liver lesion detection and the high inter-observer agreement, SE-EP-DWI sequences with low b value should be integrated in liver MRI studies. References [1] N. Kemeny, The management of resectable and unresectable liver metastases from colorectal cancer, Curr. Opin. Oncol. 22 (2010) 364–373. [2] A.E. Van der Pool, Z.S. Lalmahomed, J.H. De Wilt, A.M. Eggermont, J.N. Ijzermans, C. Verhoef, Trends in treatment for synchronous colorectal liver metastases: differences in outcome before and after 2000, J. Surg. Oncol. 102 (2010) 413–418. [3] P.P. Mainenti, M. Mancini, C. Mainolfi, L. Camera, S. Maurea, A. Manchia, M. Tanga, F. Persico, P. Addeo, D. D’Antonio, A. Speranza, L. Bucci, G. Persico, L. Pace, M. Salvatore, Detection of colo-rectal metastases: prospective comparison of contrast enhanced US, multidetector CT, PET/CT, and 1.5 tesla MR with extracellular and reticulo-endothelial cell specific contrast agents, Abdom. Imaging 35 (2010) 511–521. [4] E.D. Rappeport, A. Loft, A.K. Berthelsen, P. Von Der Recke, A. Noergaard Larsen, A.M. Mogensen, A. Wettergren, A. Rasmussen, J. Hillingsoe, P. Kirkegaard, C. Thomsen, Contrast-enhanced 18FDG-PET/CT vs. SPIO-enhanced MRI vs. 18FDG-PET vs. CT in patients with liver metastases from colorectal cancer: a prospective study with intraoperative confirmation, Acta Radiol. 4 (2007) 369–378. [5] K. Shimada, H. Isoda, Y. Hirokawa, A. Shigeki, S. Toshiya, T. Kaori, Comparison of gadolinium-EOB-DTPA enhanced and diffusion-weighted liver MRI for detection of small hepatic metastases, Eur. Radiol. 20 (2010) 2690–2698. [6] A.D. Hardie, M. Naik, E.M. Hecht, H. Chandarana, L. Mannelli, J.S. Babb, B. Taouli, Diagnosis of liver metastases: value of diffusion-weighted MRI compared with gadolinium-enhanced MRI, Eur. Radiol. 20 (2010) 1431–1441. [7] M.N. Park, Y.K. Kim, M.W. Lee, W.J. Lee, Y.S. Kim, S.H. Kim, D. Choi, H. Rhim, Small hepatocellular carcinomas: improved sensitivity by combining gadoxetic acid-enhanced and diffusion-weighted MR imaging patterns, Radiology 264 (2012) 761–770. [8] K. Holzapfel, M.J. Eiber, A.A. Fingerle, M. Bruegel, E.J. Rummeny, J. Gaa, Detection, classification, and characterization of focal liver lesions: value of diffusion-weighted MR imaging, gadoxetic acid-enhanced MR imaging and the combination of both methods, Abdom. Imaging 37 (2012) 74–82. [9] Y. Okada, K. Ohtomo, S. Kiryu, Y. Sasaki, Breath-hold T2-weighted MRI of hepatic tumors: value of echo planar imaging with diffusion-sensitizing gradient, J. Comput. Assisted Tomogr. 22 (1998) 364–371. [10] S.M. Hussain, J. De Becker, W.C. Hop, S. Dwarkasing, P.A. Wielopolski, Can a single-shot black-blood T2-weighted spin-echo echo-planar imaging sequence with sensitivity encoding replace the respiratory-triggered turbo spin-echo sequence for the liver? An optimization and feasibility study, JMRI 21 (2005) 219–229. [11] K. Coenegrachts, J. Delanote, L. Ter Beek, M. Haspeslagh, S. Bipat, J. Stoker, F. Van Kerkhove, L. Steyaert, H. Rigauts, J.W. Casselman, Improved focal liver lesion detection: comparison of single-shot diffusion-weighted echo-planar and single-shot T2 weighted turbo spin echo techniques, Br. J. Radiol. 80 (2007) 524–531. [12] C.J. Zech, K.A. Herrmann, O. Dietrich, W. Horger, M.F. Reiser, S.O. Schoenberg, Black-blood diffusion-weighted EPI acquisition of the liver with parallel imaging, Invest. Radiol. 43 (2008) 261–266. [13] T. Moteki, H. Horikoshi, N. Oya, J. Aoki, K. Endo, Evaluation of hepatic lesions and hepatic parenchyma using diffusion-weighted reordered turbo FLASH magnetic resonance images, JMRI 15 (2002) 564–572. [14] M. Bruegel, J. Gaa, S. Waldt, K. Woertler, K. Holzapfel, B. Kiefer, E.J. Rummeny, Diagnosis of hepatic metastasis: comparison of respiration-triggered diffusion-weighted echo-planar MRI and five T2-weighted turbo spin-echo sequences, AJR 191 (2008) 1421–1429. [15] S. Naganawa, H. Kawai, H. Fukatsu, Y. Sakurai, I. Aoki, S. Miura, T. Mimura, H. Kanazawa, T. Ishigaki, Diffusion-weighted imaging of the liver: technical challenges and prospects for the future, Magn. Reson. Med. Sci. 4 (2005) 175–186. [16] R.N. Low, G.D. Alzate, A. Shimakawa, Motion suppression in MR imaging of the liver: comparison of respiratory-triggered and non-triggered fast spin-echo sequences, AJR 168 (1997) 225–231. [17] B. Taouli, A. Sandberg, A. Stemmer, T. Rich, S. Wong, J. Xu, V.S. Lee, Diffusion-weighted image of the liver: comparison of navigator triggered and breath-hold acquisitions, JMRI 30 (2009) 561–568. [18] J.S. Choi, M.J. Kim, Y.E. Chung, K.A. Kim, J.Y. Choi, J.S. Lim, M.S. Park, K.W. Kim, Comparison of breath-hold, navigator-triggered, and free-breathing diffusion-weighted MRI for focal hepatic lesions, JMRI 38 (2013) 109–118. [19] T. Horie, T. Takahara, T. Ogino, T. Okuaki, M. Honda, Y. Okumura, N. Kajihara, K. Usui, I. Muro, Y. Imai, Trial of artifact reduction in body diffusion weighted imaging development and basic examination of «tracking only navigator» (TRON method), Nihon Hoshasen Gijutsu Gakkai Zasshi 64 (2008) 1157–1166. [20] T. Takahara, T.C. Kwee, S.M. Van Leeuwen, T. Ogino, T. Horie, M. Van Cauteren, G. Herigault, Y. Imai, W.P. Mali, P.R. Luijten, Diffusion-weighted magnetic
