MR of Cervical Lymph Nodes: Comparison of Fast Spin-Echo and Conventional Spin-Echo T2W Scans

Transcription

1 Clinical Radiology (1994) 49, MR of Cervical Lymph Nodes: Comparison of Fast Spin-Echo and Conventional Spin-Echo T2W Scans D. M. YOUSEM and R. W. HURST Hospital of the University of Pennsylvania, Department of Radiology, Neuroradiology Section, Philadelphia, PA, USA Purpose: In order to explore the utility offast spin-echo techniques in the neck, 50 consecutive conventional spin-echo (CSE) Iong TR examinations were compared with 50 consecutive fast spin-echo (FSE) long TR scans for cervicalymph nodes. Materials and Methods.' Standard CSE examinations used parameters of 2200/80/1 (TR/ TE/excitations), 256 x 128 matrix, 5mm thick with interslice gaps between I and 2.5mm. FSE studies were employed with TR of ,256 x 192 matrix, fat suppression and 5 mm contiguouslices. Standard Tl-weighted examinations and clinical correlation were used as proof of nodal presence. Results: Overall vascular flow artefacts, patient motion artefacts, and image quality were comparable between the two techniques. Lymph node detectability was superior with FSE scans (P < 0.05). Typical time saved was approximately 4.5 min with FSE, despite the use of larger matrices, contiguou slices and twice the excitations. Conclaion: FSE is a competitive technique to CSE images in the neck, yielding similar quality images with higher resolution, reduced scan time, and improved lymph node detectability. Yousem, D.M. & Hurst, R.W. (1994). Clinical Radiology 49,670*675. MR of Cervical Lymph Nodes: Comparison of Fast Spin-Echo and Conventional Spin-Echo T2W Scans Acceptedfor Publication 6 June 1994 The technique of fast spin-echo (FSE) or rapid acquisition relaxation enhancement (RARE) imaging [-3]has been applied to the brain and spine with considerable success [a-8]. The main advantage of the technique is the rapid acquisition of high quality proton density andt2- weighted (T2W) images without significant degradation of signal-to-noise ratio or contrast resolution. Additionally, because of the savings in time, one has the option of increasing signal to noise by increasing the number of excitations, increasing the T2-weighting by prolonging the repetition time, or performing high resolution MR images with a larger matrix size. In the neck, where rapid image acquisition is critical due to the large scanning range that is required (from the base of the skull to the upper mediastinum), conventional spin-echo long TR sequencesuffer du'6 to their long scan times. Motion artefacts from swallowing, breathing and patient restlessness'. are compounded with long scan times. Because patients with head and neck malignancies may have difficulty breathing in the supine imaging position from airway obstruction or retained secretions, rapid imaging is even more critical. In order to cover the entire scan range required, it is often necessary to increase the slice thickness and/or the interslice gap. However, since MR must be competitive with CT in the evaluation of neck pathology, it would be preferable to obtain 5 mm-thick contiguous sections similar to those obtained with CT. With standard long TR pulse sequences, this requires interleaving of two sequences and, because of the double acquisition, scan times are doubled. One often is forced to decrease matrix Correspondence to: Dr D. M. Yousem, Department of Radiology, Neuroradiology Section, Hospital of the University of Pennsylvania, 3400 Spruce Street, Philadelphia, PA 19104, USA. sizes (leading to decreased resolution), perform l/2 excitation imaging, or face 15-20min T2W scans. These factors point to the great potential for FSE imaging in the neck. Scan times are reduced, slice thicknesses can remain competitive with CT and, with fat suppression techniques applied, contrast is high. FSE's utility in the neck remains unexplored and this has led to the current study. In order to investigate the potential of FSE imaging in the neck, we compared 50 consecutive patients with conventional spin-echo scans through the neck with 50 consecutive patients with the fast spin-echo technique. The two techniques were compared for the following parameters: diagnostic quality, patient motion artefact, carotid and jugular flow artefacts, nodal detectability and scan times. We chose to study lymph node detectability because this is a common indication for scanning and would eliminate the need to control for primary site of tumour in a comparative study. METHODS MR scans from 50 consecutive patients who were evaluated for cervical lymphadenopathy using conventional spin-echo long TR images (before fast spin-echo was introduced at our institution) and 50 consecutive FSE scans were retrospectively and independently reviewed by two experienced head and neck radiologists. All patients had squamous cell carcinomas of the head and neck and were being evaluated for metastatic cervical lymphadenopathy. Conventional spin-echo images employed a TR of 2000 to 3500 ms, TEs of ms, single excitations, and fields of view from 20 to 24cm. The imaging matrix used was

