Comparison of high-resolution T1W 3D GRE (LAVA) with 2-point Dixon fat_water separation (FLEX) to T1W fast spin echo (FSE) in prostate cancer (PCa)

Clinical Imaging 40 (2016) 407413Contents lists available at ScienceDirectClinical Imagingjournal homepage: Original ArticleComparison of high-resolution T1W 3D GRE (LAVA) with 2-point Dixonfat/water separation (FLEX) to T1W fast spin echo (FSE) in prostatecancer (PCa)Karim Samji a, Abdulmohsen Alrashed a, Wael M Shabana a, Matthew DF McInnes a,Ersin Bayram b, Nicola Schieda a,abDepartment of Diagnostic Imaging, The Ottawa Hospital, 1053 Carling Avenue, Ottawa, ON, Canada K1Y 4E9Global MR Applications and Workow, GE Healthcare, Houston, TX, USAa r t i c l e i n f o a b s t r a c tArticle history:Received 9 August 2015Received in revised form 10 November 2015Accepted 19 November 2015Keywords:Prostate cancermagnetic resonance imaging MRIT1-weighted imagingDixonmetastases1. IntroductionPurpose: To compare T1-weighted (T1W) fast spin echo (FSE) to T1W 3-dimensional gradient recalled echo(LAVA) with fat water separation (FLEX) in prostate cancer (PCa).Methodology: With institutional review board waiver, 39 patients underwent 3-T magnetic resonance imagingincluding T1W LAVA FLEX (157 s)/T1W FSE (316 s). Two radiologists assessed (a) image quality/sharpness,(b) presence/severity of artifacts, and (c) skeletal (N=22)/nodal (N=9) metastases. Results were comparedusing Wilcoxon signed-rank test/receiver operator characteristic analysis.Results: With T1W LAVA FLEX, image quality/sharpness improved (Pb.001) with less motion (P=.002.03) andno difference in phase-encoding artifact (PN.05). One patient had moderate fat/water swap.Detection of skeletal metastases was unchanged (PN.05) and nodal metastases either improved (P=.002) orwere comparable (P=.16) using T1W LAVA FLEX.Conclusion: T1W LAVA FLEX improves image quality, lessens motion artifact, and is comparable or improves de-tection of metastases in PCa with reduction in acquisition time. 2016 Elsevier Inc. All rights reserved.with other sequences for detection of neurovascular bundle invasion(Fig. 2). Traditionally, T1W imaging of the pelvis is performed usingT1-weighted (T1W) imaging is an integral component of a compre-hensive magnetic resonance (MR) protocol for the evaluation of pros-tate cancer (PCa) 1. Within the prostate, T1W images provide areference for diagnosis of postbiopsy hemorrhage, which is a knownmimic of PCa on T2-weighted (T2W) imaging 2. Furthermore, T1Wimaging can be used to detect areas of cancer in men with diffusepostbiopsy hemorrhage. In the peripheral zone, an area that is relativelyspared of hemorrhage and shows corresponding low T2W signal inten-sity (SI) is highly specic for PCa (“MR exclusion sign”) 3 (Fig. 1). Largeeld of view (FOV) T1W images are important for the detection of nodalmetastases 4 and may also be used to detect or conrm skeletal metas-tases, with the latter diagnosis typically interpreted in conjunction withbone scan 5,6. In our experience, T1W images also provide excellentdepiction of the neurovascular bundles and can be used in combination There was no funding provided for the organization of this manuscript. Author EB is aGE employee, and all other authors have no disclosures. Corresponding author. Department of Diagnostic Imaging, The Ottawa Hospital, 1053Carling Avenue, Ottawa, ON, Canada K1Y 4E9. Tel.: +1-613-761-4054; fax: +1-613-761-4476.E-mail address: nschiedatoh.on.ca (N. Schieda)./10.1016/j.clinimag.2015.11.0230899-7071/ 2016 Elsevier Inc. All rights reserved.fast spin echo (FSE)/turbo spin echo. T1W FSE provides robust imagingquality but is relatively time consuming, particularly when the wholepelvis is imaged from the distal aorta to the ischial tuberosities to assessfor possible metastatic disease.A 2-point Dixon T1W in-phase and opposed-phase gradient recalledecho (GRE) imaging that automatically generates fat-only (FO) andwater-only (WO) image sets is widely available on most commercialMR systems. Recently, improved postprocessing algorithms (which cor-rect errors related to phase differences that accumulate between in-phase and opposed-phase echoes) have signicantly improved thequality of FO and WO imaging 7. Several studies have shown compara-ble or improved image quality comparing conventional (chemical) T1Wfat-suppressed (FS) GRE to FS T1W GRE obtained using fat/water sepa-ration 810. Moreover, the use of the FO image set has been proposedas a rapid method