ORIGINAL ARTICLE
ACTA RADIOLOGICA
Magnetic Resonance Imaging of the Cranial Nerves in the Posterior Fossa: A Comparative Study of T2-Weighted Spin-Echo Sequences at 1.5 and 3.0 Tesla F. FISCHBACH, M. MU¨LLER & H. BRUHN Department of Radiology, Otto von Guericke University, Medical School, Magdeburg, Germany; Department of Radiology, Charite´, Campus Virchow-Klinikum, Humboldt-University Medical School, Berlin, Germany; Department of Traumatic and Reconstructive Surgery, Charite´, Campus Mitte, Humboldt-University Medical School, Berlin, Germany Fischbach F, Mu¨ller M, Bruhn H. Magnetic resonance imaging of the cranial nerves in the posterior fossa: a comparative study of T2-weighted spin-echo sequences at 1.5 and 3.0 Tesla. Acta Radiol 2008;49:358–363. Background: High-field magnetic resonance imaging (MRI) at 3.0 Tesla (T) is rapidly gaining clinical acceptance. Whether doubling of the field strength of 1.5T and the subsequent increase in signal-to-noise ratio (SNR) leads to a significant improvement of image quality is not automatically given. Purpose: To evaluate the depiction of fine anatomic detail in the posterior fossa, focusing on brain nerves, on T2-weighted imaging, and to define the potential advantage of imaging at 3.0T versus 1.5T. Material and Methods: In total, 10 brainstem nerve pairs of 12 volunteers were identified on T2-weighted MR images of 2- and 5-mm section thickness acquired at 1.5T and 3.0T. The MR images were compared for each subject at both field strengths by three independent readers who rated image quality according to depiction of anatomic detail and contrast by using a rating scale. Results: In general, MR images at 3.0T were considered more conspicuous and less noisy than images at 1.5T. The SNR value measured was almost doubled. With respect to structural identification and contrast according to the rating scale, observer scores were significantly improved both for standard imaging with 5-mm sections and highresolution imaging with 2-mm sections at 3.0T. Direct comparison revealed a significant increase for evaluated image quality criteria and the number of nerves detected. Conclusion: The comparison revealed a clear advantage in favor of T2-weighted MRI at 3.0T vs. 1.5T in depicting the roots and course of brain nerves in the posterior fossa. Key words: Anatomy; brain/brainstem; CNS; MR imaging Frank Fischbach, Department of Radiology, Otto von Guericke University, Medical School, Magdeburg, Leipziger Strasse 44, 39120 Magdeburg, Germany (tel. +49 391 6713030, fax. +49 391 6713029, e-mail.
[email protected]) Accepted for publication November 16, 2007
High-field magnetic resonance imaging (MRI) at 3.0 Tesla (T) is rapidly gaining clinical acceptance. Among others, increased use of 3.0T imaging is based on the expectation of higher signal-to-noise ratio (SNR) traded off with higher resolution and/or shortening of examination time compared to lower field strength (1). While a higher resolution will depict smaller anatomical structures and pathologies more accurately and conspicuously, ultimately, this may also improve radiologic diagnostics. Moreover, a shortened scan time may increase DOI 10.1080/02841850701824127
patient comfort and help to decrease image artifacts caused by patient motion. Whether doubling of the field strength of 1.5T and the subsequent increase in SNR leads to a significant improvement of image quality is not automatically given. It needs to be shown for a particular application in a systematic study, because a number of other factors may also assume a significant role (2, 3). First, a higher field strength of 3.0T does not only allow for a higher SNR but also introduces complications such as a higher # 2008 Informa UK Ltd. (Informa Healthcare, Taylor & Francis AS)
MRI at 1.5 and 3.0T of the Cranial Nerves in the Posterior Fossa
susceptibility to artifacts (4) and altered relaxation times that impact on the contrast. Second, there is no direct relation between the field strength, the SNR, and the visual appreciation of the anatomical object in the image (5). Optimal selection and adjustment of additional hardware such as transmitter and receiver coils (e.g., avoiding crosstalk) may not necessarily be given. Finally, B1 inhomogeneity effects, caused for example by shorter wavelengths, may lead to signal strength variations in the image at 3.0T. Structures in the posterior fossa appear to be appropriate for evaluating the influence of higher field strength on the quality of neuroimaging, as delicate structures lie in close vicinity. The detection of abnormalities may be of crucial importance for treatment, as shown in a previous case report (6). Therefore, the aim of this study was to underpin experimentally the increased value of high-resolution T2-weighted neuroimaging at 3.0T vs. 1.5T for the depiction of anatomical details. Material and Methods Subjects Twelve healthy volunteers (seven males, five females; mean age 25¡2 years [SD]) recruited among students underwent MR imaging both at 1.5T and 3.0T after written informed consent was obtained. Ten brain nerve pairs ranging from the left and right oculomotorius nerve (N III) down to the hypoglossal nerve (N XII) were evaluated.
359
was employed with 5-mm sections, and a fastrecovered fast spin-echo (FRFSE) pulse sequence with 2-mm section thickness, respectively. Sequence parameters are summarized in Table 1. At 3.0T, the improved SNR could be used for an increased in-plane resolution of 0.4460.55 mm with 2-mm sections (field of view [FOV] 150 mm) and 0.4360.76 mm with 5-mm sections, which was not possible at 1.5T due to unacceptable image noise when keeping the number of acquisitions and scan times comparable at 3 min for the 5-mm scans and 5 min for the 2-mm scans. Analyses of SNRs in phantom measurements To provide a true comparison of the SNR at both field strengths, SNR measurements were performed on image sections of a sphere filled with 1 l of phosphate-buffered gadolinium-doted (Magnevist, 0.1%; Bayer-Schering, Berlin, Germany) water solution (MRS HD sphere; GE Healthcare, Milwaukee, Wisc., USA) using the same pulse sequences and parameters as in the scanning of volunteers. The SNR was calculated according to the equation
MR imaging Each volunteer underwent MR imaging examinations both at 1.5T (Twin-Speed) and 3.0T (Signa; both GE Healthcare, Milwaukee, Wisc., USA) using the vendor’s quadrature head coil. For comparison, a T2-weighted fast spin-echo (FSE) pulse sequence
ðSIROI SInoise Þ=SDnoise where SIROI denotes the signal intensity in a region of interest (ROI) placed at the center of the sphere, SInoise denotes the signal of the noise, and SDnoise the standard deviation of the background noise measured in the phaseencoding direction. Each SI measurement was repeated three times. As shown in Table 2, the measured SNR values were then normalized to the same voxel size in all modalities of 2-mm and 5-mm sections with the 2-mm and 5-mm sections at 1.5T each set at a base value of 100%.
Table 1. Pulse sequence protocols and imaging parameters used 1.5 Tesla
3.0 Tesla
Parameter
FSE, 5 mm
FRFSE, 2 mm
FSE, 5 mm
FRFSE, 2 mm
TE/TR, ms FA, ˚ ETL RBW, kHz FOV, mm NEX Matrix Resolution, mm Acquisition time, min:s
98/3900 90 16 16 2206220 1 2566256 0.8660.86 3:52
110/6000 90 64 16 2006200 4 3206256 0.6360.78 11:39
88/3900 90 16 16 2206220 1 5126288 0.4360.76 4:21
98/6000 90 64 16 1506150 4 3206256 0.4460.55 9:13
TE/TR: echo time/repetition time; FA: flip angle; ETL: echo train length; BW: receiver bandwidth; FOV: field of view; NEX: number of excitations. The interslice gap was kept to a minimum of 0.2 mm with an interleaved readout in order to avoid crosstalk. Acta Radiol 2008 (3)
360
F. Fischbach et al.
Table 2. Dependency of SNR values on field strength and resolution FSE, 5 mm
Resolution, mm SNR SNRvcor
FSE, 2 mm
1.5T
3.0T
1.5T
3.0T
0.8660.86 151.2 100%
0.4360.76 294.1 194%
0.6360.78 86.7 100%
0.4460.55 82.7 193%
SNRvcor: SNR corrected for voxel size given as percentage of SNR with base values at 1.5T set at 100%.
Image analyses Images were displayed as hard copies with no annotations in order to retain a level of objectivity. The optimal window and center settings were performed beforehand by an experienced technician in order to get the best contrast possible. Three radiologists independently rated the axial images according to the criteria set in an evaluation form. This score sheet guided the evaluation of each of the imaging examinations and included three defined evaluation criteria that served to rate the depiction of each anatomical structure on a fivepoint scale. The evaluation criteria included image quality factors important for clinical diagnostics such as the identification of a structure, its demarcation, and contrast to surrounding tissues. The criterion ‘‘identification of a structure’’ measured the confidence of the radiologist in denoting a structure. Five points were given for absolute confidence in denoting the structure with the root exit zone visible and the complete course through the cerebrospinal fluid (CSF), four points for a confident decision, three points for almost confident, two points when the structure could be separated from a vessel but did not precisely match the assumed course of the nerve, and one point was given when unsure and when the structure was difficult to denote unequivocally. The criterion ‘‘demarcation of the structure’’ related to the exact delineation and accuracy inherent in the display of the structure. A ‘‘very distinct’’ edge earned five points, ‘‘distinct’’ four points, ‘‘non-distinct’’ three points, ‘‘blurred’’ two points, and ‘‘difficult to recognize’’ one point. ‘‘Contrast of the structure’’ rated how much the structure differed from its surroundings. Five points were given for a dark and well-demarcated tubular structure, four points were given for a grayer tubular structure, three points when the amount of pulsation artifacts due to flow artifacts within the CSF increased, two points with surrounding pulsation artifacts, and one point when the structure was hard to differentiate from its surroundings. In addition, in the prepontine cistern, this criterion entailed the rating of the presence of disturbing pulsation artifacts or signal decreases due to flow artifacts within the CSF. Acta Radiol 2008 (3)
If more than one reviewer did not report an image quality criterion because no nerve could be detected, no valuable score was reported as a summary of the three reviewers. Statistical analyses In total, 240 structural entities (10 brain nerves of 12 volunteers, each measured twice) were evaluated by each reader according to the evaluation criteria described above. The sum of all scores attained with these three criteria was calculated and entered into a sum score for the imaging modality. The data were organized with MS Excel (MS EXCEL 2003, Microsoft Inc., Redmond, WA, USA), while SPSS version 12.0 (SPSS Inc., Chicago, Ill., USA) was used to test the data for significant differences between the individual protocols at 1.5T and 3.0T with use of nonparametric variance analyses for ordinal scaled data (7). Results SNR measurements In order to have an objective means for measuring the underlying SNR in the images at 1.5T and 3.0T acquired with the different pulse sequences and parameters, the overall SNRs were measured in an aqueous solution in a phantom sphere. Table 2 provides a summary of the measured SNRs for the four modalities. In keeping with the theory, it shows almost a doubling of SNR for 3.0T imaging at both section thicknesses when SNRs at 1.5T with 2-mm and 5-mm sections were defined as base values, i.e., 100% each. Of note, in 2-mm sections, where the SNR had been spent for a doubled in-plane resolution at 3.0T, a comparable SNR to 1.5T was obtained per voxel. Analyses of images The axial images obtained with the different examination protocols showed some basic differences in image quality already upon first inspection. MR images obtained both with the FSE and FRFSE sequences at 3.0T were considered less noisy though better resolved and more detailed than at 1.5T. To illustrate this observation, Fig. 1 shows
MRI at 1.5 and 3.0T of the Cranial Nerves in the Posterior Fossa
361
Fig. 1. Axial images on height of the cerebellopontine angle acquired with T2-weighted fast spin-echo with slice thicknesses of 5 mm (upper images) and 2 mm (lower images). The left two images were recorded at 1.5T. The right two images at 3.0T. Note the precise depiction of the vestibulocochlear nerve at 3.0T.
axial brainstem sections of 5-mm and 2-mm thicknesses (same individual) acquired with the FSE and FRFSE sequences, respectively, at the level of the vestibulocochlear nerves at 1.5T and at 3.0T. The increase of field strength markedly increased the number of nerves detected with 5mm and 2-mm sections. In 5-mm sections, the oculomotor nerve could be detected 18 times (90%) at 3.0T instead of 13 times (65%) at 1.5T. With the 2-mm sections, the number of nerves detected increased from 13 (65%) to 18 (90%) for the abducens nerve and from five (25%) to 15 (75%) for the hypoglossal nerve. Also, image quality characteristics gained higher scores in most instances; this was most obvious in all characteristics of the oculomotor nerve at sections of 5 mm and 2 mm, and in the identification of the glossopharyngeus-vagus complex at 5 mm as well as the facial-vestibulocochlear complex at 2-mm section thickness. Table 3 summarizes the observer scores for each nerve pair. The scores assigned by the three readers to each structural element were summarized for
each of the examination protocols, and averages for each evaluation criterion were calculated. Table 4 shows the averages obtained for the total score (sum of all three criteria) grouped by field strength. In general, significantly better scores were obtained in all three evaluation criteria for the images at 3.0T compared to 1.5T (Pv0.005) (Table 4). Discussion The main objective of this study was to assess and quantify the influence of higher field strength on image quality. Since the introduction of MRI into clinical imaging and the development toward higher field strengths, the trade-offs between attainable SNR, spatial resolution, and examination time have been a point of recurrent discussion (8, 9). While it is possible to perform a straightforward measurement of the SNR and contrast-to-noise ratio (CNR) in the images at different field strengths, these numbers do not allow for the assessment of esthetic appearance, the depiction of tiny structural details, the distinction of different tissues, impairment by Acta Radiol 2008 (3)
362
F. Fischbach et al.
Table 3. Summary of observer scores (mean) for brain nerves at 1.5T and 3.0T for both examination protocols 1.5T
3.0T
FSE, 5 mm
FRFSE, 2 mm
FSE, 5 mm
FRFSE, 2 mm
Nerve
n
%
I
D
C
n
%
I
D
C
n
%
I
D
C
n
%
I
D
C
Oculomotor (III) Trochlear (IV) Trigeminal (V) Abducens (VI) Facialvestibulocochlear complex (VII/VIII) Glossopharyngeusvagus complex (IX/X) Accessory (XI) Hypoglossal (XII)
13 2 20 14 20
65 10 100 70 100
1 — 3 2 3
1 — 2 1 2
1 — 1 1 1
20 8 20 13 20
100 40 100 65 100
3 — 4 1 4
2 — 2 1 2
2 — 2 1 2
18 4 20 12 20
90 20 100 60 100
2 — 4 2 3
2 — 2 1 2
1 — 2 1 1
20 10 20 18 20
100 50 100 90 100
4 1 4 3 4
3 1 3 2 3
2 1 3 1 3
16
80
2
1
1
20
100
4
2
1
18
90
3
2
2
20
100
4
2
2
0 2
0 10
— —
— —
— —
2 5
10 25
— —
— —
— —
2 8
10 40
— 1
— 1
— 1
2 15
10 75
— 1
— 1
— 1
I: identification; D: demarcation; C: contrast.
i.e., four times (40%) with 5-mm sections and 10 times (50%) when using 2-mm sections. Recent investigations at higher field strengths report a comparable qualitative improvement for the images obtained (11, 12). Of note, higher SNR may be more important for image impression than nominal higher resolution. On the other hand, in pivotal studies that we tried at 1.5T, images at the resolution acquired in this study at 3.0T appeared clearly overpowered, i.e., with insufficient SNR, when the same head coils, number of acquisitions, and imaging times were used. Brainstem structures can be better visualized at 3.0T compared to 1.5T, especially in T2-weighted MR imaging. The proposed improvement in SNR by a factor of two may be even more effective when utilizing shorter echo times (TEs), because T2 relaxation times decrease with higher fields (13). Moreover, increased susceptibility at 3.0T provides higher contrast of low-signal structures against the high signal intensity of cerebrospinal fluid. Of note, this high signal intensity may be severely impaired in fast spin-echo imaging by artifacts due to CSF flow in certain regions, particularly around the course of the abducens nerves. In contrast, T1 relaxation times increase with higher field strength as the Larmor and tumbling frequencies are more different and recovery of the magnetization is prolonged. The most important consequence is the need to increase
artifacts, and hence the diagnostic value of the images. Therefore, in order to assess image quality, we took the quantitative data for SNR and spatial resolution and added rating criteria such as identification of structures, delineation of their course, structural conspicuity, and image contrast, which were weighted by a scoring system. Accounting for pulse sequences and special hardware features (radiofrequency coils), this procedure provided solid data to prove a gain in image quality for the depiction of tiny structures in the posterior fossa at 3.0T compared to 1.5T. In the images of the highfield system, the highest number of nerves was identified compared to images at lower field. Generally, the course of brain nerves and brainstem vessels was best visualized at 3.0T. The pertinent structures were identified most confidently and were depicted most distinctly, while the number of artifacts remained equal compared to 1.5T. Most frequently, the complete course of the structures across the prepontine cistern could be delineated. As shown in Table 3, the trochlear nerve provides a good example of the increased power of high-field imaging, because it is one of the smallest and most difficult structures to visualize (10). At 1.5T, we were successful only in two out of 20 cases (10%) using 5-mm sections and in eight (40%) with 2-mm sections in detecting the trochlear nerve. At 3.0T, this nerve was identified clearly more frequently,
Table 4. Total average observer scores for the evaluated image quality criteria at 1.5T and 3.0T Identification
FSE, 5 mm FRFSE, 2 mm P
Demarcation
Contrast
1.5T
3.0T
1.5T
3.0T
1.5T
3.0T
3.45¡1.4 4.62¡0.7 0.015
4.09¡1.1 4.76¡0.6 0.011
1.65¡0.8 2.82¡0.7 0.000
2.82¡0.9 3.57¡0.7 0.000
1.47¡0.8 2.31¡1.0 0.000
2.21¡1.0 2.74¡0.9 0.000
Significance at Pv0.05; significance testing of 1.5T vs. 3.0T at both scanning protocols (FSE, 5 mm, and FRFSE, 2 mm). Acta Radiol 2008 (3)
MRI at 1.5 and 3.0T of the Cranial Nerves in the Posterior Fossa
the repetition time (TR) to maintain tissue contrast, which leads to an increase in acquisition times. If TR is not increased, then image contrast and SNR will be diminished at higher field strength (3). Moreover, the amount of T1 relaxation time increase depends on the tissue, which, in neuroimaging, leads to converging T1 relaxation times of gray matter and white matter and, hence, a decrease in contrast. Therefore, T1-weighted imaging is more difficult to improve, and inversion preparation of the pulse sequence is commonly preferred. However, T1-weighted images remain generally far inferior to T2-weighted images for imaging the posterior fossa. With respect to nerves and vessels, there may be a fundamental problem in distinguishing both hypointense structures in T2-weighted imaging when they run parallel or at superpositions and crossings. This problem cannot be remedied with higher field strength. While anatomical knowledge helps, additional techniques such as contrastenhanced angiography or inversion-prepared threedimensional (3D) gradient-echo sequences at 1-mm partitions may be needed for unambiguous assignment (10). In conclusion, this study of T2-weighted MR imaging employing fast spin-echo sequences clearly shows that, at comparable imaging times, delicate anatomical structures in the posterior fossa can generally be better visualized at 3.0T compared to 1.5T. Small structures show significant gains in sharpness, contrast, and detail, while images generally appear less grainy and more distinct. Recent studies also show the utility of virtual cisternography (14, 15) or 3D visualization software (16) in detecting pathologic nerve and vessel conflicts. The improved resolution and depiction shown here for 3.0T imaging with respect to the former ‘‘gold standard’’ at 1.5T will also lend itself to a more detailed database for such 3D reconstructions. References 1. Norris DG. High field human imaging. J Magn Reson Imaging 2003;18:519–29. 2. Frayne R, Goodyear BG, Dickhoff P. Magnetic resonance imaging at 3.0 Tesla: challenges and advantages in clinical neurological imaging. Invest Radiol 2003;38:385–402.
363
3. Takahashi M, Uematsu H, Hatabu H. MR imaging at high magnetic fields. Eur J Radiol 2003;46:45–52. 4. Abduljalil AM, Robitaille PM. Macroscopic susceptibility in ultra high field MRI. J Comput Assist Tomogr 1999;23:832–41. 5. Hart HR Jr, Bottomley PA, Edelstein WA. Nuclear magnetic resonance imaging: contrast-to-noise ratio as a function of strength of magnetic field. Am J Roentgenol 1983;141:1195–201. 6. Fischbach F, Lehmann TN, Ricke J, Bruhn H. Vascular compression in glossopharyngeal neuralgia: demonstration by high-resolution MRI at 3 Tesla. Neuroradiology 2003;45:810–1. 7. Brunner L, Langer F. Nonparametric analysis of longitudinal data in factorial experiments. Wiley series in probability and statistics. Munic: Wiley-Interscience; 2001. p. 288. 8. De Vita E, Thomas DL, Roberts S, Turner R, Yousry TA, Ordidge RJ. High resolution MRI of the brain at 4.7 Tesla using fast spin echo imaging. Br J Radiol 2003;76:631–7. 9. Novak P, Novak V, Kangarlu Al. High resolution MRI of the brainstem at 8 T. J Comput Assist Tomogr 2001;25:242–6. 10. Yousry I, Moriggl B, Dieterich M. MR anatomy of the proximal cisternal segment of the trochlear nerve: neurovascular relationships and landmarks. Radiology 2002;223:31–8. 11. Lane JI, Ward H, Witte RJ. 3-T imaging of the cochlear nerve and labyrinth in cochlear-implant candidates: 3D fast recovery fast spin-echo versus 3D constructive interference in the steady state techniques. Am J Neuroradiol 2004;25:618–22. 12. Robitaille PM, Abduljalil AM, Kangarlu A. Ultra high resolution imaging of the human head at 8 tesla: 2K62K for Y2K. J Comput Assist Tomogr 2000;24:2–8. 13. Bottomley PA, Foster TH, Argersinger RE, Pfeiler LM. A review of normal tissue hydrogen NMR relaxation times and relaxation mechanisms from 1–100MHz: dependence on tissue type, NMR frequency, temperature, species, excision, and age. Med Phys 1984;11:425–48. 14. Nowe´ V, Michiels JLP, Salgado R, De Ridder D, Van de Heyning PH, De Schepper AM, et al. High-resolution virtual MR endoscopy of the cerebellopontine angle. Am J Roentgenol 2004;182:379–84. 15. Naganawa S, Koshikawa T, Fukatsu H. MR cisternography of the cerebellopontine angle: comparison of three-dimensional constructive interference in the steady-state sequences. Am J Neuroradiol 2001;22:1179–85. 16. Naraghi R, Hastreiter P, Tomandi B. Three-dimensional visualization of neurovascular relationships in the posterior fossa: technique and clinical application. J Neurosurg 2004;100:1025–35.
Acta Radiol 2008 (3)