2 FSE VS CSE FOR NODES 671 Table I - Grading scale for various parameters Grade Carotidflow artefact Jugular vein artefact Diagnostic quality Patient motion Detectability of lymph nodes 0 I 2 J No artefact Minimal motion or signal within vessel Moderate motion but study interpretable Flow artefact obscures anatomy and study uninterpretable No artefact Minimal motion or signal within vessel Moderate motion but study interpretable Flow artefact obscures anatomy and study uninterpretable Diagnosis can be made Diagnosis cannot be made No artefact Minimal motion Moderate motion but study interpretable Motion obscures anatomy and study uninterpretable Excellent detectability Fair detectability Poor detectability either 256 x 128 (30 scans) or 256 x 192 (20 scans). The scan times varied with the TR, however, ranging from l2 to 14 min. Interslice gaps of 2.5 mm (14 cases),2.0 mm (1 case). 1.5mm (16 cases) and 1.0mm (19 cases) were used. Gradient moment nulling with inferior and superior presaturation pulses were applied. A variable bandwidth (4-16H2) was used to increase signal-to-noise ratio. Fifty consecutive patients in whom the fast spin-echo technique was employed were evaluated in a similar manner. Typical TRs were 3000 to 4000 with a TE of 72-9}ms.Imaging matrix was 256 x 192 with two excitations. Fivemm-thick contiguous slices were performed with a single echo technique. Superior and inferior presaturation pulses were applied. Frequency-selective fat suppression was employed. Water bags, filled with kaolin pectin and placed over the patients' necks, were used to improve fat suppression artefacts. The echo spacing was I 8-20 ms and echo train lengths of 8 or 16 were used. All scans were performed on a l.5t Signa (General Electric Medical Systems, Milwaukee, WI) scanner. The CSE and FSE long TR/long TE T2-weighted scans were evaluated initially without previewing any other pulse sequence in order to assess the ability of the sequence to visualize cervical lymphadenopathy. After arriving at a judgement concerning the presence or absence of cervical lymphadenopathy based on the long TR scans, the remainder of the study, consisting of axial TIW and post-gadopentetate dimeglumine TIW images with fat suppression, was reviewed. At that point a grade for the detectability of lymph nodes by the T2W scan was assigned. Thus, the remaining pulse sequences, in addition to clinical correlation with regard to palpable adenopathy, were used as the proof of diagnosis for the presence of adenopathy. The detectability of adenopathy was graded '0' for excellent detectability, '1' for fair detectability, arrd' 2' for poor detectability. The scans were also graded for carotid and jugular flow artefacts (which may obscure detection of lymphadenopathy), patient motion, image quality and scan parameters (Table l). Scans were graded on a 0 to 3 scale for evaluating motion artefacts. Zero represented no motion, I minimal motion which did not obscure visualization of the anatomy and 2 moderate motion without such obscuration of anatomy that the study was uninterpretable, and 3, motion obscuring anatomy to the exgnt that the study was uninterpretable. Flow motion artefact was evaluated on a 0 to 3 scale with 0 being no flow voids seen in the jugular and carotid blood vessels, I being minimal flow seen in portions of the jugular vein, 2 being significant flow within the jugular vein on most slices but without obscuration of evaluation of the lymph nodes or adjacent pathology to the point of uninterpretabitity of the scan, and 3 being sufficient signal intensity artefact in the jugular veins so as to obscure adjacent anatomy to the extent that the study was uninterpretable. Similar scales were used for the carotid artery. Image quality was graded as diagnostic or nondiagnostic based on the ability to detect the pathology without observing the other pulse sequences. Scan times were directly compared. All scans were expected to be obtained from the cavernous sinus to the upper mediastinum. If one or the other site was not included it was labelled suboptimal. Statistical analysis of CSE vs FSE data was performed based on the Student f-test of independent pairs of data with two tails and 98 degrees of freedom. I statistics are provided in Table 2. For unmatched pairs, using a P value of 0.05, a power factor of 0.80, and predicting a 20o/o difference in the sensitivities of the two techniques, this study would require 47 cases for statistical significance. We used 50 unmatched pairs. In l7 cases, both conventional spin-echo and fast spinecho scans were performed on the same patient. Two reviewers independently analysed these images for the Table 2 - Scores of all 50 cases Carotidflow artefact Jugular ttein artefact Diagnostic quality Detectability of lymph nodes Patient motion CSE FSE t-statistic * t.57 * Statistically significant P < Using gradescale shown in Table 1, the scores from the two reviewers were averaged for the 50 FSE and CSE scans to obtain values in the table above. Note the carotid flow artefact, jugular flow artefact, patient motion artefact and diagnostic quality were not significantly different between CSE and FSE. Detectability of lymph nodes was statistically superior on FSE than CSE.