to evaluate the skeleton for metastatic disease 11. A2-point Dixon-derived T1W GRE offers several theoretical advantagesover conventional T1W FSE, namely, that it provides four image sets(in phase, opposed phase, FO, and WO; Fig. 3) and can be acquiredwith considerably shorter imaging time. Multiparametric (MP) magnet-ic resonance imaging (MRI) (which combines 2 functional imaging se-quences with anatomic imaging) is now the reference standard for408 K. Samji et al. / Clinical Imaging 40 (2016) 407413Fig. 1. A 64-year-old male with Gleason 4+5=9 PCa. Axial T2W FSE demonstrates a rounded low T2 SI nodule in the left peripheral zone (white arrow). Axial T1W FSE image obtained atthe same level demonstrates diffuse T1W hyperintense postbiopsy hemorrhage (white arrow) in the peripheral zone and an area devoid of hemorrhage (arrow head) corresponding inlocation to the nodule detected in (a). Axial apparent diffusion coefcient map image at the same level shows marked restricted diffusion within the nodule (white arrow).evaluation of the prostate 12,13 and, due to limitations in MR spec-troscopy 14, many centers are increasingly utilizing diffusion-weighted imaging and dynamic gadolinium-enhanced MR as their func-tional sequences of choice. Another potential advantage of 2-pointDixon T1W GRE compared to conventional T1W FSE is that it can alsoprovide high-quality large FOV postgadolinium-enhanced images ofthe pelvis.A traditional limitation of GRE imaging in the pelvis is the necessarytradeoff in spatial resolution required to provide sufcient coverage ofthe whole pelvis during a single breath hold. We recently implementeda high-resolution free-breathing (FB) T1W 3-dimensional (3D) 2-pointDixon GRE sequence (FLEX) with volume acceleration (LAVA) in ourprostate MR protocol in an effort to potentially reduce examinationtimes by eliminating conventional T1W FSE imaging. The purpose ofthis study was to retrospectively compare image quality and diagnosticaccuracy for detection of skeletal/nodal metastases using a high-resolution FB T1W LAVA FLEX sequence and conventional T1W FSE.2. Materials and methods2.1. PatientsUnder a quality assurance waiver from our local institutional reviewboard, a retrospective study was performed among consecutive patientswith biopsy-proven PCa who underwent MRI for pretreatment stagingof a known intermediate-risk or high-risk (Gleason score 7) PCa. TheT1W LAVA FLEX sequence was included in our MP MRI protocol inMay 2014 in an effort to eventually reduce examination times withoutcompromising diagnostic accuracy. Between May 2014 and January2015, 67 patients were imaged with the updated MP MRI protocolthat included T1W LAVA FLEX and T1W FSE. Twenty-eight patientswere excluded because the examinations were not performed for localstaging 17 examinations performed for the detection of a targetable le-sion following a negative nontargeted transrectal ultrasound (TRUS)-guided biopsy with persisting clinical suspicion for PCa, 6 examinationsperformed before or during active surveillance, 4 examinations per-formed for posttreatment follow-up, and 1 examination performed fortreatment planning. Finally, the study population consisted of 39 pa-tients that underwent MP MRI for preoperative or preradiotherapy stag-ing of a known intermediate-risk or high-risk (Gleason score 7) PCadiagnosed at nontargeted TRUS-guided biopsy.2.2. MRI techniqueDuring the study period, every patient underwent a comprehensiveMP MRI protocol performed at 3 T using integrated spine and surfacearray coils (Table 1). In addition to our routine protocol, which includeslarge FOV imaging of the pelvis with T1W FSE, patients underwent amodied FB T1W high-resolution 3D GRE sequence with 2-pointDixon fat/water separation (T1W LAVA FLEX); pulse sequence detailsare provided in Table 1. Key aspects of the modied T1W LAVA FLEX se-quence include placement of the frequency-encoded direction along theanterior/posterior direction (to mitigate breathing artifacts), utilizationof a bipolar dual echo readout (for scan efciency and added gradientFig. 2. A 59-year-old male with Gleason score 4+3=7 PCa in the left basal lateral peripheral zone. Axial T2W FSE (a) demonstrates T2W hypointense tumor (thin white arrow) withextracapsular extension and invasion of the left neurovascular bundle. The contralateral neurovascular bundle is uninvolved (thick white arrow). Axial T1W FSE (b) shows the normalright neurovascular bundle (thick white arrow) and no normal fascicles on the left and with associated contour bulge (black arrow). Axial “fat only” image obtained from FLEX LAVA(c) demonstrates similar ndings as (b); however, it was rated of higher image quality with better image sharpness compared to (b) by both readers. Note improved delineation ofthe fascicles of the right neurovascular bundle (thick white arrow).K. Samji et al. / Clinical Imaging 40 (2016) 407413 409Fig. 3. A 64-year-old male with Gleason score 3+4=7 PCa undergoing staging MRI. Axial images obtained from T1W LAVA FLEX include: in-phase (a), opposed-phase (b), WO (c), and FO(d) image sets. WO and FO images (c and d) are derived from the in-phase and the opposed-phase data using a vendor-specic postprocessing algorithm. In this study, the margin of theprostate (thick white arrow) and seminal vesicles (data not shown) and delineation of the neurovascular bundles (arrow heads) were used to assess image sharpness.balancing for ow/motion robustness), and parallel imaging (to reduceacquisition times). Given that the phase-encoded direction is set alongright/left direction, some phase wrap artifact is expected; however, sig-nicant wrap from arms/hips would be expected to fall outside the re-gion of interest.2.3. Visual analysis2.3.1. Image quality and artifactsTwo radiologists (R1 and R2) with 5 and 12 years of experience inbody MRI (blinded) assessed the T1W FSE and T1W LAVA FLEXsequences for overall image quality, image sharpness, and the pres-ence/severity of imaging artifacts. Image sharpness was dened by thedelineation of the prostate and seminal vesicles from adjacent pelvicstructures and clarity of the neurovascular bundles (Figs. 2 and 3). Asubjective assessment of overall image quality was also performed.Both image sharpness and overall image quality were rated using a5-point Likert scale (1=poor, 2=suboptimal, 3=average, 4=above av-erage, and 5=excellent) as described previously 15. Assessment forimage artifacts included motion artifact (dened as nonperiodic ghost-ing, smearing, and blurring at edges of structures 15) and phase-encoding artifact dened as repeated/periodic ghosting in the phase-Table 1Sequence parameters for MP MRI of the prostate protocol performed with pelvic surface coila at 3 Tb including FB T1W FSE and T1W LAVA FLEXImagingplaneFOV(mm)MatrixsizeSlice thickness/gap (mm)TR/TE (ms) Echo trainlengthFlipangleAccelerationfactorReceiverbandwidthAcquisitiontimeNumber ofsignals(Hz/voxel) averagedT1 FSEcT1 3D dual echo GREdT1W LAVA FLEXeT2 FSEAxialAxialAxialCoronal3503502402403203802202203203202922243303443202565.0/0.04.0/1.02.0/1.04.0/0720/8144.8//1.238905250/1051253N/AN/A27351111210111N/A22N/A244558166.71225 min 16 sBreath hold2 min 37 s4 min21112SagittalAxial3.0/03.0/04 min4 minDWIfT1 GREg dynamic contrastAxialAxial2802802202201288012812835.0/04.0/04200/904.3/1.31N/A90122219504885 min6 min4101abcdefIntegrated pelvic surface coils (16 channels) with activated spine coils (8 channels).Clinical 3-T system: Discovery 750W (General Electric, Milwaukee, WI).FSE, acquired FB.Conventional breath hold T1W in-phase and opposed-phase GRET1W 3D GRE with volume acceleration (LAVA) using 2-point Dixon fat/water separation (T1W LAVA FLEX).DWI=diffusion-weighted imaging performed with spectral fat suppression echo planar imaging with tridirectional motion probing gradients and B values of 0,500,1000 with auto-matic apparent diffusion coefcient map generation.g Dynamic fast-spoiled 2-dimensional GRE performed with a temporal resolution of 10 s after injection of 0.1 mmol/kg of gadobutrol (Gadovist, Bayer Inc., Toronto, ON) at a rate of3 ml/s.410 K. Samji et al. / Clinical Imaging 40 (2016) 407413encoding direction (set left to right for both T1W FSE and T1W LAVAFLEX sequences) 16. Additional image artifacts that were assessedspecically on the T1W LAVA FLEX sequence included “fat/waterswap” (dened as the computational misrepresentation of water as fatsignal 17) (Fig. 4) and susceptibility artifact (dened as the geographicdistortion in regions adjacent to disturbed magnetic eld homogeneityobserved with GRE imaging 18). Assessment for all image artifacts wasalso performed using a 5-point Likert scale (1=none; 2=minimal arti-fact, no impact on image quality; 3=mild artifact, slight reduction inimage quality that remains diagnostic; 4=moderate artifact, artifactthat reduces the diagnostic quality of the examination; 5=severe arti-fact, artifact renders the study nondiagnostic) as described previously. Diagnostic accuracy for metastatic diseaseTwo radiologists (R1 and R3) with 5 and 12 years of experience inbody MRI (blinded) also assessed the T1W FSE and T1W LAVA FLEX se-quences for the presence of skeletal and nodal metastases. For the T1WLAVA FLEX sequence, both radiologists considered a focal lesion in theskeleton that was of low SI (similar to muscle) detected on the FO im-ages as a potential metastasis, as has been described previously 19(Fig. 5). For metastatic lymphadenopathy, a short-axis diameter ofN1 cm was set as the threshold for potential metastatic disease 20.The presence of skeletal metastases was conrmed in all patientsusing whole-body technetium-99m methylene diphosphonate bonescan with single photon emission computed tomography (SPECT) com-puted tomography (CT). The presence of nodal metastases (dened asshort-axis diameter N1 cm as described above) was conrmed withCT of the abdomen and pelvis (which is routinely performed in conjunc-tion with MRI at our institution to complete initial staging ofintermediate-risk and high-risk PCa). The bone scan and the CT wereboth performed within 3 months of the MRI examination and werereviewed by a fellowship-trained abdominal radiologist with 7 yearsof experience to document the number of metastatic osseous lesionsand abnormal lymph nodes. R1 and R3 assessed the T1W FSE andT1W LAVA FLEX sequences in an interleaved fashion in two differentreading sessions, one session for each sequence, with each session sep-arated by a minimum of 8 weeks to reduce bias.2.4. Statistical analysisParametric data are presented as meanstandard deviation. Imagequality, sharpness, frequency, and severity of imaging artifacts weretabulated and comparisons between pulse sequences were performedfor each reader using the Wilcoxon signed-rank test. The sensitivity/specicity condence intervals (CI) for each reader for each pulse se-quence was calculated for diagnosis of lymphadenopathy and skeletalmetastases. Receiver operator characteristic (ROC) analysis was per-formed and area under curve for ROC was compared between pulse se-quences. P value b.05 was considered statistically signicant andstatistical analyses were performed with STATA Data Analysis and Sta-tistical Software v13.0 (StataCorp, College Station, TX).3. ResultsThe mean patient age was 65.36.7 years and mean PSA was 18.914.3 ng/ml. There were four patients with 22 pelvic skeletal metastasesconrmed by bone scan and ve patients with 9 pelvic nodal metastases(short-axis size N1 cm) conrmed by CT.Image quality and imaging artifacts for T1W FSE and T1W LAVA FLEXGRE are summarized in Table 2. A single patient was judged to havepoor or suboptimal image quality both on T1W FSE (by Reader 2) andT1W LAVA FLEX (by Reader 1) sequences but average image qualityby the same reader on the corresponding T1W LAVA FLEX and T1WFSE sequence. All other patients were judged to have at least averageimage quality on both sequences. Image quality scores were signicant-ly higher with T1W LAVA FLEX compared to T1W FSE for both readers,Pb.0001. Image sharpness was also signicantly better with T1W LAVAFLEX compared to T1W FSE for both readers, Pb.0001; however, imagesharpness was considered average or above average in the majority ofpatients with T1W FSE. Reader 2 identied seven patients with moder-ate motion artifact and four patients with moderate phase-encoding ar-tifact on T1W FSE. No patients had severe artifact on T1W FSE for eitherreader. Reader 2 identied three patients with moderate motion artifactand seven patients with moderate phase-encoding artifact on T1WLAVA FLEX. No patients had severe artifact on T1W LAVA FLEX for eitherreader. Motion artifact was worse with T1W FSE compared to T1WLAVA FLEX for both readers (R1, P= .03; R2, P= .002) and there wasno difference in phase-encoding artifact comparing the two pulse se-quences (R1, P=.72; R2, P=.6). With respect to fat/water swap artifactand susceptibility artifact encountered with T1W LAVA FLEX, combiningboth readers interpretations, only one patient had moderate suscepti-bility and fat/water swap artifact (identied by Reader 2 only).Diagnostic accuracies for detection of skeletal and nodal metastasesare summarized in Table 3. There was no difference in diagnostic accu-racy for detection of skeletal metastases comparing T1W FSE to T1WLAVA FLEX for either reader (PN.05). The sensitivity for detection ofskeletal metastases was 76.2% with T1W FSE and 71.4% with T1WLAVA FLEX for both reade