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model EURR-7169; No. of Pages 9
ARTICLE IN PRESS S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
[21]
[22]
[23] [24]
resonance imaging of the liver using tracking only navigator echo, Invest Radiol. 45 (2010) 57–63. K.M. Ivancevic, T.C. Kwee, T. Takahara, T. Ogino, H.K. Hussain, P.S. Liu, T.L. Chenevert, Diffusion-weighted MR imaging in the liver at 3.0 tesla using tracking only navigator echo (TRON): a feasibility study, JMRI 30 (2009) 1027–1033. T. Parikh, S.J. Drew, V.S. Lee, S. Wong, E.M. Hecht, J.S. Babb, B. Taouli, Focal liver lesion detection and characterization with diffusion-weighted MR imaging: comparisons with standard breath-hold T2-weighted imaging, Radiology 246 (2008) 812–822. D. Le Bihan, C. Poupon, A. Amadon, F. Lethimonnier, Artifacts and pitfalls in diffusion MRI, JMRI 24 (2006) 478–488. K.P. Pruessmann, M. Weiger, M.B. Scheidegger, P. Boesiger, SENSE sentivity encoding for fast MRI, Magn. Reson. Med. 42 (1999) 952–962.
9
[25] B. Taouli, D.M. Koh, Diffusion-weighted MR imaging of the liver, Radiology 254 (2010) 147–166. [26] F. Donati, P. Boraschi, R. Gigoni, S. Salemi, F. Falaschi, C. Bartolozzi, Focal nodular hyperplasia of the liver: diffusion and perfusion MRI characteristics, Magn. Res. Imaging 31 (2013) 10–16. [27] International Working Party, Terminology of nodular hepatocellular lesions of the liver, Hepatology 22 (1995) 983–993. [28] D. Lowenthal, M. Zeile, W.Y. Lim, C. Wybranski, F. Fischbach, G. Wieners, 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. 21 (2011) 832–840.
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
ARTICLE IN PRESS European Journal of Radiology xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad
Focal liver lesions detection: Comparison of respiratory-triggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging Said El Bouchaibi a , Kenneth Coenegrachts b , Maria Antonietta Bali a , Julie Absil a , Thierry Metens a,∗ , Celso Matos a a b
Department of Radiology, Clinics of MRI, Hôpital Erasme Université Libre de Bruxelles, 808 Route de Lennik B, 1070 Bruxelles, Belgium Department of Radiology AZ Sint-Jan Brugge – Oostende AV, Campus Brugge, Ruddershove 10, B-8000 Brugge, Belgium
a r t i c l e
i n f o
Article history: Received 10 November 2014 Received in revised form 14 June 2015 Accepted 16 June 2015 Keywords: Liver Focal liver lesions Diffusion-weighted MRI Respiratory synchronization Navigators
a b s t r a c t Purpose: To compare low b value (10 s/mm2 ) spin-echo echo-planar (SE-EP) diffusion-weighted imaging (DWI) acquired with respiratory-triggering (RT), triggering and tracking navigator (TT), tracking only navigator (TRON) techniques for image quality and focal liver lesions (FLL) detection in non-cirrhotic patients. Material and methods: This bi-centric study was approved by the institutional review boards; informed consent was obtained. Eighty-three patients were prospectively included and SE-EP-DWI with RT, TT and TRON techniques were performed. DWI sequences were randomized and independently analyzed by two readers. The qualitative evaluation was based on a 3-point score for axial artifacts (motion, ghost, susceptibility artifacts and distortion) and stair-step artifacts. Sensitivity of FLL detection was calculated for all lesions together and after lesion size stratification (≤10 mm, >10–20 mm and >20 mm). The standard of reference consisted of a retrospective reading of the conventional MRI, the three DWI sequences and by follow-up (12 months): a total of 409 FLL were detected. Data between sequences was compared with non-parametric tests. Cohen’s kappa coefficient was used for inter-observer agreement. Results: Image quality was comparable for RT and TT. TRON showed statistically significantly more axial artifacts for the two readers (p < 0.05). Stair-step artifacts were not statistically significantly different between DWI sequences. Overall sensitivities for RT, TT, TRON were 85%, 86%, 82% and 86%, 89% 83%, respectively, for readers 1 and 2. The inter-observer agreement was very good. Conclusion: Image quality was better for RT and TT compared to TRON. Overall sensitivities for FLL detection were comparable between techniques and readers. © 2015 Elsevier Ireland Ltd. All rights reserved.
