- Email: [email protected]
Magnetic resonance imaging: Spinal cord imaging with the turbo-fluid attenuated inversion recovery (FLAIR) pulse sequence
ClinicalRadiology(1995) 50, 1-5
Magnetic Resonance Imaging: Spinal Cord Imaging With the Turbo-Fluid Attenuated Inversion Recovery (FLAIR) Pulse Sequence J. V. HAJNAL, L. KASUBOSKI*, N. M. deSOUZAt and G. M. BYDDER~f
Picker International, Borehamwood, Herts, UK, *Picker International, Highland Heights, OH, USA, and t The Robert Steiner Magnetic Resonance Unit, Hammersmith Hospital, Royal Postgraduate Medical School, London, UK Twelve patients with suspected or proven spinal cord disease were studied with a fluid attenuated repeated echo sequence employing a fast spin-echo data acquisition (TurboFLAIR) and the results were compared with those obtained with conventional Tl- and T2weighted spin-echo sequences. The Turbo-FLAIR sequence utilized an unselected 180 ~ pulse to reduce the signal from CSF followed by a spin-echo sequence based on the Rapid Acquisition with Relaxation Enhancement (RARE) pulse sequence for data acquisition. Three and four echo versions were used. Four low signal intensity!esions (cysts or syrinxes) were seen of which three were better seen with the Turbo-FLAIR sequence than with the other sequences. The remaining lesion was equally well seen. Of the seven high signal lesions found, six were better seen with the TurboFLAIR sequence and one was equally well seen. The Turbo-FLAIR sequence is more rapid than the T2-weighted spin-echo sequence and compares favourably with the combined results of both the Tl- and T2-weighted spin-echo sequences for lesion conspicuity. Hajnal, J.V., Kasuboski, L., deSouza, N.M. & Bydder, G.M. (1995). Clinical Radiology 50, 1-5. Magnetic Resonance Imaging: Spinal Cord Imaging With the Turbo-Fluid Attenuated Inversion Recovery (FLAIR) Pulse Sequence
Accepted for Publication 15 August 1994
Although T2-weighted spin-echo sequences are widely accepted as the most sensitive technique for the detection of intracerebral lesions, their use in the spinal cord is more problematic. Partial volume effects and motion artefacts from CSF produce high signal intensity artefacts which may simulate disease even when gradient moment nulling and cardiac gating are used. By applying an inversion pulse 2000-2500 ms before the T2-weighted spin-echo sequence, it is possible to null or greatly reduce the signal from CSF and produce highly T2-weighted images of the spinal cord with a low level of artefact from CSF. This approach also provides high contrast between the spinal cord and CSF and is well suited to detecting lesions such as cysts which have a very long Tl. Fluid attenuated inversion recovery (FLAIR) sequences can provide increased sensitivity to disease in the brain and spinal cord [1,2]. The principal disadvantage of the FLAIR sequence is the increase in time which it requires. The repetition time (TR) is typically 5000-6000 ms compared with the 25003000ms for conventional T2-weighted sequences. By utilizing an inversion pulse with a spin-echo sequence in which several phase encoding steps are performed after each excitation it is possible to reduce the imaging time [3] and still achieve T2-weighted images with good control of CSF artefact. The resulting Turbo-FLAIR sequence offers the potential for more rapid imaging with high sensitivity to disease. Correspondenceto: ProfessorG. M. Bydder,The RobertSteinerMR Unit, HammersmithHospital,Du CaneRoad,LondonW120HS,UK.
In this paper we illustrate the use of this technique in 12 patients with suspected or proven spinal cord disease and compare the results with conventional Tl- and T2weighted images. PATIENTS AND METHODS With the permission of the Royal Postgraduate Medical School Research Ethics Committee 12 patients with the diagnosis of spinal cord disease were examined using a Picker 1.0T HPQ system. Their clinical diagnoses are summarized in Table 1. The Turbo-FLAIR sequence used in this study consisted of a non-selective 180~ pulse followed by a multislice multi-echo spin-echo sequence designed to produce a set of slices at different levels. Sequence times were chosen so that the central slice had an inversion time (TI) appropriate for the null point of CSF (about 2200 ms at 1.0 T). The CSF signal was not completely nulled for the slices acquired at the start and finish of the multislice set, but provided these were within about 500 ms of the null point, the CSF signal was still sufficiently reduced to obtain useful benefit. Since the signal from all the CSF within the spinal column was suppressed the scans were insensitive to artefacts from CSF flow. Subjects were examined using three and four echo sequences in which the first echo was phase encoded to provide low spatial frequencies and subsequent echoes were separately phase encoded to obtain higher spatial frequencies. The T2 weighting of the images was primarily.
2
CLINICAL RADIOLOGY
Table 1 - Clinical
diagnoses (n =
12)
Syrinx and post-traumatic cyst (3) Post-traumatic cyst (2) Syrinx Cord compression(5) Disc (2) Malignant metastasis (3) Multiple sclerosis(2) Tumours (2) Ependymoma (2)
(SE2500/80) emplopying gradient moment nulling were obtained in each case using a 128 x 256 matrix, two data acquisitions and matched 3 mm slice widths (7.5 min). Scans were studied by two observers for lesion number, extent and conspicuity and a consensus assessment was made of all conventional images compared to the T u r b o - F L A I R images.
RESULTS
Syrinx and Post-Traumatic Cyst (3) determined by the time of the first echo (40, 60 or 80 ms), generally referred to as a pseudo echo time (pTE). The inter-echo spacing of the sequences was 38ms. Up to 15 levels were obtained with slice widths of 3 mm and one data acquisition. The phase encoded direction was superior/inferior for coronal and sagittal images and 256 x 256 data matrices were acquired in 6.6 min. Comparable 3 m m unenhanced T2-weighted scans (128 x 256) (SE500-700/20) were obtained in all cases with 4 data acquisitions (3.7 min). Intravenous gadopentetate dimeglumine (0.1mmol/kg) followed by a T1weighed scan was used in five cases (one case of multiple sclerosis, two cases of cord compression and two cases of intramedullary tumour). T2-weighted spin-echo sequences
Post-Traumatic Cyst (2) Two patients were examined one 8 months and the other 8 years after trauma. In the first case the cystic lesion in the cervical cord was better seen with the TurboF L A I R sequence (Fig. 1). The lower thoracic cord cyst in the second patient was equally well seen with the T1weighted SE and T u r b o - F L A I R sequences.
Syrinx This patient presented with a progressive quadriplegia 4 years after a motor vehicle accident which had rendered him paraplegic. The lower extent o f the thoracic syrinx
(a) (b) Fig. 1 - Post-traumatic cyst: SE560/20 (a) and T u r b o - F L A I R 6050/60/2180 (b) scans. The cyst (arrow) is m o r e clearly seen in (b).
3
MRI USING TURBO-FLAIR
was seen more clearly using the Turbo-FLAIR sequence than with conventional scans (Fig. 2).
Cord Compression (5) Three of the five patients showed increased signal within or around the cord using Turbo-FLAIR sequences; no change in signal intensity was seen within the cord in the remaining two cases. In no case was increased signal within the cord apparent using the conventional SE2500/80 sequence. A patient with a single level compression due to a prolapsed disc is illustrated in Fig. 3. The increased signal intensity in his cord may have been clue to oedema or gliosis. Multiple Sclerosis (2) One patient with definite multiple sclerosis (MS) showed inhomogeneity in the spinal cord with the SE2500/80 sequence. This lesion was shown more clearly with the Turbo-FLAIR sequence. An extensive area of abnormality was seen on the Turbo-FLAIR in the second case although conventional imaging was negative (Fig. 4).
(a)
(a) (b) Fig. 2 - Syrinx following trauma: SE560/20 (a) and Turbo-FLAIR 5040/80/2270 (b) scans. The lower extent of the syrinx is best demonstrated in (b) (arrow).
Fig. 3 - Cord compression due to disc protrusion lateral slice: SE580/20 (a), SE2500/80 (b) and Turbo-FLAIR 6060/40/2160 (e) images. The cord displays an increase in signal intensity at the level of the stenosis on (c) (arrows).
4
CLINICAL RADIOLOGY
Fig. 3 (c)
Fig. 3 (b)
Tumours (2)
Ependymoma (2) Both cases had a previous history of surgery and radiotherapy. The first patient presented with back pain during pregnancy three years after surgery for an ependymoma of the conus. Several nerve roots and the lower end of the conus showed contrast enhancement with the Tl-weighted spin-echo. The same areas showed high signal intensity using the Turbo-FLAIR sequence, without the need for contrast enhancement. The second patient had a cervical and thoracic cord ependymoma incompletely removed at surgery 8 years previously. A cystic surgical defect was apparent as a low signal area on the Tl-weighted image but the SE2500/80 scan showed poor detail. The cyst and some high signal regions were better shown with the Turbo-FLAIR sequence. Overall, in all cases, the four low signal intensity lesions due to a long Tl (either cysts or syrinxes) were better seen with the Turbo-FLAIR sequence in three cases and equally well seen with the best of the other sequences in one case. Of the seven high signal areas due to increased T2, six were most clearly seen with the Turbo-FLAIR sequence and one was equally well seen with either the Tl-weighted spin-echo (with or without enhancement) or the T2weighted spin-echo sequence.
DISCUSSION In this small series, the Turbo-FLAIR sequence provided better or equal conspicuity in both long T1 and long T 2 lesions and was performed in less time than conventional T2-weighted spin-echo sequences with acceptable levels of reduction in signal-to-noise ratio and edge definition. The basic FLAIR sequence is easy to implement but relatively slow (12.8 min for an 11 slice set), The use of multiecho techniques to acquire more than one phase encode step per excitation allowed the imaging time to be reduced and the advantages over conventional techniques to be maintained, but it also introduces some disadvantages. There is a loss in signal-to-noise ratio of the images as compared to the conventional single echo sequences. This is in part attributable to residual uncompensated eddy currents, which can be substantially reduced by the use of self-shielded gradient coils. Another disadvantage of the sequence is the mixed T2 weighting it produces with an associated loss of edge information [4,5]. This is not a particular problem in spinal cord imaging when the phase encoding direction is from head to foot, except in the case of very small lesions. Finally, and unique to this particular sequence, use of multiple echoes leads to loss of efficiency with which successive slices can be acquired within the range o f inversion times that produce substantial attenuation of
MRI USING TURBO-FLAIR
(a) Fig. 4
5
(17)
Multiple sclerosis: SE2500 80 (a) and Turbo-EkAIR 6040 60 2130 (h) scar:> There is an extensixc high signal region in (b) (arrows).
the CSF signal. This reduces the number of useful slices that can be acquired for a given pTE, but is not generally a problem for sagittal or coronal imaging of the cord because relatively few slices are required for the examination.
2
3 Acknowledgements. We are grateful to the Medical Research Council and the Department of Health for their continued support.
4
REFERENCES
5
1 De Coene B, Hajnal JV, Gatehouse P et al. MRI of the brain using
fluid attenuated inversion recovery (FLAIR) pulse sequences. A mcri~an Journal ~)/?, euroradiology 1992:13:1555 1564. White SJ. Hajnal JV. Young IR el al. Use of fluid attenuated inxersion recovery (FLAIR) pulse sequences for imaging the spinal cord and nerve roots. Magnetic Resonance in Medicine 1992; 28:153 162. Hennig J. Naureth A, Friedburgh H. RARE (Rapid Acquisition with Relaxation Enhancement) imaging: a fast imaging method for clinical MR. Magnetic Resonance in Medicine 1986;3:823 832. Constable RT, Anderson AW, Zhong J e t al. Factors influencing contrast in fast spin echo M R imaging. Magnetic Resonance Imaging 1992;10:497 511. Jones KM, Mulkern RV, Schwartz KB et al. Fast spin echo M R imaging of the brain and spine: current concepts. American Journal o f Roentgenology 1992;158:1313-1320.
Magnetic Resonance Imaging: Spinal Cord Imaging With the Turbo-Fluid Attenuated Inversion Recovery (FLAIR) Pulse Sequence J. V. HAJNAL, L. KASUBOSKI*, N. M. deSOUZAt and G. M. BYDDER~f
Picker International, Borehamwood, Herts, UK, *Picker International, Highland Heights, OH, USA, and t The Robert Steiner Magnetic Resonance Unit, Hammersmith Hospital, Royal Postgraduate Medical School, London, UK Twelve patients with suspected or proven spinal cord disease were studied with a fluid attenuated repeated echo sequence employing a fast spin-echo data acquisition (TurboFLAIR) and the results were compared with those obtained with conventional Tl- and T2weighted spin-echo sequences. The Turbo-FLAIR sequence utilized an unselected 180 ~ pulse to reduce the signal from CSF followed by a spin-echo sequence based on the Rapid Acquisition with Relaxation Enhancement (RARE) pulse sequence for data acquisition. Three and four echo versions were used. Four low signal intensity!esions (cysts or syrinxes) were seen of which three were better seen with the Turbo-FLAIR sequence than with the other sequences. The remaining lesion was equally well seen. Of the seven high signal lesions found, six were better seen with the TurboFLAIR sequence and one was equally well seen. The Turbo-FLAIR sequence is more rapid than the T2-weighted spin-echo sequence and compares favourably with the combined results of both the Tl- and T2-weighted spin-echo sequences for lesion conspicuity. Hajnal, J.V., Kasuboski, L., deSouza, N.M. & Bydder, G.M. (1995). Clinical Radiology 50, 1-5. Magnetic Resonance Imaging: Spinal Cord Imaging With the Turbo-Fluid Attenuated Inversion Recovery (FLAIR) Pulse Sequence
Accepted for Publication 15 August 1994
Although T2-weighted spin-echo sequences are widely accepted as the most sensitive technique for the detection of intracerebral lesions, their use in the spinal cord is more problematic. Partial volume effects and motion artefacts from CSF produce high signal intensity artefacts which may simulate disease even when gradient moment nulling and cardiac gating are used. By applying an inversion pulse 2000-2500 ms before the T2-weighted spin-echo sequence, it is possible to null or greatly reduce the signal from CSF and produce highly T2-weighted images of the spinal cord with a low level of artefact from CSF. This approach also provides high contrast between the spinal cord and CSF and is well suited to detecting lesions such as cysts which have a very long Tl. Fluid attenuated inversion recovery (FLAIR) sequences can provide increased sensitivity to disease in the brain and spinal cord [1,2]. The principal disadvantage of the FLAIR sequence is the increase in time which it requires. The repetition time (TR) is typically 5000-6000 ms compared with the 25003000ms for conventional T2-weighted sequences. By utilizing an inversion pulse with a spin-echo sequence in which several phase encoding steps are performed after each excitation it is possible to reduce the imaging time [3] and still achieve T2-weighted images with good control of CSF artefact. The resulting Turbo-FLAIR sequence offers the potential for more rapid imaging with high sensitivity to disease. Correspondenceto: ProfessorG. M. Bydder,The RobertSteinerMR Unit, HammersmithHospital,Du CaneRoad,LondonW120HS,UK.
In this paper we illustrate the use of this technique in 12 patients with suspected or proven spinal cord disease and compare the results with conventional Tl- and T2weighted images. PATIENTS AND METHODS With the permission of the Royal Postgraduate Medical School Research Ethics Committee 12 patients with the diagnosis of spinal cord disease were examined using a Picker 1.0T HPQ system. Their clinical diagnoses are summarized in Table 1. The Turbo-FLAIR sequence used in this study consisted of a non-selective 180~ pulse followed by a multislice multi-echo spin-echo sequence designed to produce a set of slices at different levels. Sequence times were chosen so that the central slice had an inversion time (TI) appropriate for the null point of CSF (about 2200 ms at 1.0 T). The CSF signal was not completely nulled for the slices acquired at the start and finish of the multislice set, but provided these were within about 500 ms of the null point, the CSF signal was still sufficiently reduced to obtain useful benefit. Since the signal from all the CSF within the spinal column was suppressed the scans were insensitive to artefacts from CSF flow. Subjects were examined using three and four echo sequences in which the first echo was phase encoded to provide low spatial frequencies and subsequent echoes were separately phase encoded to obtain higher spatial frequencies. The T2 weighting of the images was primarily.
2
CLINICAL RADIOLOGY
Table 1 - Clinical
diagnoses (n =
12)
Syrinx and post-traumatic cyst (3) Post-traumatic cyst (2) Syrinx Cord compression(5) Disc (2) Malignant metastasis (3) Multiple sclerosis(2) Tumours (2) Ependymoma (2)
(SE2500/80) emplopying gradient moment nulling were obtained in each case using a 128 x 256 matrix, two data acquisitions and matched 3 mm slice widths (7.5 min). Scans were studied by two observers for lesion number, extent and conspicuity and a consensus assessment was made of all conventional images compared to the T u r b o - F L A I R images.
RESULTS
Syrinx and Post-Traumatic Cyst (3) determined by the time of the first echo (40, 60 or 80 ms), generally referred to as a pseudo echo time (pTE). The inter-echo spacing of the sequences was 38ms. Up to 15 levels were obtained with slice widths of 3 mm and one data acquisition. The phase encoded direction was superior/inferior for coronal and sagittal images and 256 x 256 data matrices were acquired in 6.6 min. Comparable 3 m m unenhanced T2-weighted scans (128 x 256) (SE500-700/20) were obtained in all cases with 4 data acquisitions (3.7 min). Intravenous gadopentetate dimeglumine (0.1mmol/kg) followed by a T1weighed scan was used in five cases (one case of multiple sclerosis, two cases of cord compression and two cases of intramedullary tumour). T2-weighted spin-echo sequences
Post-Traumatic Cyst (2) Two patients were examined one 8 months and the other 8 years after trauma. In the first case the cystic lesion in the cervical cord was better seen with the TurboF L A I R sequence (Fig. 1). The lower thoracic cord cyst in the second patient was equally well seen with the T1weighted SE and T u r b o - F L A I R sequences.
Syrinx This patient presented with a progressive quadriplegia 4 years after a motor vehicle accident which had rendered him paraplegic. The lower extent o f the thoracic syrinx
(a) (b) Fig. 1 - Post-traumatic cyst: SE560/20 (a) and T u r b o - F L A I R 6050/60/2180 (b) scans. The cyst (arrow) is m o r e clearly seen in (b).
3
MRI USING TURBO-FLAIR
was seen more clearly using the Turbo-FLAIR sequence than with conventional scans (Fig. 2).
Cord Compression (5) Three of the five patients showed increased signal within or around the cord using Turbo-FLAIR sequences; no change in signal intensity was seen within the cord in the remaining two cases. In no case was increased signal within the cord apparent using the conventional SE2500/80 sequence. A patient with a single level compression due to a prolapsed disc is illustrated in Fig. 3. The increased signal intensity in his cord may have been clue to oedema or gliosis. Multiple Sclerosis (2) One patient with definite multiple sclerosis (MS) showed inhomogeneity in the spinal cord with the SE2500/80 sequence. This lesion was shown more clearly with the Turbo-FLAIR sequence. An extensive area of abnormality was seen on the Turbo-FLAIR in the second case although conventional imaging was negative (Fig. 4).
(a)
(a) (b) Fig. 2 - Syrinx following trauma: SE560/20 (a) and Turbo-FLAIR 5040/80/2270 (b) scans. The lower extent of the syrinx is best demonstrated in (b) (arrow).
Fig. 3 - Cord compression due to disc protrusion lateral slice: SE580/20 (a), SE2500/80 (b) and Turbo-FLAIR 6060/40/2160 (e) images. The cord displays an increase in signal intensity at the level of the stenosis on (c) (arrows).