3 672 CLINICAL RADIOLOGY \0) lig. t Grading ofarrefacts. (a) Non-diagnostic (grade l) CSE study demonstrating patient motion artefact (grade 2) is well as vasculai artefact fromjugular veins causing obscuration of anatomy to the ooint where the study was uninterpretable (grade 3). (D) Axial-3000/g0 fast spin-echo fat-suppressed scan in a patient with tymphoma demonstrat_ ing excellent detectability of lynph nodes and noiignificant vascular artefact. Retropharyngeal, high jugular and posteriol triangular lym_ phadenopathy is evident. same parameters as the 50 patient data set (jugular and carotid flow artefacts, patient motion, detectibility of lyrnphadenopathy and diagnostic quality). For a statisti_ cally significant result assuming the same parameters above, but for 17 matched pairs of data, ihe power factor would be 0.65, moderate in signiflcance. RESULTS (see Table 2) Fifty Patient Data Sets The mean scan time for the fast spin-echo T2W scans in the 50 patients was 6.3min. This compared with fig. 2, -. lomnarison of quality of CSE and FSE images in same patient. (a) Axial conventional spin-echo image lhrough thi vocal cords demonstrating relatively low signal-to-noise ratio due to one excitation scan with decreased resolution with256 x 12g matrix. (b) Comparison is made with fast spin-echo imaee with iai suppression in which signal-to-noise ratio a;d resolution is improved due to 256 x l92matrix and two excitation scanning. Vascular artefact is negligible. These scans were obtained through the iame patient as (a). mean scan time of ll.0min for the conventional soinecho technique. Therefore the mean time savins ;was 4.7 min. When the studies were evaluated for the presence or

4 FSE VS CSE FOR NODES 673 (b) Fig. 3 - Comparison of CSE and FSE for nodal detection. (a) Axial CSE scan at the mandibular tip demonstrating adequate visualization of the lymph nodes around the left submandibular gland (arrow). The signal-to-noise ratio (SNR) and resolution are fair. (b) The lymph node to background contrast is excellent on this axial 4000/80 fat-suppressed FSE scan through the same level on the same patient. SNR is improved as well. absence of patient motion, the average grading for the fast spin-echo technique was 0.73; however, it was 0.52 for the conventional spin-echo technique,, not statistically signihcant. Non-diagnostic studies due to patient motion occurred with equal frequency on FSE (2) and CSE (2) scans. ft) Fig. 4 - Improved nodal detection with FSE vs CSE. (a) Axial fatsuppressed FSE scan (4000/80/2) showing bilateral lymph nodes posterior to the submandibular gland (arrows) in high jugular chains, graded 'excellent visualization'. Mild vascular artefact is seen on the right side emanating from thejugular vein. Scan time for 44 contiguous 5mm sections was 7min. (6) Axial CSE (2000/90/l) scan also suffers from vascular flow artefact. Lymph node detectability was graded 'lair'. Scan time for 30 5mm sections with l mm interslice gaps was l0min, z+ s.