1. Introduction The detection and characterization of focal liver lesions (FLL), especially malignant primary and secondary tumors, are mandatory for treatment planification and patient prognostic assessment [1,2]. In this setting, magnetic resonance imaging (MRI) is the imaging tool of choice since it has shown higher sensitivity and
Abbreviations: BH, breath-hold; DWI, diffusion-weighted imaging; FLL, focal liver lesion; MRI, magnetic resonance imaging; RT, respiratory-triggering; SE-EP, spin-echo echo-planar; TRON, tracking only navigator; TT, trigerring and tracking. ∗ Corresponding author. Fax: +32 25553994. E-mail address: [email protected] (T. Metens).
specificity compared to multidetector CT and PET-CT, especially for the detection of liver metastases of gastro-intestinal tract tumors [3,4]. Diffusion-weighted imaging (DWI) is a MRI acquisition technique sensitive to the Brownian motion of water molecules. In the detection and characterisation of FLL, DWI has shown added value to conventional MRI sequences [4–8]. When spin-echo echo-planar (SE-EP) DWI is performed in the liver with low b values, i.e., equal or inferior to 50 s/mm2 , the signal of flowing spins is suppressed and therefore, the signal of liver vasculature is suppressed. This phenomena known as black-blood (BB) effect, may improve lesion conspicuity when compared to conventional T2-weighted images [9–14]. Thereby, in the setting of FLL detection, black-blood SE-
http://dx.doi.org/10.1016/j.ejrad.2015.06.018 0720-048X/© 2015 Elsevier Ireland Ltd. All rights reserved.
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model EURR-7169; No. of Pages 9
ARTICLE IN PRESS S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
2
EP-DWI may replace the T2-weighted sequence as suggested by Hussain et al. [10]. SE-EP-DWI sequences, with low b value, used in several published studies have been acquired either with respiratory triggered (RT) [10,11,14] or breath-hold (BH) techniques [9,12,13]. The RT approach synchronizes the image acquisition with the patient’s breathing cycle and acquires the imaging data during the end expiratory phase to avoid motion artifacts [15]. The BH technique has the advantage of being faster, however it has poorer signal-to-noise ratio, greater sensitivity to distortions and more ghosting artifacts compared to RT [15,16]. Technical advances have been recently implemented to improve image quality and to reduce acquisition time. The navigator based triggering is a respiratory-triggered prospective acquisition technique in which the position of the diaphragm is measured by the navigator-echo and triggered on the expiratory phase [14,17]. This approach has been compared to BH and free-breathing (FB) techniques for image quality and FLL detection [17,18]. The triggering and tracking (TT) technique is a navigator based triggering technique with an additional real time correction of the selection gradients in order to realign the slice position according to the diaphragm position at the end expiratory phase. Tracking only navigator (TRON) is a modified navigator-echo technique that allows real-time slice tracking and position correction during the entire respiratory cycle, without using any gating window. Few reports have been published evaluating the TRON, limited by a low number of subjects and assessing only image quality [19–21]. The aim of the present study was therefore to compare the image quality of SE-EP-DWI sequences acquired with a b value of 10 s/mm2 and performed with RT, TT and TRON techniques and to assess the sensitivity of these three SE-EP-DWI sequences in the detection of FLL.
Gadovist© , Bayer Healthcare, Leverkusen, Germany). The acquisition sequences included axial and coronal respiratory triggered T2-weighted turbo spin echo, axial T2-weighted turbo spin echo with spectral fat suppression, axial T1-weighted gradient-echo in and out of phase and axial three-dimensional T1-weighted gradient-echo with spectral fat suppression sequences (before and after administration of intravenous contrast agent). The three SE-EP-DWI sequences were acquired in the axial plane with a spectral fat-suppressed SE-EP sequence, b value of 10 s/mm2 in three orthogonal diffusion-sensitized directions and with four repetitions using the following parameters: TE 56 ms, EPI factor 63, half Fourier factor 0.60, Sense factor 2, anterior–posterior phase encoding direction, 30 slices with no interslice gap, voxel size 1.9 mm × 2.4 mm × 5 mm, and a field of view of 290 mm × 305 mm. SE-EP-DWI sequences were consecutively acquired just before administration of intravenous contrast agent following the same order for each patient: (1) RT sequence used an air-filled sensor placed on the hypocondrial region and fixed with an elastic belt. Data were acquired at the end- expiratory phase; (2) TT sequence using the navigator-echo to detect and to trigger the acquisition at the end expiratory phase and additionally tracking for the actual diaphragm position. For RT and TT data were acquired during 1200 ms. During these 1200 ms, 8 slices were consecutively acquired meaning that 4 packages were required to acquire the 30 slices. The mean acquisition time for RT and TT (considering a 4 s respiratory cycle duration) was estimated to 3 min 12 s; (3) TRON technique was used to track and correct for liver displacement, using a navigator-echo to measure the displacement of the diaphragm, between the liver and the lung and to update the gradients in order to align the slice position to the actual anatomical position (Tracking). Data were acquired during the entire respiratory cycle. A fixed TR of 4200 ms was used. The acquisition time was 50 s. All the acquired axial images were then reformatted in the coronal plane.
2. Material and methods This bi-centric study was approved by the institutional review boards of each institution. Written informed consent was obtained from all participants. 2.1. Study population During a 15-month period between March 2010 and June 2011, a total of 103 consecutive liver MRI investigations were prospectively conducted in the two institutions. The inclusion criteria for enrolment were the suspicion or presence of a FLL based on other imaging modalities (US, CT, PET/CT) and/or on the basis of laboratory findings (tumor markers, hepatic enzymes). Exclusion criteria were cirrhosis and general contraindication to MRI. Twenty patients were later excluded: seven patients because of an incomplete examination (at least one diffusion sequence and/or one conventional sequence was lacking) and thirteen patients because of cirrhosis. Finally 83 patients were included in this study, 47 men and 36 women (mean age: 60.6 years, age range: 27–83 years). 2.2. MRI protocol In the two institutions, MRI examinations were performed using a 1.5-T Unit (Achieva; Philips Medical System, Best, The Netherlands) equipped with a 16-channel phased-array surface coil and with parallel imaging capabilities (sensitivity encoding). All patients were placed in the magnet in a supine position. The conventional MRI protocol consisted of unenhanced and contrast-enhanced MRI-sequences after intravenous administration of gadolinium chelates (0.1 ml/kg) (Gadobutrol 1 mmol/ml,
2.3. Image analysis The three SE-EP-DWI sequences were randomized using a table of randomization (randomizer.org) and then transferred for analysis to an independent diagnostic workstation (Viewforum, Philips Medical System, Best, The Netherlands). The randomized SE-EPDWI sequences were analyzed by two independents readers: a senior radiologist, reader 1, and a radiologist in training, reader 2, with respectively, 13 years and 1 year of experience in liver MRI interpretation. Both readers were blind to the type of SE-EP-DWI sequence under evaluation. They were unaware of any information regarding patient history, clinical examination, laboratory and other imaging findings and the final diagnosis. One month before the beginning of the independent reading sessions, a learning consensus session took place. This session was performed using images from patients that were not included in the study and served to define axial artifacts (motion, ghost, susceptibility artifacts and distortions) and stair-step artifacts (i.e., stair-step-like deformation of organ contours, such as those of the liver and gallbladder, due to a mismatch between two sequential image acquisition) (Figs. 1 and 2). A 3-point score, grading axial and stair-step artifacts, was also defined: (1) good, when no artifacts at all were observed; (2) moderate, when minimal artifacts were present with little perturbations on interpretation; (3) poor, when serious artifacts with difficulties on interpretation were present. 2.4. Image quality analysis Image quality was assessed in the axial plane for the presence of motion, ghosts, susceptibility artifacts and distortions; and in the
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model EURR-7169; No. of Pages 9
ARTICLE IN PRESS S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
3
Fig. 1. Spin-echo echo-planar diffusion-weighted imaging (b value: 10 s/mm2 ), in the axial plane performed with (a) respiratory-triggering (RT), (b) triggering and tracking (TT), (c) tracking only navigator echo (TRON) techniques. No artifacts are visualized on the RT image, score 1. On the TT image, moderate artifacts are present, score 2. On the TRON image, severe artifacts are present, score 3. The focal liver lesion localised in the segment IV is better depicted and delimited on the RT image (arrow).
coronal reformatted images for stair-step artifacts using the 3-point score grading defined in the consensus lecture.
2.5. Lesion detection analysis On DW images FLL was defined as an intrahepatic hyperintense signal spot, after exclusion of residual signal from vessels, bile ducts or artifacts. The major diameter of each lesion and its location, according to the Couinaud classification, were recorded. When multiple lesions were present, only ten, those with the greatest diameter, were selected for analysis. A lesion detected on conventional MRI and SE-EP-DWI sequences was considered as true positive. A lesion seen on conventional MRI sequences and missed on a SE-EP-DWI sequence was considered as false negative. A lesion seen on a SE-EP-DWI sequence but not identified with the standard of reference was considered as false positive. Finally we provided the detection analysis when it is limited to malignant lesions only.
2.6. Standard of reference The standard of reference consisted of a lecture session performed by a third reader, radiologist with 25 years of experience in liver MRI interpretation, who independently from the other two readers identified the FLL on the basis of findings obtained from all the MRI sequences, i.e., unenhanced and contrast-enhanced conventional MRI and the three SE-EP-DWI sequences. Imaging follow-up at 12 months was further considered to confirm or exclude the presence of FLL. For each lesion, location and size were recorded. Lesions were further stratified according to their size in three sets: ≤10 mm, >10–20 mm, >20 mm. 2.7. Statistical analysis Two different software programs were used for statistical analysis: SPSS (Version 20, SPSS, Chicago, Illinois, USA) and MEDCALC (Version 12.5.0, Medcalc software, Ostend, Belgium). To assess image quality, a patient-based analysis was performed for axial and stair-step artifacts. The scores for each sequence were reported as mean and standard deviation for each reader. The
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model EURR-7169; No. of Pages 9
ARTICLE IN PRESS S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
4
Fig. 2. Coronal reformatted images from axial spin-echo echo-planar diffusion-weighted imaging (b value: 10 s/mm2 ) with (a) respiratory-triggering (RT), (b) triggering and tracking (TT), (c) tracking only navigator echo (TRON) techniques. On the RT image, severe stair- step artifacts are present, score 3. On the TT image, no artifacts are present, score 1. On the TRON image presence of moderate artifacts, score 2. The focal liver lesion is better depicted on the TT image (arrow).
non-parametric Wilcoxon test was used to identify significant differences between sequences for each reader. A linear Cohen’s kappa coefficient was computed to compare scores between the two readers for the same SE-EP-DWI sequence. The level of agreement k was considered poor if 20 mm; the mean size of the lesions was 14.1 mm (range of size: 3–100 mm). Sensitivities for FLL detection calculated for each reader and the inter-observer agreement for each sequence are shown in Table 3(a). Table 4(a) shows no statistically significant differences between the three SE-EP-DWI sequences sensitivities calculated for each reader, except between TRON and TT for both readers. When restricting the analysis to malignant lesions only, no statistically significant differences between the three SE-EP-DWI sequences sensitivities were demonstrated (Tables 3 and 4(b)) and sensitivities for detection remain high. Table 5 reports the sensitivity of the three SE-EP-DWI sequences calculated by the two readers after stratification according to the
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model
ARTICLE IN PRESS
EURR-7169; No. of Pages 9
S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
5
Table 1 Image quality assessment with comparisons for each reader. SE-EP-DWI Artifacts Axial artifacts Reader 1 Reader 2 Stair-step artifacts Reader 1 Reader 2
p of pairwise comparisons
RT
TT
TRON
RT/TT
RT/TRON
TT/TRON
1.30 ± 0.56 1.32 ± 0.56
1.37 ± 0.55 1.38 ± 0.56
1.75 ± 0.5 1.72 ± 0.5
0.25 0.35
10–20 mm >20 mm
Reader 1 77.4% (48/62) 16.1% (10/62) 6.5% (4/62)
Reader 2 77.6% (45/58) 15.5% (9/58) 6.9% (4/58)
Reader 1 78.9% (45/57) 15.8% (9/57) 5.3% (3/57)
Reader 2 73.9% (34/46) 19.6% (9/46) 6.5% (3/46)
Reader 1 74.7% (56/75) 18.7% (14/75) 6.7% (5/75)
Reader 2 76.8% (53/69) 20.3% (14/69) 2.9% (2/69)
SE-EP-DWI: spin-echo echo-planar diffusion-weighted imaging; RT: respiratory-triggering; TT: triggering and tracking; TRON: tracking only navigator.
of a biliary duct and, in the last case, was due to a non-saturated intrahepatic vessel. 4. Discussion The results of the present study demonstrate that between the three SE-EP-DWI sequences, the image quality with TRON, based on the axial artifacts, is statistically significantly lower compared to RT and TT for the two readers. TRON allows continuous real-time slice tracking and position correction due to the action of a navigator-echo that measures the displacement of the diaphragm between the liver and the lung, without using any gating window. Compared to RT and TT, the TRON technique samples continuously during the respiratory cycle
including the inspiration phase and is therefore much more prone to axial artifacts, explaining the lower image quality observed in our study. Our results are in disagreement with a previous study that reported an equal image quality between the TRON and RT sequences [20]. Takahara et al. evaluated the TRON sequence compared to RT and FB techniques, using a b value of 50 s/mm2 in young healthy volunteers. The main difference in our work compared to that of Takahara et al. is the study population. In their population, mean age was 31.7 years, however in our patient population it was 60.6 years. When compared to volunteers, our population may present an irregular respiratory cycle, thus generating more artifacts.
Fig. 3. (a) Axial contrast-enhanced 3D T1-weighted image acquired during portal phase showing two liver metastases in the segment VII (arrows). Axial spin-echo echo-planar diffusion-weighted imaging (SE-EP-DWI) (b value: 10 s/mm2 ) acquired with (b) respiratory triggering (RT), (c) triggering and tracking (TT) and (d) tracking only navigator echo (TRON) techniques. Susceptibility artifacts in the sub-phrenic areas are present on the three SE-EP DWI images (dotted arrows) but more importantly on TRON image. Both readers missed the more medial liver metastasis on TRON (arrowheads).
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model EURR-7169; No. of Pages 9
ARTICLE IN PRESS S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
7
Fig. 4. (a) Axial contrast-enhanced 3D T1-weighted image acquired during the arterial phase showing a nodular focal hyperplasia (FNH) in the segment VIII (arrow). Axial spin- echo echo-planar diffusion-weighted imaging (SE-EP-DWI) (b value: 10 s/mm2 ) acquired with (b) respiratory triggering (RT), (c) triggering and tracking (TT) and (d) tracking only navigator echo (TRON) techniques. The FNH was missed by the two readers on the three SE-EP DWI due to the low lesion-to-liver contrast.
In a study performed in healthy volunteers (mean age 33 years), Ivancevic et al. demonstrated, in agreement with our study, a lower image quality with TRON sequence compared to RT [21]. The image quality based on the axial artifacts is equivalent between RT and TT with, respectively, a score of 1.30 ± 0.56 and 1.37 ± 0.55 for reader 1 and 1.32 ± 0.56 and 1.38 ± 0.56 for reader 2. These sequences were acquired using, respectively, an air-filled sensor and a navigator-echo to trigger the end expiratory phase. These two SE-EP-DWI sequences were acquired using a sampling duration fixed at 1200 ms underlying the hypothesis that, with a regular respiratory cycle, the inspiration phase would not start during the sampling. The image quality assessment evaluated on the coronal reformatted images, based on the stair-step-like deformation, reported better values for TT compared to TRON and RT, even though not statistically significant. The combined effect in the TT technique of the sampling in end expiratory phase and the tracking due to the navigator-echo, may explain these observations. The absence of difference concerning the stair-step artifacts between the TRON and RT sequences in our study is in agreement with the results of the work of Takahara et al. [20]. However, our results are in dis-
agreement with the study of Ivancevic et al. reporting less stair-step artifacts with TRON in comparison with RT and FB. In their study, Ivancevic et al. have modified the original TRON sequence [21]. They established a maximum threshold of displacement of liver motion in the superior-inferior direction and, hence, all exceeding displacement detected by the navigator-echo was rejected. This modification could explain the lower level of stair-step artifacts of the TRON compared to the other SE-EP-DWI sequences. Furthermore, this modification in the TRON technique increases the acquisition time especially in patients with irregular breathing cycle. Our results for the stair-step artifacts demonstrate the effectiveness of the navigator-echo in the motion correction of the liver in the craniocaudal axis including the phases of inspiration and early expiration of the respiratory cycle. The absence of motion correction of liver in the anterior-posterior axis especially in the phase of inspiration and early expiration can explain the lower image quality in the axial plane with TRON compared to RT and TT. Our standard of reference consisted of a combined lecture of the conventional MRI and the three SE-EP-DWI sequences. Fiftysix lesions were retrospectively detected on conventional MRI
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model EURR-7169; No. of Pages 9 8
ARTICLE IN PRESS S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
sequences after integrating SE-EP-DWI sequences in the interpretation. This added value of SE-EP-DWI sequences in lesion detection has already been reported, especially when dealing with lesions of small size, i.e., inferior to 10 mm [8]. The overall sensitivity among the three SE-EP-DWI techniques is equivalent, though a statistically significant difference was observed for both readers between TRON and TT, with TT reporting a higher percentage. The per-lesion analysis performed after sizebased stratification also showed a statistically significant lower sensitivity with TRON compared to TT for reader 2 for the detection of lesions measuring ≤10 mm. For focal liver lesion detection, it is known that high image quality is mandatory. The lower image quality of TRON compared to RT and TT can explain the lower overall and after size-based stratification sensitivity of this technique. Globally, our results are in agreement with previous published data [12,22]. The majority of false negative lesions for both readers have a diameter ≤10 mm. Furthermore, most of those lesions are localized in regions of the liver that are more susceptible to susceptibility artifacts and distortions, such as close to the lung and stomach, as has been previously reported [11,23–25]. The missed FNH may be explained by the fact that the typical FNH has a cellular structure similar to that of normal hepatic parenchyma, and as a consequence there is a low lesion-to-liver contrast resolution [26,27]. Interestingly when the detection is limited to malignant lesions, the sensitivity remains elevated with no significant differences between techniques. However as high b value DW images were not included in the present study we cannot characterize the malignancy and therefore specificities could not be compared between techniques. It is known that DWI sequences artifacts can simulate suspicious lesions and result in false-positive findings as we experienced in our study [12,21,24]. In fifty percent of our cases, the error of interpretation can be explained by susceptibility artifacts that are especially present in sub-phrenic areas [12,24]. For extra-hepatic lesions, the most likely explanation is the lower signal-to-noise ratio and lower spatial resolution in the SE-EP-DWI sequences compared to the conventional MRI sequences [12,24]. Finally, we also reported a very high inter-observer agreement, suggesting that SE-EP-DWI sequences with low b value interpretation can be considered straightforward, with no possible intermediate solution. This simplicity in interpretation may represent an advantage for non-experienced radiologists [6]. There are several limitations to our study. We only have assessed the image quality on the presence or absence of qualitative criteria evaluated by the two readers. Even though the inter-observer agreement is very high, this approach is based on a subjective evaluation. A more objective approach could have been the estimation of the signal-to-noise ratio. We qualitatively evaluated the three SE-EP-DWI sequences and their sensitivity. This approach is not consistent with a clinical practice in which DWI is used not only for lesion detection but also for lesion characterization by performing DWI sequences with higher b values (>500 s/mm2 ) and apparent diffusion coefficient calculation. More generally, in non-cirrhotic liver the detection of liver lesions and possibly their characterization using DWI with both low and high b values should be compared to the use of Gd-EOB-DTPA, maybe even combined with it [28]. Finally, RT and TT techniques provided image quality advantages compared to the TRON approach, especially in older patients as in our study. In patients with an irregular respiratory cycle this superior image quality might be at the cost of 2–3 min longer acquisition time, which is worth, considering that robust lesion detection on DW images is an important pre-condition to the interpretation of all other MR sequences.
In conclusion, our results show that among the three SE-EP-DWI sequences with a b value of 10 s/mm2 , while RT and TT techniques are equivalent, the use of TRON induces lower image quality. Considering the overall sensitivity for focal liver lesion detection and the high inter-observer agreement, SE-EP-DWI sequences with low b value should be integrated in liver MRI studies. References [1] N. Kemeny, The management of resectable and unresectable liver metastases from colorectal cancer, Curr. Opin. Oncol. 22 (2010) 364–373. [2] A.E. Van der Pool, Z.S. Lalmahomed, J.H. De Wilt, A.M. Eggermont, J.N. Ijzermans, C. Verhoef, Trends in treatment for synchronous colorectal liver metastases: differences in outcome before and after 2000, J. Surg. Oncol. 102 (2010) 413–418. [3] P.P. Mainenti, M. Mancini, C. Mainolfi, L. Camera, S. Maurea, A. Manchia, M. Tanga, F. Persico, P. Addeo, D. D’Antonio, A. Speranza, L. Bucci, G. Persico, L. Pace, M. Salvatore, Detection of colo-rectal metastases: prospective comparison of contrast enhanced US, multidetector CT, PET/CT, and 1.5 tesla MR with extracellular and reticulo-endothelial cell specific contrast agents, Abdom. Imaging 35 (2010) 511–521. [4] E.D. Rappeport, A. Loft, A.K. Berthelsen, P. Von Der Recke, A. Noergaard Larsen, A.M. Mogensen, A. Wettergren, A. Rasmussen, J. Hillingsoe, P. Kirkegaard, C. Thomsen, Contrast-enhanced 18FDG-PET/CT vs. SPIO-enhanced MRI vs. 18FDG-PET vs. CT in patients with liver metastases from colorectal cancer: a prospective study with intraoperative confirmation, Acta Radiol. 4 (2007) 369–378. [5] K. Shimada, H. Isoda, Y. Hirokawa, A. Shigeki, S. Toshiya, T. Kaori, Comparison of gadolinium-EOB-DTPA enhanced and diffusion-weighted liver MRI for detection of small hepatic metastases, Eur. Radiol. 20 (2010) 2690–2698. [6] A.D. Hardie, M. Naik, E.M. Hecht, H. Chandarana, L. Mannelli, J.S. Babb, B. Taouli, Diagnosis of liver metastases: value of diffusion-weighted MRI compared with gadolinium-enhanced MRI, Eur. Radiol. 20 (2010) 1431–1441. [7] M.N. Park, Y.K. Kim, M.W. Lee, W.J. Lee, Y.S. Kim, S.H. Kim, D. Choi, H. Rhim, Small hepatocellular carcinomas: improved sensitivity by combining gadoxetic acid-enhanced and diffusion-weighted MR imaging patterns, Radiology 264 (2012) 761–770. [8] K. Holzapfel, M.J. Eiber, A.A. Fingerle, M. Bruegel, E.J. Rummeny, J. Gaa, Detection, classification, and characterization of focal liver lesions: value of diffusion-weighted MR imaging, gadoxetic acid-enhanced MR imaging and the combination of both methods, Abdom. Imaging 37 (2012) 74–82. [9] Y. Okada, K. Ohtomo, S. Kiryu, Y. Sasaki, Breath-hold T2-weighted MRI of hepatic tumors: value of echo planar imaging with diffusion-sensitizing gradient, J. Comput. Assisted Tomogr. 22 (1998) 364–371. [10] S.M. Hussain, J. De Becker, W.C. Hop, S. Dwarkasing, P.A. Wielopolski, Can a single-shot black-blood T2-weighted spin-echo echo-planar imaging sequence with sensitivity encoding replace the respiratory-triggered turbo spin-echo sequence for the liver? An optimization and feasibility study, JMRI 21 (2005) 219–229. [11] K. Coenegrachts, J. Delanote, L. Ter Beek, M. Haspeslagh, S. Bipat, J. Stoker, F. Van Kerkhove, L. Steyaert, H. Rigauts, J.W. Casselman, Improved focal liver lesion detection: comparison of single-shot diffusion-weighted echo-planar and single-shot T2 weighted turbo spin echo techniques, Br. J. Radiol. 80 (2007) 524–531. [12] C.J. Zech, K.A. Herrmann, O. Dietrich, W. Horger, M.F. Reiser, S.O. Schoenberg, Black-blood diffusion-weighted EPI acquisition of the liver with parallel imaging, Invest. Radiol. 43 (2008) 261–266. [13] T. Moteki, H. Horikoshi, N. Oya, J. Aoki, K. Endo, Evaluation of hepatic lesions and hepatic parenchyma using diffusion-weighted reordered turbo FLASH magnetic resonance images, JMRI 15 (2002) 564–572. [14] M. Bruegel, J. Gaa, S. Waldt, K. Woertler, K. Holzapfel, B. Kiefer, E.J. Rummeny, Diagnosis of hepatic metastasis: comparison of respiration-triggered diffusion-weighted echo-planar MRI and five T2-weighted turbo spin-echo sequences, AJR 191 (2008) 1421–1429. [15] S. Naganawa, H. Kawai, H. Fukatsu, Y. Sakurai, I. Aoki, S. Miura, T. Mimura, H. Kanazawa, T. Ishigaki, Diffusion-weighted imaging of the liver: technical challenges and prospects for the future, Magn. Reson. Med. Sci. 4 (2005) 175–186. [16] R.N. Low, G.D. Alzate, A. Shimakawa, Motion suppression in MR imaging of the liver: comparison of respiratory-triggered and non-triggered fast spin-echo sequences, AJR 168 (1997) 225–231. [17] B. Taouli, A. Sandberg, A. Stemmer, T. Rich, S. Wong, J. Xu, V.S. Lee, Diffusion-weighted image of the liver: comparison of navigator triggered and breath-hold acquisitions, JMRI 30 (2009) 561–568. [18] J.S. Choi, M.J. Kim, Y.E. Chung, K.A. Kim, J.Y. Choi, J.S. Lim, M.S. Park, K.W. Kim, Comparison of breath-hold, navigator-triggered, and free-breathing diffusion-weighted MRI for focal hepatic lesions, JMRI 38 (2013) 109–118. [19] T. Horie, T. Takahara, T. Ogino, T. Okuaki, M. Honda, Y. Okumura, N. Kajihara, K. Usui, I. Muro, Y. Imai, Trial of artifact reduction in body diffusion weighted imaging development and basic examination of «tracking only navigator» (TRON method), Nihon Hoshasen Gijutsu Gakkai Zasshi 64 (2008) 1157–1166. [20] T. Takahara, T.C. Kwee, S.M. Van Leeuwen, T. Ogino, T. Horie, M. Van Cauteren, G. Herigault, Y. Imai, W.P. Mali, P.R. Luijten, Diffusion-weighted magnetic
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
G Model EURR-7169; No. of Pages 9
ARTICLE IN PRESS S.E. Bouchaibi et al. / European Journal of Radiology xxx (2015) xxx–xxx
[21]
[22]
[23] [24]
resonance imaging of the liver using tracking only navigator echo, Invest Radiol. 45 (2010) 57–63. K.M. Ivancevic, T.C. Kwee, T. Takahara, T. Ogino, H.K. Hussain, P.S. Liu, T.L. Chenevert, Diffusion-weighted MR imaging in the liver at 3.0 tesla using tracking only navigator echo (TRON): a feasibility study, JMRI 30 (2009) 1027–1033. T. Parikh, S.J. Drew, V.S. Lee, S. Wong, E.M. Hecht, J.S. Babb, B. Taouli, Focal liver lesion detection and characterization with diffusion-weighted MR imaging: comparisons with standard breath-hold T2-weighted imaging, Radiology 246 (2008) 812–822. D. Le Bihan, C. Poupon, A. Amadon, F. Lethimonnier, Artifacts and pitfalls in diffusion MRI, JMRI 24 (2006) 478–488. K.P. Pruessmann, M. Weiger, M.B. Scheidegger, P. Boesiger, SENSE sentivity encoding for fast MRI, Magn. Reson. Med. 42 (1999) 952–962.
9
[25] B. Taouli, D.M. Koh, Diffusion-weighted MR imaging of the liver, Radiology 254 (2010) 147–166. [26] F. Donati, P. Boraschi, R. Gigoni, S. Salemi, F. Falaschi, C. Bartolozzi, Focal nodular hyperplasia of the liver: diffusion and perfusion MRI characteristics, Magn. Res. Imaging 31 (2013) 10–16. [27] International Working Party, Terminology of nodular hepatocellular lesions of the liver, Hepatology 22 (1995) 983–993. [28] D. Lowenthal, M. Zeile, W.Y. Lim, C. Wybranski, F. Fischbach, G. Wieners, 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. 21 (2011) 832–840.
Please cite this article in press as: S.E. Bouchaibi, et al., Focal liver lesions detection: Comparison of respiratorytriggering, triggering and tracking navigator and tracking-only navigator in diffusion-weighted imaging, Eur J Radiol (2015), http://dx.doi.org/10.1016/j.ejrad.2015.06.018