4
CLINICAL RADIOLOGY
Fig. 3 (c)
Fig. 3 (b)
Tumours (2)
Ependymoma (2) Both cases had a previous history of surgery and radiotherapy. The first patient presented with back pain during pregnancy three years after surgery for an ependymoma of the conus. Several nerve roots and the lower end of the conus showed contrast enhancement with the Tl-weighted spin-echo. The same areas showed high signal intensity using the Turbo-FLAIR sequence, without the need for contrast enhancement. The second patient had a cervical and thoracic cord ependymoma incompletely removed at surgery 8 years previously. A cystic surgical defect was apparent as a low signal area on the Tl-weighted image but the SE2500/80 scan showed poor detail. The cyst and some high signal regions were better shown with the Turbo-FLAIR sequence. Overall, in all cases, the four low signal intensity lesions due to a long Tl (either cysts or syrinxes) were better seen with the Turbo-FLAIR sequence in three cases and equally well seen with the best of the other sequences in one case. Of the seven high signal areas due to increased T2, six were most clearly seen with the Turbo-FLAIR sequence and one was equally well seen with either the Tl-weighted spin-echo (with or without enhancement) or the T2weighted spin-echo sequence.
DISCUSSION In this small series, the Turbo-FLAIR sequence provided better or equal conspicuity in both long T1 and long T 2 lesions and was performed in less time than conventional T2-weighted spin-echo sequences with acceptable levels of reduction in signal-to-noise ratio and edge definition. The basic FLAIR sequence is easy to implement but relatively slow (12.8 min for an 11 slice set), The use of multiecho techniques to acquire more than one phase encode step per excitation allowed the imaging time to be reduced and the advantages over conventional techniques to be maintained, but it also introduces some disadvantages. There is a loss in signal-to-noise ratio of the images as compared to the conventional single echo sequences. This is in part attributable to residual uncompensated eddy currents, which can be substantially reduced by the use of self-shielded gradient coils. Another disadvantage of the sequence is the mixed T2 weighting it produces with an associated loss of edge information [4,5]. This is not a particular problem in spinal cord imaging when the phase encoding direction is from head to foot, except in the case of very small lesions. Finally, and unique to this particular sequence, use of multiple echoes leads to loss of efficiency with which successive slices can be acquired within the range o f inversion times that produce substantial attenuation of
MRI USING TURBO-FLAIR
(a) Fig. 4
5
(17)
Multiple sclerosis: SE2500 80 (a) and Turbo-EkAIR 6040 60 2130 (h) scar:> There is an extensixc high signal region in (b) (arrows).
the CSF signal. This reduces the number of useful slices that can be acquired for a given pTE, but is not generally a problem for sagittal or coronal imaging of the cord because relatively few slices are required for the examination.
2
3 Acknowledgements. We are grateful to the Medical Research Council and the Department of Health for their continued support.
4
REFERENCES
5
1 De Coene B, Hajnal JV, Gatehouse P et al. MRI of the brain using
fluid attenuated inversion recovery (FLAIR) pulse sequences. A mcri~an Journal ~)/?, euroradiology 1992:13:1555 1564. White SJ. Hajnal JV. Young IR el al. Use of fluid attenuated inxersion recovery (FLAIR) pulse sequences for imaging the spinal cord and nerve roots. Magnetic Resonance in Medicine 1992; 28:153 162. Hennig J. Naureth A, Friedburgh H. RARE (Rapid Acquisition with Relaxation Enhancement) imaging: a fast imaging method for clinical MR. Magnetic Resonance in Medicine 1986;3:823 832. Constable RT, Anderson AW, Zhong J e t al. Factors influencing contrast in fast spin echo M R imaging. Magnetic Resonance Imaging 1992;10:497 511. Jones KM, Mulkern RV, Schwartz KB et al. Fast spin echo M R imaging of the brain and spine: current concepts. American Journal o f Roentgenology 1992;158:1313-1320.