5 674 CLINICAL RADIOLOGY Twenty-two of 50 CSE scans failed to cover the entire region of interest of a standard head and neck MR, from the cavernous sinus to the thoracic inlet-upper mediastinum. Four FSE scans were suboptimal m tns regard. Images were graded as being diagnostic in 44 of 50 (88%) fast spin-echo and 44 of 50 (88%) conventional spin-echo MR scans. Flow artefacts (6), poor fat suppression (2), and patient motion (4) accounted for the non-diagnostic studies. The. average grade for the fast spin-echo technique for the flow artefact within the carotid arteries was 0.3b. For the conventional spin-echo the average grade was In six cases the conventional spin-echo technique had sufficient signal within the artery so as to be iabelled moderate in severity; this was present in only two FSE studies. There were no cases where carotid flow artefact obscured anatomy. For jugular flow artefacts, the mean FSE grade was 1.02 and for CSE was There was no statistically significant difference between CSE and FSE in carotid or jugular flow artefacts. However, in seven FSE cases the jugular vein had sufficient artefact from flow signal in the phase encoding plane that it obscured anatomy; this was present in only one CSE case (Figs I and 2). The FSE technique had a mean grade of 0.82 in the readers'ability to detect lymph nodes. The corresponding value of CSE was The improved ability to detecl lymph nodes with FSE was statistically significant to a p value (0.05. Poor nodal detectability was noted in 18 CSE and eight FSE studies. Comparative Studies Between CSE and FSE in Same Patient When the CSE studies were compared directly with FSE studies of the same patients, the findings were essentially similar to the analysis of the 50 patient categories. The major findings noted were: (1) no statistically significant differences in jugular vein and carotid artefact scores; (2) two non-diagnostic FSE cases because of patient motion artefact, but all CSE cases were diagnostic; (3) improved nodal detectability with FSE over CSE in eight cases, equivalent conspicuity in six cases, and inferior detectability by FSE in three cases (Figs 3 and 4). DISCUSSION The fast spin-echo technique (previously known as rapid acquisition relaxation enhancement) offers the ability to reduce scan times to as much as 1/l6th that of conventional spin-echo technique [l-4]. Typically, however, because scans are performed with multiple echoes the time saving is less than the full potential. The technique has generated a lot ofenthusiasm because of the potential for increasing patient throughput while maintaining good quality imaging. However, some early problems with the technique as implemented in the CNS (image blurring on the first echo sequence, difficulty in evaluating the proton density images because of the higher intensity of the CSF, vascular flow artefacts, and the potential for misreading areas of cortical abnormality because of their isointensity with the CSF) has led to some caution in wholeheartedly endorsine the technique [9,10]. Another of the curious characteristics of this technique that has been noticed is the high signal intensity of thetat on the long TE sequence [8,11]. The closely spaced 1g0" refocusing pulses cause a decrease in the contribution of fat to the T2 decay process leading to bright lipid signal [8]. For this reason, fast spin-echo images through the orbit and head and neck are generally performed with fat suppression. We chose to explore the fat-suppressed FSE technique in the neck in order to determine its utility visd-vis conventional spin-echo T2-weighted scans, since the long CSE scan times often lead to sub-optimal studies. The use of fat suppression raises the spe-tre of incomplete or inhomogeneous suppression which may degrade scan quality. This was a problem in 4oh of our cases and can be obviated with the use of water bass filled with kaolin pectin or other 'sat-pads' currently o=n the market. The proton density-weighted image, most helpful in evaluating cortical lesions in the brain, is of limited use in the MR evaluation of neck masses. Because of this, fast spin-echo images with a single long (T2-weighted) echo may be performed, thereby gaining a greater advantage with the fast spin-echo technique. With fat suppression, lesions which are inflammatory and have a high signal intensity on T2-weighted scans may be easily dete-ted against dark fat, and lesions which are of intermediate signal intensity (most highly cellular malignancies such as squamous cell carcinoma) will still be of greater intensity than the suppressed fat. The lymph nodes in the neck, which are sometimes isointense with muscle and hard to see on short TR images, are well demonstrated as high signal intensity structures on T2-weighted scans. Detection of lymphadenopathy is one of the primary roles for T2W scans in the neck and that is why we chose to analyse this variable in our FSE vs CSE study. Standard spin-echo images require flow compensation techniques such as gradient moment nulling in order to reduce the flow aftefact from blood vessel pulsations. Employing gradient moment nulling (GMI.Q decreases the number of scans per TR because of the larger TE needed to apply the gradients to compensate for the flow [12]. This also decrease signal-to-noise ratio. However, GMN also yields high signal in slow flowing blood vessels which can lead to misdiagnosis of lymphadenopathy where jugular venous branches are outlined in fat. However, analysis of sequential images showing the tubular nature of the vessel should reduce this possible error. Our study has shown that the fast spin-echo technique does not require flow compensation to reduce flow artefacts to the equivalent status of conventional spin-echo images. Without GMN, the blood vessels remain black because of the saturation occurring with the rapid 180" pulse profiles. The flow artefacts are tolerable. Ideally, this study should be followed with a prospective protocol performing CSE and FSE studies on all patients, who are matched for primary sites of tumour and staging, to assess the role of FSE for routine neck imaging. This would require a very large population of head and neck lesions to evaluate; for this reason we chose cervical lymphadenopathy alone. Because of the difference in image quality, blinding reviewers to the two types of scans is not possible. Also using another 'gold' standard for proof of adenopathy, such as pathologic confirmation rather than the remaining pulse sequences, would also be difficult to achieve retrospectively.

6 FSE VS CSE FOR NODES 675 In this study we have demonstrated that the artefacts from patient motion, and jugular and carotid flow are comparable and not statistically significant between FSE and CSE images. Lymph node detectability shows a statistically significant (P < 0.05) improvement with FSE over CSE. In addition, with FSE, one has the advantage of 256 x 192 images (vs 256 x 128 with CSE), two excitation scanning (vs one excitation with CSE) for improved signal-to-noise ratio, and contiguous 5mm slices (vs mm interslice gaps with CSE) while obtaining scan times that are 4.5min shorter. Instead of skimping on full coverage of the neck (seen in 23 of 50 CSE studies presumably because of time constraints), full coverage from the base of the skull to the upper mediastinum can be obtained with FSE scans in 6.3min. We believe the technique has tremendous potential in neck imaging and will replace CSE scans. If high resolution (256 x 256) images are required the technique can be adjusted with a continued benefit in time savings compared with conventional spin-echo technique. Acknowledgements. The authors acknowledge the assistance of Paul Wang and Wayne Alves in the development and analysis of this project. REFERENCF,s I Hennig J, Nauerth A, Friedburg H. RARE imaging: a fast imaging method for clinical MR. Magnetic Resonance in Medicine 1986; 3: Hennig J, Friedburg H. Clinical applications and methodological developments of the RARE technique. Magnetic Resonance Imaging 1988;6: Mulkem RV, Wong STS, Winalski C et al. Contrzst manipulation and artifact assessment of 2D and 3D RARE sequences. Magnetic Resonance Imaging I 990;8: Oshio K, Jolesz FA, Melki PS et al. T2-weighted thin-section imaging with the multislab three-dimensional RARE technique. Journal of Magnetic Resonance Imagr.ng 1991;l: Jones KM, Mulkern RV, Mantell MT et al. Brain hemorrhage: evaluation with fast spin-echo and conventional dual spin-echo images. Radio lo g y 1992;182: Sze G. MR imaging of the spinal cord: current status and future advances. American Journal of Roentgenology 1992;159 :l 49-l Sze G, Merriam M, Oshio K et al. Fast spin echo in the evaluation ofintradural disease ofthe spine. American Journal of Neuroradiology 1992;13: Jones KM, Mulkern RV, Schwartz RB et al. Fast spin-echo MR imaging of the brain and spine: current concepts. American Joumal of Roentgenology 1992;158: Atlas SW, Hackney DB, Listerud J et al. Fast spin echo MR imaging: blinded comparison with conventional spin echo imaging in the detection offocal brain lesions. Society of Magnetic Resonance in Medictue 1991: Abstract No. 41. l0 Hackney DB, Listerud J, Yousem DM et al. Optimization of parameters for fast spin echo imaging of the brain. Society of Magnetic Resonance in Medicine 1991; Abstract No l Melki PS, Mulkern RV, Panych LP et al. Comparing the FAISE method with conventional dual-echo sequences. Journal of Magnetic Re sonance Imag ing 19 ;l : l2 Mugler JP, Brookeman JR. The design of pulse sequences emplofng spatial presaturation for the suppression of flow artifacts. Magnetic Resonance in Medicine 1992;23: