- Email: [email protected]
Optic nerve diffusion tensor imaging in optic neuritis
www.elsevier.com/locate/ynimg NeuroImage 30 (2006) 498 – 505
Optic nerve diffusion tensor imaging in optic neuritis S. Anand Trip,a,b,* Claudia Wheeler-Kingshott,a Stephen J. Jones,c Wai-Yung Li,d Gareth J. Barker,e Alan J. Thompson,a Gordon T. Plant,b and David H. Miller a a
NMR Research Unit, Department of Neuroinflammation, Institute of Neurology, University College London, Queen Square, London WC1N 3BG, UK Department of Neuro-Ophthalmology, Moorfields Eye Hospital, City Road, London, UK c Department of Clinical Neurophysiology, National Hospital for Neurology and Neurosurgery, London, UK d Lysholm Department of Neuroradiology, The National Hospital for Neurology and Neurosurgery, Queen Square, London, UK e Centre for Neuroimaging Sciences, Department of Neurology, Institute of Psychiatry, Kings College London, London, UK b
Received 7 June 2005; revised 13 September 2005; accepted 19 September 2005 Available online 20 October 2005
Diffusion tensor magnetic resonance imaging (DT-MRI) provides in vivo information about the pathology of multiple sclerosis lesions. Increases in mean diffusivity (MD) and reductions in fractional anisotropy (FA) have been found and may represent axonal disruption. The optic nerve is an ideal structure for study by DT-MRI but previous clinical studies did not obtain the full diffusion tensor necessary to calculate MD and FA. In this study, a technique that was specifically developed to achieve full diffusion tensor measurements from the optic nerve (zonal oblique multislice (ZOOM) echoplanar imaging) was applied to 25 patients with a single unilateral episode of optic neuritis at least one year previously, and 15 controls. The intraorbital nerves were segmented on non-diffusion-weighted images and the regions of interest transferred to MD, FA, and eigenvalue maps to obtain quantitative data. Quantitative visual testing and electrophysiology were also performed. In affected nerves, mean MD and mean orthogonal eigenvalue l– – were elevated, and mean FA reduced compared with clinically unaffected contralateral nerves (P < 0.001) and control nerves (P < 0.001). The mean principal eigenvalue l\\ was significantly increased in affected nerves compared to contralateral unaffected nerves (P == 0.04) but not compared to control nerves (P == 0.13). There was no association of clinical measures of visual function in affected eyes with the DT-MRI parameters but there was a significant correlation of the whole field visual evoked potential (VEP) amplitude with MD (r == 0.57, P == 0.006) and l– = – (r = 0.56, P == 0.007). These findings suggest that optic nerve DT-MRI measures provide an indication of the structural integrity of axons. D 2005 Elsevier Inc. All rights reserved.
* Corresponding author. NMR Research Unit, Department of Neuroinflammation, Institute of Neurology, University College London, Queen Square, London WC1N 3BG, UK. Fax: +44 0 20 7278 5616. E-mail address: [email protected] (S.A. Trip). Available online on ScienceDirect (www.sciencedirect.com). 1053-8119/$ - see front matter D 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2005.09.024
Introduction Diffusion is the random motion of molecules in any fluid system including biological tissue. Diffusion-weighted magnetic resonance imaging (DW-MRI) refers to the process of making magnetic resonance imaging (MRI) sensitive to the molecular motion of water molecules, allowing the Apparent Diffusion Coefficient (ADC) to be measured in one or more directions (Wheeler-Kingshott et al., 2004). This is potentially a useful technique for studying white matter structure, as white matter tracts in the central nervous system consist of bundles of axons usually orientated in the same direction. Water diffusion can occur in any direction but occurs preferentially parallel to the orientation of axons (Takahashi et al., 2000) because their cell membranes and other oriented micro-structures act as barriers to diffusion (Von Meerwall and Fergusson, 1981). Such diffusion is said to be anisotropic and is dependent on the structural integrity of white matter tracts. Diffusion tensor magnetic resonance imaging (DTMRI) is an extension of DW-MRI in which 6 or more measurements probe diffusion in different directions and is one method by which the diffusion tensor (DT) – a complete description of diffusion in three dimensions – is calculated. Any disruption to white matter tracts or change in axonal membrane permeability would be expected to change DT indices (Horsfield and Jones, 2002; Ford et al., 1998), and in particular to lead to an increase in the mean diffusivity (MD), a measure of average molecular motion, and also to a decrease in fractional anisotropy (FA), a measure of the preponderance of diffusion direction (Basser et al., 1994). Multiple sclerosis (MS) is a common central nervous system (CNS) disease characterised pathologically by development of multifocal inflammatory demyelinating white matter lesions. It often results in major neurological disability and current disease modifying treatments have limited efficacy. The pathophysiological mechanisms of disease evolution are only partly understood and the use of MR techniques to study the evolution of lesions in
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505
vivo is an important tool for obtaining better insights in to such mechanisms. DW-MRI and DT-MRI have supplemented standard MR techniques for studying pathology in vivo. Studies in MS have shown increased ADC and MD with decreased FA in chronic T1hypointense lesions compared to T1-isointense lesions which is compatible with evidence that T1-hypointense lesions represent more extensive tissue loss (Barkhof et al., 2000; Filippi and Inglese, 2001). FA is lower in acute, gadolinium-enhancing lesions compared to non-enhancing lesions, probably because extracellular oedema alters the anisotropic pattern of diffusion (Werring et al., 1999; Filippi and Inglese, 2001). ADC and MD values are elevated, but the extent may depend upon the lesion age (Werring et al., 1999; Roychowdhury et al., 2000). Optic neuritis is a common manifestation in MS, and its study offers special opportunities to explore the pathophysiology of individual inflammatory demyelinating CNS lesions. This is because the optic nerve is a discrete anatomical structure, the function of which can be more reliably assessed by quantitative clinical and electrophysiological measures than is possible for other CNS pathways, e.g. motor function. The lesion of optic neuritis can also be clearly identified with MRI provided that fatsaturation methods are used (Gass et al., 1996). As the optic nerve is almost a pure white matter tract, it should lend itself to study with DT-MRI. There are, however, particular challenges associated with in vivo imaging of the optic nerve with MRI (Barker, 2000), and especially with DW-MRI/DT-MRI (Barker, 2001). The optic nerves are small, mobile structures surrounded by CSF and orbital fat. High resolution CSF and fat-saturated techniques are desirable particularly as these minimise the effects of partial volume in quantitative imaging. DW-MRI is sensitised to microscopic motion of water molecules, but this raises additional problems as it is extremely sensitive to macroscopic motion to which the optic nerves are susceptible due to eye movement. Furthermore, the high resolution needed to image small structures such as the optic nerve results in low MR signal-to-noise ratio (SNR) which needs special attention during image processing and analysis. Until recently, attempts at DW-MRI of the optic nerve have been restricted to measuring the ADC along a limited number of diffusion directions (Iwasawa et al., 1997; Freeman et al., 1998; Wheeler-Kingshott et al., 2002a; Hickman et al., 2005) and have therefore not obtained the full DT which can provide rotationally invariant indices including MD, FA, and diffusion eigenvalues. A CSF- and fatsuppressed zonal oblique multislice echoplanar imaging (ZOOMEPI) sequence (Wheeler-Kingshott et al., 2002a) has recently been developed further in the optic nerve by increasing the number of
499
diffusion directions from three to six, thus allowing full DT analysis and the calculation of MD and FA in healthy volunteers (Wheeler-Kingshott et al., 2004). In the current study, we present the results of DT-MRI in a cohort of patients who had a previous episode of unilateral optic neuritis and compare the findings with those obtained in healthy controls. The aims of the study were to investigate if the rotationally invariant indices produced by the DT were different in nerves affected by optic neuritis, and whether any changes were related to measures of visual function and/or electrophysiological markers of optic nerve function.
Methods Subjects Twenty-five patients with a single previous attack of acute unilateral optic neuritis and no recurrence were recruited from the case records of the Neuro-ophthalmology clinic, Moorfields Eye Hospital, London. In all patients, appropriate investigations had been made to exclude an alternative aetiology to the optic neuropathy. In order to study a range of visual deficits, there was a deliberate selection bias towards those with incomplete visual recovery. Patient demographic data are summarised in Table 1. Patients with a clinical diagnosis of MS (Poser et al., 1983) were included in the study if they had only one attack of unilateral optic neuritis. Both the eye affected by optic neuritis and the clinically unaffected eye were investigated in all patients. Fifteen controls subjects were also studied, none of whom were known to have any ophthalmological or neurological disorder. The mean age was 36 years (range 30 – 56 years) with 9 women and 6 men. One eye from each control subject was randomly chosen for study. Ethical approval for the study was obtained from the joint ethics committee of the Institute of Neurology and the National Hospital for Neurology and Neurosurgery and the Moorfields Eye Hospital Research Governance Committee. Informed consent in writing was obtained from all subjects, in accordance with the Declaration of Helsinki. DT-MRI Images were acquired with a 1.5 T GE Signa scanner (General Electric, Milwaukee, WI) with maximum gradient strength of 22 mTm 1. A quadrature birdcage coil was used as both transmitter
Table 1 Demographic data and visual function of patients Male:Female MS:CIS Age (years) Time since onset of optic neuritis (years) LogMAR visual acuity Visual field mean deviation (dB) Colour vision (FM 100-Hue score)
Affected Unaffected Affected Unaffected Affected Unaffected
Mean
SD
Range
Ratio
– – 40.6 3.1 +0.17 0.07 6.6 2.9 19.2 11.1
– – 9.4 1.7 0.32 0.09 4.7 3.8 7.9 2.6
– – 22.4 – 57.0 1.0 – 6.7 0.08 – +1.40 0.20 – +0.20 18.5 – 0.9 14.8 – +0.68 10.2 – 36.6 7.5 – 16.1
12:13 10:15 – – – – – – – –
CIS indicates clinically isolated syndrome; MAR, minimum angle of resolution; FM, Farnsworth – Munsell.
500
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505
and receiver. Subjects were asked to close their eyes and avoid deliberate eye movements during scanning. Coronal optic nerve images were obtained using a single-shot ZOOM-EPI DT-MRI sequence (Wheeler-Kingshott et al., 2004) with the following parameters: field of view 80 40 mm (with only the inner volume, IV = 26 mm, fully excited); matrix size 64 32; pixel size 1.25 1.25 mm. To increase the efficiency of the sequence, diffusion gradients were applied at maximum strength along 6 pairs of orthogonal directions (1, 1, 0; 1, 0, 1; 0, 1, 1; 1, 1, 0; 1, 0, 1; 0, 1, 1). The diffusion parameters were d, D = 20, 27 ms; b = 600 s _ mm 2. The echo time TE was 83.6 ms. One non-DW (b 0) set of images was acquired. To reduce the partial volume effects due to CSF and fat signal, an inversion pulse was used to null signal from CSF (TI = 1400 ms, TR = 3400 ms) and a frequency-selective spectral – spatial pulse (exciting only water) was used for excitation. ZOOM-EPI requires a spatial gap between slices acquired consecutively in time, therefore we chose to image 3 slices with a thickness of 4 mm and slice gap of 8 mm. The motion of the ON is ‘‘frozen’’ during each single-shot acquisition of one image, but motion between successive acquisitions is extremely difficult to control; averaging over many images was found to be the best way to allow for this, resulting in high quality images of the mean position of the nerve (Wheeler-Kingshott et al., 2000). Averaging was also necessary because of the low SNR which characterises the images (SNR of DW < 5:1). For our protocol, 44 averages were collected for off-line averaging (see below). The total acquisition time for the DT-MRI (with 3 dummy cycles) was 17 min 40 s for a total of 924 images. The averaging process to obtain the final b 0 and 6 DW images was performed on the magnitude signal. The low SNR of the magnitude images meant that a bias could be introduced in the DT calculations. An extra scan of just noise (total scan time = 3 min) was acquired and used during post-processing to perform Rayleigh
noise correction of the signal intensity, before DT calculations (Miller and Joseph, 1993; Wheeler-Kingshott et al., 2002a). A nonlinear smoothing algorithm was applied to each set of images to further reduce noise while preserving structure (Parker et al., 2000). Review of the acquired DT images revealed that the most anterior of the 3 slices was contaminated by the posterior aspect of the globe in 52% of nerves affected by optic neuritis, in 60% of clinically unaffected patient nerves, and in 47% of control nerves. Because the optic nerve could not be clearly segmented in so many of these slices, the anterior slice could not be used for analysis. In the posterior intracanalicular/intracranial slice, the optic nerve could not be adequately identified in 28% of nerves affected by optic neuritis, in 32% of clinically unaffected patient nerves, and in 33% of control nerves. Again, this slice was not used for analysis. The intraorbital nerve was clearly seen in the middle slice of all subjects and was used for analysis. The DT images were processed and analysed on Unix workstations (Sun Microsystems, Mountain View, CA, USA). MD, FA, and eigenvector maps were calculated on a pixel by pixel basis. Images were displayed and analysed using the DispImage display tool (Plummer, 1992). To avoid bias, each optic nerve was manually segmented by a single blinded observer from the nondiffusion-weighted (b 0) image. A square 2 2 voxel region of interest (ROI) was placed over the centre of each optic nerve on the mean b 0 image averaged over all 44 acquisitions, using maximum signal intensity and minimum standard deviation as guidance towards the optimum position. These ROIs were then automatically transferred to the Rayleigh noise corrected MD, FA, and eigenvector maps (Fig. 1) where mean values from the 4 voxels were obtained. Three eigenvalues were produced from each ROI: the largest was the principal eigenvalue (k \\) representing the diffusion coefficient along the principal direction of diffusion parallel to
Fig. 1. Intraorbital slice from a 36-year-old female patient 2 years after left sided optic neuritis. (A) Non-diffusion weighted b 0 image, (B) MD map, (C) FA map, (D) principal eigenvalue k \\ map, and (E) one of two orthogonal eigenvalue k – maps. The arrow indicates the affected nerve.
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505
the nerve; the other two were the diffusion coefficients orthogonal to the nerve. The average of the two orthogonal eigenvalues was expressed as k –. To avoid the introduction of bias by partial volume effect, careful note was taken of the signal characteristics of each voxel within the ROI when transferred to the DT maps. Any ROI which contained one or more ‘‘empty’’ voxels, where the DT-MRI calculation failed (usually due to low SNR), was excluded. This led to the exclusion of 3 affected and 2 clinically unaffected patient nerves and 1 control nerve. To assess measurement reproducibility, measurements were repeated in 10 patients and 10 controls, who were all randomly selected and again analysed in a blinded fashion. Lesion identification using fast spin echo (FSE) Patients also had optic nerve imaging with a dual echo FSE sequence (coronal-oblique, TR 2300 ms, TEeffective 58/145 ms, echo-train length 8, 2 excitations, matrix size 512 384, field of view 24 18 cm, in-plane resolution 0.5 0.5 mm, 16 3 mm interleaved contiguous slices, 11 min acquisition time). A neuroradiologist (WYL), blinded to the side of clinical involvement and severity of visual impairment, identified and measured the length of the lesion (the number of slices affected 3 mm) on the FSE images.
501
horizontally by 20- vertically, using corneal surface electrodes (DTL Plus, Retina Technologies) referred to skin-surface electrodes over the ipsilateral outer canthus. 40V check sizes were used, reversing 4.3 times per second. The luminance of the bright squares was 60 cd/m2 and of the dark squares was 4 cd/m2. The amplifier corner frequencies were 1 and 1000 Hz. The sampling rate was 3 samples/ms and the sweep duration was 170 ms. We made 3 averages of 200 responses and subsequently averaged them together. Sweeps containing artefacts of more than 165AV were automatically rejected. Analysis was performed blind to the status of each subject. Statistical analysis Statistical analyses were performed using SPSS 11.5 for Windows (SPSS, Chicago, IL, USA). In order to account for potential age and gender effects, differences between patients and controls were assessed using multiple linear regression models, with the MRI parameter as the response variable, disease status, and gender as categorical variables, and age as a continuous covariate. Differences between affected eye and unaffected eye within patients were explored using the Wilcoxon signed ranks test. Spearman rank correlations were calculated between MRI parameters, visual function, and electrophysiological variables. Data are presented uncorrected for multiple comparisons and therefore appropriate allowance should be made by the reader when interpreting the data.
Clinical testing of visual function Results Visual acuity of patients with appropriate refraction was measured using a retro-illuminated ETDRS chart and recorded as the 4 m logMAR acuity (Ferris et al., 1982). The central 30- of the visual field was analysed using the 30 – 2 program on the Humphrey field analyser (Allergan-Humphrey Inc., San Leandro, CA, USA). Wide-angle lenses were used to correct refractive errors where necessary. The overall visual field mean deviation was compared with a reference field derived from control data provided by the manufacturer. One patient, who had near normal acuity and normal visual fields by clinical examination, was unable to reliably complete automated perimetry due to high levels of fixation losses and false negative responses. Colour vision was assessed using the Farnsworth – Munsell 100-Hue test (Farnsworth, 1943) and scored as square root of the error score (FM 100-Hue score) because this follows a normal distribution. Two patients with a congenital anomaly of colour vision were excluded from this test. These three visual function parameters were used because they give continuously variable measures that are amenable to statistical analysis. Electrophysiological assessment of optic nerve function Visual evoked potential (VEP) amplitudes and latencies were recorded to monocular stimuli using skin-surface EEG electrodes attached over the occiput, 5 cm above the inion, referred to a frontal electrode at Fz (10 – 20 System). The stimuli comprised reversal of a checkerboard pattern in the ‘‘whole field’’ and in the ‘‘central field’’ as previously described (Brusa et al., 2001). Central field responses were unobtainable in one patient and one control. The pattern electroretinogram (PERG) N95 component is thought to be of retinal ganglion cell origin and can thus be reduced in optic neuropathies including optic neuritis by a process of retrograde degeneration (Holder, 1991). The PERG was recorded to binocular stimulation of the ‘‘whole field’’, subtending 28-
Clinical data of patients Table 1 summarises patient demographics including range of visual function. Electrophysiology data Table 2 presents the electrophysiology data. In affected eyes, there was no correlation of visual acuity with electrophysiological measures but the visual field mean deviation correlated with the whole field (r = 0.46, P = 0.03) and central field (r = 0.46, P = 0.03) VEP amplitudes, and colour vision (FM 100-Hue score) was correlated with whole field VEP (r = 0.49, P = 0.02), central field VEP (r = 0.69, P < 0.001), and PERG N95 (r = 0.47, P = 0.02) amplitudes. No correlations were present with VEP latencies. Measurement reproducibility for DT-MRI As the ROI size was fixed at 2 2 voxels and the placement was guided by quantitative measures, the repeat measurements for all DT-MRI parameters were exactly the same. Group differences in DT-MRI data Table 3 summarises all the group DT-MRI data. The mean MD from nerves affected by optic neuritis was elevated at 1611 [SD 290] 10 6 mm2s 1, compared with 1191 [SD 198] 10 6 mm2s 1 from clinically unaffected contralateral nerves ( P < 0.001 versus affected nerves) and 1083 [SD 168] 10 6 mm2s 1 from control nerves ( P < 0.001 versus affected nerves). There was no significant difference between control nerves and patients’ unaffected contralateral nerves ( P = 0.29).
502
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505
Table 2 Electrophysiology group data Control nerves (n = 15)
Unaffected patient nerves (n = 25)
Affected patient nerves (n = 25) 5.6 [3.4] P c = 0.003* P f = 0.001* 2.9 [1.8] P c = 0.002* P f = 0.004* 110.9 [18.3] P c = 0.02* P f = 0.21 115.9 [23.2] P c = 0.001* P f = 0.01* 5.2 [2.1] P c = 0.47 P f = 0.001* 98.6 [14.8] P c = 0.06 P f = 0.003*
Whole field VEP amplitude, AV
9.9 [5.4]
8.9 [4.8] P c = 0.78
Central field VEP amplitudea, AV
5.9 [3.3]
5.3 [3.5] P c = 0.84
Whole field VEP latency, ms
97.0 [5.8]
106.6 [17.8] P c = 0.06
Central field VEP latencya, ms
91.8 [6.9]
98.7 [20.0] P c = 0.26
6.0 [2.7]
6.7 [2.2] P c = 0.13
87.9 [9.0]
89.3 [7.7] P c = 0.99
PERG N95 amplitude, AV PERG N95 latency, ms
There were no significant differences in any of the DT parameters between the 15 patients with clinically isolated optic neuritis and the 10 patients with MS. Correlation of MR-DTI data with visual function and electrophysiology There was no association of the clinical measures of visual function with any of the DT-MRI parameters. However, there was a significant correlation of the whole field VEP amplitude with MD in both affected eyes (r = 0.57, P = 0.006) and unaffected contralateral eyes (r = 0.43, P = 0.04), and also with k – in affected eyes (r = 0.56, P = 0.007). Central field VEP amplitude in unaffected contralateral eyes correlated with MD (r = 0.44, P = 0.04) and k – (r = 0.54, P = 0.008) whereas there were only trends in affected eyes. The PERG N95 amplitude showed non-significant trends for correlation with (i) affected eye MD (r = 0.42, P = 0.06) and (ii) unaffected contralateral eye MD (r = 0.35, P = 0.10) as well as a significant correlation with unaffected contralateral eye k – (r = 0.45, P = 0.03). There were also non-significant trends for correlation of affected eye FA with whole field VEP amplitude (r = 0.37, P = 0.09) and unaffected contralateral eye FA with central field VEP amplitude (r = 0.39, P = 0.06) (Table 4). No association was found with any of the DT-MRI measures and either VEP or PERG latencies.
Values are mean [SD]. P c values are comparisons between control nerves and patient nerves. P f values are comparisons between unaffected contralateral nerves and affected nerves in patients. a Controls n = 14, patients n = 24. * P < 0.05.
Lesion length on FSE A single lesion was identified in all of the 25 affected optic nerves. Mean lesion length was 26 mm [SD 7] with a range of 12 – 39 mm. The FSE imaging coordinates were used to determine if the 4 mm intraorbital slice contained lesioned nerve. In 22 of the 25 patients, the DT-MRI slices contained lesion throughout, in 2 it partly contained lesion, and in 1 there was no lesion within the slice (in this patient, there was a short canalicular lesion visible on FSE). Correlations with visual function and electrophysiology did not change significantly when the part-lesion and lesion-free cases were excluded from analysis. There was no significant correlation between lesion length and DT-MRI parameters but there was a borderline association with k – (r = 0.40, P = 0.07).
Mean FA from affected nerves was lower at 0.435 [SD 0.100], compared with 0.602 [SD 0.100] from clinically unaffected contralateral nerves (P < 0.001) and 0.669 [SD 0.093] from control nerves (P < 0.001). There was a non-significant trend for FA to be lower in patients’ unaffected contralateral nerves versus control nerves (P = 0.07). The mean orthogonal eigenvalue k – from affected nerves was raised at 1217 [SD 326] 10 6 mm2s 1, compared with 730 [SD 196] 10 6 mm2s 1 from clinically unaffected contralateral nerves (P < 0.001) and 629 [SD 199] 10 6 mm2s 1 from control nerves (P < 0.001). There was no significant difference between control nerves and patients’ unaffected contralateral nerves (P = 0.26). The mean principal eigenvalue k \\ was marginally higher in affected nerves at 2355 [SD 509] 10 6 mm2s 1 compared to 2117 [SD 336] 10 6 mm2s 1 in unaffected contralateral nerves (P = 0.04) and 2088 [SD 309] 10 6 mm2s 1 in control nerves (P = 0.13). There was no significant difference between control nerves and patients’ unaffected contralateral nerves (P = 0.80).
Discussion Using a ZOOM-EPI technique that was specifically developed for investigating the optic nerve, it was possible to measure the full DT in the optic nerves of controls and patients with previous
Table 3 DT-MRI parameter group data MD (10 Control nerves (n = 14) Patient unaffected nerves (n = 23) Patient affected nerves (n = 22)
6
1083 [168] 1191 [198] P c = 0.29 1611 [290] P c < 0.001* P f < 0.001*
mm2s 1)
6
FA
k \\ (10
0.669 [0.093] 0.602 [0.100] P c = 0.07 0.435 [0.100] P c < 0.001* P f < 0.001*
2088 [309] 2117 [336] P c = 0.80 2355 [509] P c = 0.13 P f = 0.04*
mm2s 1)
k – (10
6
mm2s 1)
629 [199] 730 [196] P c = 0.26 1217 [326] P c < 0.001* P f < 0.001*
Values are mean [SD]. P c values are comparisons between control nerves and patient nerves. P f values are comparisons between unaffected contralateral nerves and affected nerves in patients. * P < 0.05.
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505
503
Table 4 Correlation of DT-MRI measures with electrophysiology amplitude data in patients MD
Whole field VEP amplitude Central field VEP amplitude PERG N95 amplitude
k \\
FA
k–
Affected
Unaffected
Affected
Unaffected
Affected
Unaffected
Affected
Unaffected
r= P= r= P= r= P=
r= P= r= P= r= P=
r = 0.37 P = 0.09 r = 0.28 P = 0.23 r = 0.30 P = 0.17
r = 0.01 P = 0.96 r = 0.39 P = 0.06 r = 0.13 P = 0.57
r= P= r= P= r= P=
r= P= r= P= r= P=
r= P= r= P= r= P=
r= P= r= P= r= P=
0.57 0.006* 0.37 0.10 0.42 0.06
0.43 0.04* 0.44 0.04* 0.35 0.10
0.14 0.54 0.06 0.80 0.26 0.24
0.33 0.13 0.06 0.80 0.07 0.75
0.56 0.007* 0.42 0.06 0.24 0.28
0.17 0.43 0.54 0.008* 0.45 0.03*
r values are Spearman Rank Correlation Coefficients. * P < 0.05.
unilateral optic neuritis in clinically acceptable scan times, with fatand CSF-suppression to minimise partial volume effects. This was achieved by increasing the number of diffusion directions from three (sufficient for ADC measurement only) to six, and by using a high number of averages for each diffusion weighting to compensate for motion of the optic nerves. The DT produced rotationally invariant indices: MD, FA, and diffusion eigenvalues which provided useful in vivo information about the changes that occur after optic neuritis. Although it was hoped to obtain quantitative data from three 4 mm slices, contamination of the optic nerve by the globe resulted from the anterior slice being prescribed too anteriorly in some subjects, and the posterior slice did not allow reliable segmentation of the optic nerve. The intraorbital slice (the middle of the 3) did, however, allow reliable segmentation of the optic nerve on non-diffusion weighted b 0 images in all subjects. Special care was taken to exclude a small number of nerves in which the ROI did not completely contain sufficient optic nerve signal when transferred to the DT-MRI maps, to ensure that partial volume effects were kept to a minimum. Previous studies of optic nerve DW-MRI have not reported the full DT, or did not include methods to suppress fat and CSF. Iwasawa et al. (1997) used a spin-echo sequence with low diffusion weighting and cardiac gating to study a single intraorbital optic nerve slice. Three diffusion directions were acquired but due to motion artefact, ADC could only be measured in the y and z directions and not always reliably. A small 1 mm diameter ROI was used. The mean ADC (4180 10 6 mm2s 1, n = 9) in the optic nerves previously affected by optic neuritis was significantly higher than that in controls (1559 10 6 mm2s 1, n = 14) and that in the nerves with acute optic neuritis (941 10 6 mm2s 1, n = 4) ( P < 0.001). Hickman et al. (2005) used ZOOM-EPI to study the optic nerve ADC in 16 patients who were seen 1 year following an attack of optic neuritis, and in 10 controls. Four 4 mm slices covering the intraorbital optic nerve were used to determine mean ADC which was 1324 10 6 mm2s 1 in affected nerves, compared to 990 10 6 mm2s 1 in unaffected contralateral nerves ( P = 0.005) and 928 10 6 mm2s 1 in control nerves ( P = 0.006). The affected nerve ADC correlated with measures of visual function (visual acuity, visual field, and colour vision) and electrophysiological parameters (VEP amplitudes and latencies). The only study which attempted to measure FA as well as ADC was that by Chabert et al. (2002) who found – using a non CPMGFSE sequence – that the MD was 1670 10 6 mm2s 1 with a mean FA of 0.59 in four healthy volunteers. They found that the fibre directions followed the expected nerve fibre directions on anisotropy maps. However, the sequence was neither CSF- nor fat-
suppressed so the results may have been affected by partial volume effects. This may account for the higher MD and lower FA that was obtained compared to our 15 controls. The optic nerve MD and ADC values of controls in all of these studies and in our control nerves (using different sequences and different scanners) are consistently higher than those in deep brain white matter and are likely to reflect structural differences between the optic nerve and brain white matter tracts. The Raleigh noise correction compensates for the otherwise artefactually low values due to low SNR without correction rather than erroneously elevating diffusivity (Wheeler-Kingshott et al., 2002a). Further evidence supporting the latter conclusion comes from a study of normal cervical cord structure using ZOOM-EPI with the same scanner as the present study which found comparable cervical cord MD values (940 10 6 mm2s 1) to the optic nerve despite the absence of Raleigh noise correction (Wheeler-Kingshott et al., 2002b). The current study has found that, as expected, MD is significantly increased and FA is significantly reduced in optic nerves affected by optic neuritis compared to unaffected contralateral nerves and healthy control nerves. These findings are compatible with a process of axonal disruption or loss, although demyelination and gliosis might also contribute. The increase in MD and decrease in FA seen in what were predominantly chronic optic nerve lesions resemble those reported in studies of chronic brain lesions in MS (Werring et al., 1999; Bammer et al., 2000; Filippi and Inglese, 2001; Dong et al., 2004), although the absolute values for both MD and FA seen in the present study were somewhat greater than those reported in brain lesions. The higher FA may reflect the greater baseline anisotropy of the optic nerves compared to brain white matter tracts in the normal state and the higher MD could be due to the bias to poor recovery we had in our cohort. However, differences in the DT-MRI sequences used may also be a factor. The present study did not find an association of any DT parameter with measures of visual function. A previous study by Hickman et al. (2005) found such an association but differed in that their patients were all examined at the same 1 year time point following optic neuritis when one would expect a similar pattern of recovery processes taking place in each patient, whereas the present cohort was heterogeneous in terms of disease duration and therefore may have differing degrees of recovery processes. Perhaps more importantly, the present study surveyed only 4 mm of optic nerve compared to the 16 mm of nerve surveyed in the study of Hickman et al. (2005); the less comprehensive sampling of the lesions may have reduced the opportunity to detect clinically relevant DT changes. The VEP amplitude is believed to reflect the number of optic nerve fibres in functional continuity (Jones and Brusa, 2003) and a
504
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505
reduction in amplitude in optic neuritis could be due to two mechanisms. In the acute phase of optic neuritis associated with substantial visual impairment, there is a marked loss of VEP amplitude (Jones, 1993), often abolishing the response completely, which usually resolves over a period of a few weeks concurrent with the period when gadolinium enhancement on MRI disappears (Youl et al., 1991; Hickman et al., 2004). This suggests that the VEP amplitude reduction in the acute phase, characterised by blood – brain barrier leakage, inflammation, oedema, and demyelination, is caused by temporary conduction block in optic nerve axons. Conduction block from inflammation and oedema is unlikely to be the cause of the reduced VEP amplitude in the present cohort as they were all at least 1 year from onset of optic neuritis. The reduced VEP amplitude arising from the chronic lesions that we studied could reflect either: (i) conduction block from persistent demyelination or (ii) axonal loss. The latter will likely have resulted from acute inflammation-induced axonal transection (a well described finding in acute inflammatory lesion in multiple sclerosis; Trapp et al., 1998), although post-inflammatory mechanisms for axonal loss may also contribute (Compston and Coles, 2002). The significant correlation of VEP amplitude reduction with MD increase suggests that the increased diffusivity is of functional relevance and may reflect either axonal loss or conduction block in surviving axons. This association was found in the clinically affected eyes (whole field only) and also in the clinically unaffected eyes (whole and central fields). The latter finding is of interest as two previous follow-up studies have found nonsignificant marginal reductions in the VEP amplitude of unaffected fellow eyes over time which has been postulated to be due to an insidious process of axonal degeneration (Brusa et al., 1999, 2001). The findings in the asymptomatic nerves in the present study would support this hypothesis. The lack of correlation with central field VEP amplitude in affected eyes may be due to the lower SNR of this measure. As the PERG N95 response is believed to be of retinal ganglion cell origin and shows less consistent changes than the VEP in optic neuritis (Holder, 1991), it is not unexpected that there were no convincing correlations with DT-MRI, and the correlation demonstrated with the unaffected eye k – is more likely to be due to type 1 error. The mean of the orthogonal eigenvalues k –, which reflects diffusion orthogonal to the axis of the optic nerve, behaved in a similar way to MD in that it was significantly increased in affected nerves and correlated with whole field VEP amplitude to a similar extent. The principal eigenvalue k \\, however, was not significantly different between the three groups of optic nerves. A possible explanation for this is that although some axons are lost following optic neuritis, the remaining axons are still aligned in the same direction (along the optic nerve) thus maintaining a relatively normal k \\. The spaces between the remaining axons may allow increased diffusion of water molecules in the parallel direction which could counteract the effects of axonal disruption reducing anisotropy. At the same time, increased spaces between axons would be expected to contribute to an increase in k – and MD. Changes in FA values may occur secondary to changes in any or all of the three eigenvalues. In inflammatory conditions such as optic neuritis, all three eigenvalues could increase and therefore any percentage change in FA is likely to be less than the change in any given eigenvalue. If all three eigenvalues increased together to the same extent, then it would be possible for FA to be unchanged thus making a case that eigenvalues may be more sensitive than FA in
quantifying pathology. It may be of interest in future studies to look at both FA and eigenvalues. Future studies will need to concentrate on overcoming the methodological difficulties specific to DT-MRI of the optic nerve. Higher resolution imaging would help to decrease partial volume; improved SNR should result from increases in gradient strength, the use of multichannel receiver coils, and higher field scanners. Parallel imaging techniques were not widely available at the time this study was started. The main advantage of parallel imaging is that it is implemented by the manufacturer and therefore does not require special set up. It is not limited, like ZOOM-EPI, to noncontiguous slices/multiple acquisitions. The disadvantage is that it will reduce the already very low SNR still further and although this can, in theory, be made up by further averaging (albeit at the expense of longer scan times), it is not obvious what the effect of such low SNR will be on the algorithms used for the image reconstruction, nor how Rayleigh noise correction will interact with the parallel imaging reconstruction. More work would be needed to determine at what point the noise correction was best applied and to devise algorithms that can cope with position dependent sensitivity and ‘‘geometry factor’’. The challenge will be to produce a sequence with good coverage of the whole optic nerve with clinically acceptable scanning times; development of such methods could provide a powerful tool to study the effect of novel therapies that protect the optic nerve and its axons from damage in optic neuritis and other optic neuropathies.
Acknowledgments The NMR Research Unit is supported by the Multiple Sclerosis Society of Great Britain and Northern Ireland. We thank Dr. Simon Hickman for helpful comments.
References Bammer, R., Augustin, M., Strasser-Fuchs, S., et al., 2000. Magnetic resonance diffusion tensor imaging for characterizing diffuse and focal white matter abnormalities in multiple sclerosis. Magn. Reson. Med. 44, 583 – 591. Barker, G.J., 2000. Technical issues for the study of the optic nerve with MRI. J. Neurol. Sci. 172 (Suppl. 1), S13 – S16. Barker, G.J., 2001. Diffusion-weighted imaging of the spinal cord and optic nerve. J. Neurol. Sci. 186 (Suppl. 1), S45 – S49. Barkhof, F., Karas, G.B., van Walderveen, M.A., 2000. T1 hypointensities and axonal loss. Neuroimaging Clin. N Am. 10 (4), 739 – 752 (Nov). Basser, P.J., Mattiello, J., LeBihan, D., 1994. MR diffusion tensor spectroscopy and imaging. Biophys. J. 66, 259 – 267. Brusa, A., Jones, S.J., Kapoor, R., Miller, D.H., Plant, G.T., 1999. Long-term recovery and fellow eye deterioration after optic neuritis, determined by serial visual evoked potentials. J. Neurol. 246 (9), 776 – 782 (Sep). Brusa, A., Jones, S.J., Plant, G.T., 2001. Long-term remyelination after optic neuritis: a 2-year visual evoked potential and psychophysical serial study. Brain 124, 468 – 479. Chabert, S., Molko, N., Cointepas, Y., Le Roux, P., Le Bihan, D., 2002. Functional diffusion tensor imaging of the optic nerve using a non CPMG fast spin echo sequence. Proc. Intl. Soc. Mag. Reson. Med., 1115. Compston, A., Coles, A., 2002. Multiple sclerosis. Lancet 359, 1221 – 1231.
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505 Dong, Q., Welsh, R.C., Chenevert, T.L., et al., 2004. Clinical applications of diffusion tensor imaging. J. Magn. Reson. Imaging 19, 6 – 18. Farnsworth, D., 1943. The Farnsworth – Munsell 100-hue and dichotomous tests for color vision. J. Opt. Soc. Am. 33, 568 – 578. Ferris, F.L. III, Kassoff, A., Bresnick, G.H., Bailey, I., 1982. New visual acuity charts for clinical research. Am. J. Ophthalmol. 94, 91 – 96. Filippi, M., Inglese, M., 2001. Overview of diffusion-weighted magnetic resonance studies in multiple sclerosis. J. Neurol. Sci. 186 (Suppl. 1), S37 – S43. Ford, J.C., Hackney, D.B., Lavi, E., Phillips, M., Patel, U., 1998. Dependence of apparent diffusion coefficients on axonal spacing, membrane permeability, and diffusion time in spinal cord white matter. J. Magn. Reson. Imaging 8 (4), 775 – 782 (Jul – Aug). Freeman, A.J., Ballinger, R., Werner, M., Guy, J., 1998. Diffusion weighted EPI of the human optic nerve at 3 T. Proc. Int. Soc. Mag. Reson. Med. 6, 1265. Gass, A., Moseley, I.F., Barker, G.J., et al., 1996. Lesion discrimination in optic neuritis using high-resolution fat-suppressed fast spin-echo MRI. Neuroradiology 38, 317 – 321. Hickman, S.J., Toosy, A.T., Miszkiel, K.A., et al., 2004. Visual recovery following acute optic neuritis—A clinical, electrophysiological and magnetic resonance imaging study. J. Neurol. 251, 996 – 1005. Hickman, S.J., Wheeler-Kingshott, C.A., Jones, S.J., Miszkiel, K.A., Barker, G.J., Plant, G.T., Miller, D.H., 2005. Optic nerve diffusion measurement from diffusion weighted imaging in optic neuritis. AJNR Am. J. Neuroradiol. 26, 951 – 956. Holder, G.E., 1991. The incidence of abnormal pattern electroretinography in optic nerve demyelination. Electroencephalogr. Clin. Neurophysiol. 78, 18 – 26. Horsfield, M.A., Jones, D.K., 2002. Applications of diffusion-weighted and diffusion tensor MRI to white matter diseases—A review. NMR Biomed. 15 (78), 570 – 577. Iwasawa, T., Matoba, H., Ogi, A., et al., 1997. Diffusion-weighted imaging of the human optic nerve: a new approach to evaluate optic neuritis in multiple sclerosis. Magn. Reson. Med. 38, 484 – 491. Jones, S.J., 1993. Visual evoked potentials after optic neuritis. Effect of time interval, age and disease dissemination. J. Neurol. 240, 489 – 494. Jones, S.J., Brusa, A., 2003. Neurophysiological evidence for long-term repair of MS lesions: implications for axon protection. J. Neurol. Sci. 206, 193 – 198. Miller, A.J., Joseph, P.M., 1993. The use of power images to perform quantitative analysis on low SNR MR images. Magn. Reson. Imaging 11, 1051 – 1056.
505
Parker, G.J., Schnabel, J.A., Symms, M.R., Werring, D.J., Barker, G.J., 2000. Nonlinear smoothing for reduction of systematic and random errors in diffusion tensor imaging. J. Magn. Reson. Imaging 11, 702 – 710. Plummer, D.L., 1992. DispImage: a display and analysis tool for medical images. Rivista di Neuroradiologia 5, 489 – 495. Poser, C.M., Paty, D.W., Scheinberg, L., et al., 1983. New diagnostic criteria for multiple sclerosis: guidelines for research protocols. Ann. Neurol. 13, 227 – 231. Roychowdhury, S., Maldjian, J.A., Grossman, R.I., 2000. Multiple sclerosis: comparison of trace apparent diffusion coefficients with MR enhancement pattern of lesions. AJNR Am. J. Neuroradiol. 21, 869 – 874. Takahashi, M., Ono, J., Harada, K., Maeda, M., Hackney, D.B., 2000. Diffusional anisotropy in cranial nerves with maturation: quantitative evaluation with diffusion MR imaging in rats. Radiology 216 (3), 881 – 885 (Sep). Trapp, B.D., Peterson, J., Ransohoff, R.M., et al., 1998. Axonal transection in the lesions of multiple sclerosis. N. Engl. J. Med. 338, 278 – 285. Von Meerwall, E., Fergusson, R.D., 1981. Interpreting pulsed-gradient spinecho diffusion experiments with permeable membranes. J. Chem. Phys. 74, 6956 – 6959. Werring, D.J., Clark, C.A., Barker, G.J., Thompson, A.J., Miller, D.H., 1999. Diffusion tensor imaging of lesions and normal-appearing white matter in multiple sclerosis. Neurology 52, 1626 – 1632. Wheeler-Kingshott, C.A.M., Ciccarelli, O., Parker, G.J.M., Barker, G.J., 2000. Quantification of the motion of the optic nerve. Proc. ESMRMB 17th Annual, p. 443. Wheeler-Kingshott, C.A., Parker, G.J., Symms, M.R., et al., 2002. ADC mapping of the human optic nerve: increased resolution, coverage, and reliability with CSF-suppressed ZOOM-EPI. Magn. Reson. Med. 47, 24 – 31. Wheeler-Kingshott, C.A., Hickman, S.J., et al., 2002. Investigating cervical spinal cord structure using axial diffusion tensor imaging. NeuroImage 16 (1), 93 – 102. Wheeler-Kingshott, C.A., Trip, A., Symms, M.R., Parker, G.J.M., Barker, G.J., Miller, D.H., 2004. DTI of the optic nerve. Proc. Int. Soc. Mag. Reson. Med. 12, 2471. Youl, B.D., Turano, G., Miller, D.H., et al., 1991. The pathophysiology of acute optic neuritis. An association of gadolinium leakage with clinical and electrophysiological deficits. Brain 114 (Pt. 6), 2437 – 2450.
Optic nerve diffusion tensor imaging in optic neuritis S. Anand Trip,a,b,* Claudia Wheeler-Kingshott,a Stephen J. Jones,c Wai-Yung Li,d Gareth J. Barker,e Alan J. Thompson,a Gordon T. Plant,b and David H. Miller a a
NMR Research Unit, Department of Neuroinflammation, Institute of Neurology, University College London, Queen Square, London WC1N 3BG, UK Department of Neuro-Ophthalmology, Moorfields Eye Hospital, City Road, London, UK c Department of Clinical Neurophysiology, National Hospital for Neurology and Neurosurgery, London, UK d Lysholm Department of Neuroradiology, The National Hospital for Neurology and Neurosurgery, Queen Square, London, UK e Centre for Neuroimaging Sciences, Department of Neurology, Institute of Psychiatry, Kings College London, London, UK b
Received 7 June 2005; revised 13 September 2005; accepted 19 September 2005 Available online 20 October 2005
Diffusion tensor magnetic resonance imaging (DT-MRI) provides in vivo information about the pathology of multiple sclerosis lesions. Increases in mean diffusivity (MD) and reductions in fractional anisotropy (FA) have been found and may represent axonal disruption. The optic nerve is an ideal structure for study by DT-MRI but previous clinical studies did not obtain the full diffusion tensor necessary to calculate MD and FA. In this study, a technique that was specifically developed to achieve full diffusion tensor measurements from the optic nerve (zonal oblique multislice (ZOOM) echoplanar imaging) was applied to 25 patients with a single unilateral episode of optic neuritis at least one year previously, and 15 controls. The intraorbital nerves were segmented on non-diffusion-weighted images and the regions of interest transferred to MD, FA, and eigenvalue maps to obtain quantitative data. Quantitative visual testing and electrophysiology were also performed. In affected nerves, mean MD and mean orthogonal eigenvalue l– – were elevated, and mean FA reduced compared with clinically unaffected contralateral nerves (P < 0.001) and control nerves (P < 0.001). The mean principal eigenvalue l\\ was significantly increased in affected nerves compared to contralateral unaffected nerves (P == 0.04) but not compared to control nerves (P == 0.13). There was no association of clinical measures of visual function in affected eyes with the DT-MRI parameters but there was a significant correlation of the whole field visual evoked potential (VEP) amplitude with MD (r == 0.57, P == 0.006) and l– = – (r = 0.56, P == 0.007). These findings suggest that optic nerve DT-MRI measures provide an indication of the structural integrity of axons. D 2005 Elsevier Inc. All rights reserved.
* Corresponding author. NMR Research Unit, Department of Neuroinflammation, Institute of Neurology, University College London, Queen Square, London WC1N 3BG, UK. Fax: +44 0 20 7278 5616. E-mail address: [email protected] (S.A. Trip). Available online on ScienceDirect (www.sciencedirect.com). 1053-8119/$ - see front matter D 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2005.09.024
Introduction Diffusion is the random motion of molecules in any fluid system including biological tissue. Diffusion-weighted magnetic resonance imaging (DW-MRI) refers to the process of making magnetic resonance imaging (MRI) sensitive to the molecular motion of water molecules, allowing the Apparent Diffusion Coefficient (ADC) to be measured in one or more directions (Wheeler-Kingshott et al., 2004). This is potentially a useful technique for studying white matter structure, as white matter tracts in the central nervous system consist of bundles of axons usually orientated in the same direction. Water diffusion can occur in any direction but occurs preferentially parallel to the orientation of axons (Takahashi et al., 2000) because their cell membranes and other oriented micro-structures act as barriers to diffusion (Von Meerwall and Fergusson, 1981). Such diffusion is said to be anisotropic and is dependent on the structural integrity of white matter tracts. Diffusion tensor magnetic resonance imaging (DTMRI) is an extension of DW-MRI in which 6 or more measurements probe diffusion in different directions and is one method by which the diffusion tensor (DT) – a complete description of diffusion in three dimensions – is calculated. Any disruption to white matter tracts or change in axonal membrane permeability would be expected to change DT indices (Horsfield and Jones, 2002; Ford et al., 1998), and in particular to lead to an increase in the mean diffusivity (MD), a measure of average molecular motion, and also to a decrease in fractional anisotropy (FA), a measure of the preponderance of diffusion direction (Basser et al., 1994). Multiple sclerosis (MS) is a common central nervous system (CNS) disease characterised pathologically by development of multifocal inflammatory demyelinating white matter lesions. It often results in major neurological disability and current disease modifying treatments have limited efficacy. The pathophysiological mechanisms of disease evolution are only partly understood and the use of MR techniques to study the evolution of lesions in
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505
vivo is an important tool for obtaining better insights in to such mechanisms. DW-MRI and DT-MRI have supplemented standard MR techniques for studying pathology in vivo. Studies in MS have shown increased ADC and MD with decreased FA in chronic T1hypointense lesions compared to T1-isointense lesions which is compatible with evidence that T1-hypointense lesions represent more extensive tissue loss (Barkhof et al., 2000; Filippi and Inglese, 2001). FA is lower in acute, gadolinium-enhancing lesions compared to non-enhancing lesions, probably because extracellular oedema alters the anisotropic pattern of diffusion (Werring et al., 1999; Filippi and Inglese, 2001). ADC and MD values are elevated, but the extent may depend upon the lesion age (Werring et al., 1999; Roychowdhury et al., 2000). Optic neuritis is a common manifestation in MS, and its study offers special opportunities to explore the pathophysiology of individual inflammatory demyelinating CNS lesions. This is because the optic nerve is a discrete anatomical structure, the function of which can be more reliably assessed by quantitative clinical and electrophysiological measures than is possible for other CNS pathways, e.g. motor function. The lesion of optic neuritis can also be clearly identified with MRI provided that fatsaturation methods are used (Gass et al., 1996). As the optic nerve is almost a pure white matter tract, it should lend itself to study with DT-MRI. There are, however, particular challenges associated with in vivo imaging of the optic nerve with MRI (Barker, 2000), and especially with DW-MRI/DT-MRI (Barker, 2001). The optic nerves are small, mobile structures surrounded by CSF and orbital fat. High resolution CSF and fat-saturated techniques are desirable particularly as these minimise the effects of partial volume in quantitative imaging. DW-MRI is sensitised to microscopic motion of water molecules, but this raises additional problems as it is extremely sensitive to macroscopic motion to which the optic nerves are susceptible due to eye movement. Furthermore, the high resolution needed to image small structures such as the optic nerve results in low MR signal-to-noise ratio (SNR) which needs special attention during image processing and analysis. Until recently, attempts at DW-MRI of the optic nerve have been restricted to measuring the ADC along a limited number of diffusion directions (Iwasawa et al., 1997; Freeman et al., 1998; Wheeler-Kingshott et al., 2002a; Hickman et al., 2005) and have therefore not obtained the full DT which can provide rotationally invariant indices including MD, FA, and diffusion eigenvalues. A CSF- and fatsuppressed zonal oblique multislice echoplanar imaging (ZOOMEPI) sequence (Wheeler-Kingshott et al., 2002a) has recently been developed further in the optic nerve by increasing the number of
499
diffusion directions from three to six, thus allowing full DT analysis and the calculation of MD and FA in healthy volunteers (Wheeler-Kingshott et al., 2004). In the current study, we present the results of DT-MRI in a cohort of patients who had a previous episode of unilateral optic neuritis and compare the findings with those obtained in healthy controls. The aims of the study were to investigate if the rotationally invariant indices produced by the DT were different in nerves affected by optic neuritis, and whether any changes were related to measures of visual function and/or electrophysiological markers of optic nerve function.
Methods Subjects Twenty-five patients with a single previous attack of acute unilateral optic neuritis and no recurrence were recruited from the case records of the Neuro-ophthalmology clinic, Moorfields Eye Hospital, London. In all patients, appropriate investigations had been made to exclude an alternative aetiology to the optic neuropathy. In order to study a range of visual deficits, there was a deliberate selection bias towards those with incomplete visual recovery. Patient demographic data are summarised in Table 1. Patients with a clinical diagnosis of MS (Poser et al., 1983) were included in the study if they had only one attack of unilateral optic neuritis. Both the eye affected by optic neuritis and the clinically unaffected eye were investigated in all patients. Fifteen controls subjects were also studied, none of whom were known to have any ophthalmological or neurological disorder. The mean age was 36 years (range 30 – 56 years) with 9 women and 6 men. One eye from each control subject was randomly chosen for study. Ethical approval for the study was obtained from the joint ethics committee of the Institute of Neurology and the National Hospital for Neurology and Neurosurgery and the Moorfields Eye Hospital Research Governance Committee. Informed consent in writing was obtained from all subjects, in accordance with the Declaration of Helsinki. DT-MRI Images were acquired with a 1.5 T GE Signa scanner (General Electric, Milwaukee, WI) with maximum gradient strength of 22 mTm 1. A quadrature birdcage coil was used as both transmitter
Table 1 Demographic data and visual function of patients Male:Female MS:CIS Age (years) Time since onset of optic neuritis (years) LogMAR visual acuity Visual field mean deviation (dB) Colour vision (FM 100-Hue score)
Affected Unaffected Affected Unaffected Affected Unaffected
Mean
SD
Range
Ratio
– – 40.6 3.1 +0.17 0.07 6.6 2.9 19.2 11.1
– – 9.4 1.7 0.32 0.09 4.7 3.8 7.9 2.6
– – 22.4 – 57.0 1.0 – 6.7 0.08 – +1.40 0.20 – +0.20 18.5 – 0.9 14.8 – +0.68 10.2 – 36.6 7.5 – 16.1
12:13 10:15 – – – – – – – –
CIS indicates clinically isolated syndrome; MAR, minimum angle of resolution; FM, Farnsworth – Munsell.
500
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505
and receiver. Subjects were asked to close their eyes and avoid deliberate eye movements during scanning. Coronal optic nerve images were obtained using a single-shot ZOOM-EPI DT-MRI sequence (Wheeler-Kingshott et al., 2004) with the following parameters: field of view 80 40 mm (with only the inner volume, IV = 26 mm, fully excited); matrix size 64 32; pixel size 1.25 1.25 mm. To increase the efficiency of the sequence, diffusion gradients were applied at maximum strength along 6 pairs of orthogonal directions (1, 1, 0; 1, 0, 1; 0, 1, 1; 1, 1, 0; 1, 0, 1; 0, 1, 1). The diffusion parameters were d, D = 20, 27 ms; b = 600 s _ mm 2. The echo time TE was 83.6 ms. One non-DW (b 0) set of images was acquired. To reduce the partial volume effects due to CSF and fat signal, an inversion pulse was used to null signal from CSF (TI = 1400 ms, TR = 3400 ms) and a frequency-selective spectral – spatial pulse (exciting only water) was used for excitation. ZOOM-EPI requires a spatial gap between slices acquired consecutively in time, therefore we chose to image 3 slices with a thickness of 4 mm and slice gap of 8 mm. The motion of the ON is ‘‘frozen’’ during each single-shot acquisition of one image, but motion between successive acquisitions is extremely difficult to control; averaging over many images was found to be the best way to allow for this, resulting in high quality images of the mean position of the nerve (Wheeler-Kingshott et al., 2000). Averaging was also necessary because of the low SNR which characterises the images (SNR of DW < 5:1). For our protocol, 44 averages were collected for off-line averaging (see below). The total acquisition time for the DT-MRI (with 3 dummy cycles) was 17 min 40 s for a total of 924 images. The averaging process to obtain the final b 0 and 6 DW images was performed on the magnitude signal. The low SNR of the magnitude images meant that a bias could be introduced in the DT calculations. An extra scan of just noise (total scan time = 3 min) was acquired and used during post-processing to perform Rayleigh
noise correction of the signal intensity, before DT calculations (Miller and Joseph, 1993; Wheeler-Kingshott et al., 2002a). A nonlinear smoothing algorithm was applied to each set of images to further reduce noise while preserving structure (Parker et al., 2000). Review of the acquired DT images revealed that the most anterior of the 3 slices was contaminated by the posterior aspect of the globe in 52% of nerves affected by optic neuritis, in 60% of clinically unaffected patient nerves, and in 47% of control nerves. Because the optic nerve could not be clearly segmented in so many of these slices, the anterior slice could not be used for analysis. In the posterior intracanalicular/intracranial slice, the optic nerve could not be adequately identified in 28% of nerves affected by optic neuritis, in 32% of clinically unaffected patient nerves, and in 33% of control nerves. Again, this slice was not used for analysis. The intraorbital nerve was clearly seen in the middle slice of all subjects and was used for analysis. The DT images were processed and analysed on Unix workstations (Sun Microsystems, Mountain View, CA, USA). MD, FA, and eigenvector maps were calculated on a pixel by pixel basis. Images were displayed and analysed using the DispImage display tool (Plummer, 1992). To avoid bias, each optic nerve was manually segmented by a single blinded observer from the nondiffusion-weighted (b 0) image. A square 2 2 voxel region of interest (ROI) was placed over the centre of each optic nerve on the mean b 0 image averaged over all 44 acquisitions, using maximum signal intensity and minimum standard deviation as guidance towards the optimum position. These ROIs were then automatically transferred to the Rayleigh noise corrected MD, FA, and eigenvector maps (Fig. 1) where mean values from the 4 voxels were obtained. Three eigenvalues were produced from each ROI: the largest was the principal eigenvalue (k \\) representing the diffusion coefficient along the principal direction of diffusion parallel to
Fig. 1. Intraorbital slice from a 36-year-old female patient 2 years after left sided optic neuritis. (A) Non-diffusion weighted b 0 image, (B) MD map, (C) FA map, (D) principal eigenvalue k \\ map, and (E) one of two orthogonal eigenvalue k – maps. The arrow indicates the affected nerve.
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505
the nerve; the other two were the diffusion coefficients orthogonal to the nerve. The average of the two orthogonal eigenvalues was expressed as k –. To avoid the introduction of bias by partial volume effect, careful note was taken of the signal characteristics of each voxel within the ROI when transferred to the DT maps. Any ROI which contained one or more ‘‘empty’’ voxels, where the DT-MRI calculation failed (usually due to low SNR), was excluded. This led to the exclusion of 3 affected and 2 clinically unaffected patient nerves and 1 control nerve. To assess measurement reproducibility, measurements were repeated in 10 patients and 10 controls, who were all randomly selected and again analysed in a blinded fashion. Lesion identification using fast spin echo (FSE) Patients also had optic nerve imaging with a dual echo FSE sequence (coronal-oblique, TR 2300 ms, TEeffective 58/145 ms, echo-train length 8, 2 excitations, matrix size 512 384, field of view 24 18 cm, in-plane resolution 0.5 0.5 mm, 16 3 mm interleaved contiguous slices, 11 min acquisition time). A neuroradiologist (WYL), blinded to the side of clinical involvement and severity of visual impairment, identified and measured the length of the lesion (the number of slices affected 3 mm) on the FSE images.
501
horizontally by 20- vertically, using corneal surface electrodes (DTL Plus, Retina Technologies) referred to skin-surface electrodes over the ipsilateral outer canthus. 40V check sizes were used, reversing 4.3 times per second. The luminance of the bright squares was 60 cd/m2 and of the dark squares was 4 cd/m2. The amplifier corner frequencies were 1 and 1000 Hz. The sampling rate was 3 samples/ms and the sweep duration was 170 ms. We made 3 averages of 200 responses and subsequently averaged them together. Sweeps containing artefacts of more than 165AV were automatically rejected. Analysis was performed blind to the status of each subject. Statistical analysis Statistical analyses were performed using SPSS 11.5 for Windows (SPSS, Chicago, IL, USA). In order to account for potential age and gender effects, differences between patients and controls were assessed using multiple linear regression models, with the MRI parameter as the response variable, disease status, and gender as categorical variables, and age as a continuous covariate. Differences between affected eye and unaffected eye within patients were explored using the Wilcoxon signed ranks test. Spearman rank correlations were calculated between MRI parameters, visual function, and electrophysiological variables. Data are presented uncorrected for multiple comparisons and therefore appropriate allowance should be made by the reader when interpreting the data.
Clinical testing of visual function Results Visual acuity of patients with appropriate refraction was measured using a retro-illuminated ETDRS chart and recorded as the 4 m logMAR acuity (Ferris et al., 1982). The central 30- of the visual field was analysed using the 30 – 2 program on the Humphrey field analyser (Allergan-Humphrey Inc., San Leandro, CA, USA). Wide-angle lenses were used to correct refractive errors where necessary. The overall visual field mean deviation was compared with a reference field derived from control data provided by the manufacturer. One patient, who had near normal acuity and normal visual fields by clinical examination, was unable to reliably complete automated perimetry due to high levels of fixation losses and false negative responses. Colour vision was assessed using the Farnsworth – Munsell 100-Hue test (Farnsworth, 1943) and scored as square root of the error score (FM 100-Hue score) because this follows a normal distribution. Two patients with a congenital anomaly of colour vision were excluded from this test. These three visual function parameters were used because they give continuously variable measures that are amenable to statistical analysis. Electrophysiological assessment of optic nerve function Visual evoked potential (VEP) amplitudes and latencies were recorded to monocular stimuli using skin-surface EEG electrodes attached over the occiput, 5 cm above the inion, referred to a frontal electrode at Fz (10 – 20 System). The stimuli comprised reversal of a checkerboard pattern in the ‘‘whole field’’ and in the ‘‘central field’’ as previously described (Brusa et al., 2001). Central field responses were unobtainable in one patient and one control. The pattern electroretinogram (PERG) N95 component is thought to be of retinal ganglion cell origin and can thus be reduced in optic neuropathies including optic neuritis by a process of retrograde degeneration (Holder, 1991). The PERG was recorded to binocular stimulation of the ‘‘whole field’’, subtending 28-
Clinical data of patients Table 1 summarises patient demographics including range of visual function. Electrophysiology data Table 2 presents the electrophysiology data. In affected eyes, there was no correlation of visual acuity with electrophysiological measures but the visual field mean deviation correlated with the whole field (r = 0.46, P = 0.03) and central field (r = 0.46, P = 0.03) VEP amplitudes, and colour vision (FM 100-Hue score) was correlated with whole field VEP (r = 0.49, P = 0.02), central field VEP (r = 0.69, P < 0.001), and PERG N95 (r = 0.47, P = 0.02) amplitudes. No correlations were present with VEP latencies. Measurement reproducibility for DT-MRI As the ROI size was fixed at 2 2 voxels and the placement was guided by quantitative measures, the repeat measurements for all DT-MRI parameters were exactly the same. Group differences in DT-MRI data Table 3 summarises all the group DT-MRI data. The mean MD from nerves affected by optic neuritis was elevated at 1611 [SD 290] 10 6 mm2s 1, compared with 1191 [SD 198] 10 6 mm2s 1 from clinically unaffected contralateral nerves ( P < 0.001 versus affected nerves) and 1083 [SD 168] 10 6 mm2s 1 from control nerves ( P < 0.001 versus affected nerves). There was no significant difference between control nerves and patients’ unaffected contralateral nerves ( P = 0.29).
502
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505
Table 2 Electrophysiology group data Control nerves (n = 15)
Unaffected patient nerves (n = 25)
Affected patient nerves (n = 25) 5.6 [3.4] P c = 0.003* P f = 0.001* 2.9 [1.8] P c = 0.002* P f = 0.004* 110.9 [18.3] P c = 0.02* P f = 0.21 115.9 [23.2] P c = 0.001* P f = 0.01* 5.2 [2.1] P c = 0.47 P f = 0.001* 98.6 [14.8] P c = 0.06 P f = 0.003*
Whole field VEP amplitude, AV
9.9 [5.4]
8.9 [4.8] P c = 0.78
Central field VEP amplitudea, AV
5.9 [3.3]
5.3 [3.5] P c = 0.84
Whole field VEP latency, ms
97.0 [5.8]
106.6 [17.8] P c = 0.06
Central field VEP latencya, ms
91.8 [6.9]
98.7 [20.0] P c = 0.26
6.0 [2.7]
6.7 [2.2] P c = 0.13
87.9 [9.0]
89.3 [7.7] P c = 0.99
PERG N95 amplitude, AV PERG N95 latency, ms
There were no significant differences in any of the DT parameters between the 15 patients with clinically isolated optic neuritis and the 10 patients with MS. Correlation of MR-DTI data with visual function and electrophysiology There was no association of the clinical measures of visual function with any of the DT-MRI parameters. However, there was a significant correlation of the whole field VEP amplitude with MD in both affected eyes (r = 0.57, P = 0.006) and unaffected contralateral eyes (r = 0.43, P = 0.04), and also with k – in affected eyes (r = 0.56, P = 0.007). Central field VEP amplitude in unaffected contralateral eyes correlated with MD (r = 0.44, P = 0.04) and k – (r = 0.54, P = 0.008) whereas there were only trends in affected eyes. The PERG N95 amplitude showed non-significant trends for correlation with (i) affected eye MD (r = 0.42, P = 0.06) and (ii) unaffected contralateral eye MD (r = 0.35, P = 0.10) as well as a significant correlation with unaffected contralateral eye k – (r = 0.45, P = 0.03). There were also non-significant trends for correlation of affected eye FA with whole field VEP amplitude (r = 0.37, P = 0.09) and unaffected contralateral eye FA with central field VEP amplitude (r = 0.39, P = 0.06) (Table 4). No association was found with any of the DT-MRI measures and either VEP or PERG latencies.
Values are mean [SD]. P c values are comparisons between control nerves and patient nerves. P f values are comparisons between unaffected contralateral nerves and affected nerves in patients. a Controls n = 14, patients n = 24. * P < 0.05.
Lesion length on FSE A single lesion was identified in all of the 25 affected optic nerves. Mean lesion length was 26 mm [SD 7] with a range of 12 – 39 mm. The FSE imaging coordinates were used to determine if the 4 mm intraorbital slice contained lesioned nerve. In 22 of the 25 patients, the DT-MRI slices contained lesion throughout, in 2 it partly contained lesion, and in 1 there was no lesion within the slice (in this patient, there was a short canalicular lesion visible on FSE). Correlations with visual function and electrophysiology did not change significantly when the part-lesion and lesion-free cases were excluded from analysis. There was no significant correlation between lesion length and DT-MRI parameters but there was a borderline association with k – (r = 0.40, P = 0.07).
Mean FA from affected nerves was lower at 0.435 [SD 0.100], compared with 0.602 [SD 0.100] from clinically unaffected contralateral nerves (P < 0.001) and 0.669 [SD 0.093] from control nerves (P < 0.001). There was a non-significant trend for FA to be lower in patients’ unaffected contralateral nerves versus control nerves (P = 0.07). The mean orthogonal eigenvalue k – from affected nerves was raised at 1217 [SD 326] 10 6 mm2s 1, compared with 730 [SD 196] 10 6 mm2s 1 from clinically unaffected contralateral nerves (P < 0.001) and 629 [SD 199] 10 6 mm2s 1 from control nerves (P < 0.001). There was no significant difference between control nerves and patients’ unaffected contralateral nerves (P = 0.26). The mean principal eigenvalue k \\ was marginally higher in affected nerves at 2355 [SD 509] 10 6 mm2s 1 compared to 2117 [SD 336] 10 6 mm2s 1 in unaffected contralateral nerves (P = 0.04) and 2088 [SD 309] 10 6 mm2s 1 in control nerves (P = 0.13). There was no significant difference between control nerves and patients’ unaffected contralateral nerves (P = 0.80).
Discussion Using a ZOOM-EPI technique that was specifically developed for investigating the optic nerve, it was possible to measure the full DT in the optic nerves of controls and patients with previous
Table 3 DT-MRI parameter group data MD (10 Control nerves (n = 14) Patient unaffected nerves (n = 23) Patient affected nerves (n = 22)
6
1083 [168] 1191 [198] P c = 0.29 1611 [290] P c < 0.001* P f < 0.001*
mm2s 1)
6
FA
k \\ (10
0.669 [0.093] 0.602 [0.100] P c = 0.07 0.435 [0.100] P c < 0.001* P f < 0.001*
2088 [309] 2117 [336] P c = 0.80 2355 [509] P c = 0.13 P f = 0.04*
mm2s 1)
k – (10
6
mm2s 1)
629 [199] 730 [196] P c = 0.26 1217 [326] P c < 0.001* P f < 0.001*
Values are mean [SD]. P c values are comparisons between control nerves and patient nerves. P f values are comparisons between unaffected contralateral nerves and affected nerves in patients. * P < 0.05.
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505
503
Table 4 Correlation of DT-MRI measures with electrophysiology amplitude data in patients MD
Whole field VEP amplitude Central field VEP amplitude PERG N95 amplitude
k \\
FA
k–
Affected
Unaffected
Affected
Unaffected
Affected
Unaffected
Affected
Unaffected
r= P= r= P= r= P=
r= P= r= P= r= P=
r = 0.37 P = 0.09 r = 0.28 P = 0.23 r = 0.30 P = 0.17
r = 0.01 P = 0.96 r = 0.39 P = 0.06 r = 0.13 P = 0.57
r= P= r= P= r= P=
r= P= r= P= r= P=
r= P= r= P= r= P=
r= P= r= P= r= P=
0.57 0.006* 0.37 0.10 0.42 0.06
0.43 0.04* 0.44 0.04* 0.35 0.10
0.14 0.54 0.06 0.80 0.26 0.24
0.33 0.13 0.06 0.80 0.07 0.75
0.56 0.007* 0.42 0.06 0.24 0.28
0.17 0.43 0.54 0.008* 0.45 0.03*
r values are Spearman Rank Correlation Coefficients. * P < 0.05.
unilateral optic neuritis in clinically acceptable scan times, with fatand CSF-suppression to minimise partial volume effects. This was achieved by increasing the number of diffusion directions from three (sufficient for ADC measurement only) to six, and by using a high number of averages for each diffusion weighting to compensate for motion of the optic nerves. The DT produced rotationally invariant indices: MD, FA, and diffusion eigenvalues which provided useful in vivo information about the changes that occur after optic neuritis. Although it was hoped to obtain quantitative data from three 4 mm slices, contamination of the optic nerve by the globe resulted from the anterior slice being prescribed too anteriorly in some subjects, and the posterior slice did not allow reliable segmentation of the optic nerve. The intraorbital slice (the middle of the 3) did, however, allow reliable segmentation of the optic nerve on non-diffusion weighted b 0 images in all subjects. Special care was taken to exclude a small number of nerves in which the ROI did not completely contain sufficient optic nerve signal when transferred to the DT-MRI maps, to ensure that partial volume effects were kept to a minimum. Previous studies of optic nerve DW-MRI have not reported the full DT, or did not include methods to suppress fat and CSF. Iwasawa et al. (1997) used a spin-echo sequence with low diffusion weighting and cardiac gating to study a single intraorbital optic nerve slice. Three diffusion directions were acquired but due to motion artefact, ADC could only be measured in the y and z directions and not always reliably. A small 1 mm diameter ROI was used. The mean ADC (4180 10 6 mm2s 1, n = 9) in the optic nerves previously affected by optic neuritis was significantly higher than that in controls (1559 10 6 mm2s 1, n = 14) and that in the nerves with acute optic neuritis (941 10 6 mm2s 1, n = 4) ( P < 0.001). Hickman et al. (2005) used ZOOM-EPI to study the optic nerve ADC in 16 patients who were seen 1 year following an attack of optic neuritis, and in 10 controls. Four 4 mm slices covering the intraorbital optic nerve were used to determine mean ADC which was 1324 10 6 mm2s 1 in affected nerves, compared to 990 10 6 mm2s 1 in unaffected contralateral nerves ( P = 0.005) and 928 10 6 mm2s 1 in control nerves ( P = 0.006). The affected nerve ADC correlated with measures of visual function (visual acuity, visual field, and colour vision) and electrophysiological parameters (VEP amplitudes and latencies). The only study which attempted to measure FA as well as ADC was that by Chabert et al. (2002) who found – using a non CPMGFSE sequence – that the MD was 1670 10 6 mm2s 1 with a mean FA of 0.59 in four healthy volunteers. They found that the fibre directions followed the expected nerve fibre directions on anisotropy maps. However, the sequence was neither CSF- nor fat-
suppressed so the results may have been affected by partial volume effects. This may account for the higher MD and lower FA that was obtained compared to our 15 controls. The optic nerve MD and ADC values of controls in all of these studies and in our control nerves (using different sequences and different scanners) are consistently higher than those in deep brain white matter and are likely to reflect structural differences between the optic nerve and brain white matter tracts. The Raleigh noise correction compensates for the otherwise artefactually low values due to low SNR without correction rather than erroneously elevating diffusivity (Wheeler-Kingshott et al., 2002a). Further evidence supporting the latter conclusion comes from a study of normal cervical cord structure using ZOOM-EPI with the same scanner as the present study which found comparable cervical cord MD values (940 10 6 mm2s 1) to the optic nerve despite the absence of Raleigh noise correction (Wheeler-Kingshott et al., 2002b). The current study has found that, as expected, MD is significantly increased and FA is significantly reduced in optic nerves affected by optic neuritis compared to unaffected contralateral nerves and healthy control nerves. These findings are compatible with a process of axonal disruption or loss, although demyelination and gliosis might also contribute. The increase in MD and decrease in FA seen in what were predominantly chronic optic nerve lesions resemble those reported in studies of chronic brain lesions in MS (Werring et al., 1999; Bammer et al., 2000; Filippi and Inglese, 2001; Dong et al., 2004), although the absolute values for both MD and FA seen in the present study were somewhat greater than those reported in brain lesions. The higher FA may reflect the greater baseline anisotropy of the optic nerves compared to brain white matter tracts in the normal state and the higher MD could be due to the bias to poor recovery we had in our cohort. However, differences in the DT-MRI sequences used may also be a factor. The present study did not find an association of any DT parameter with measures of visual function. A previous study by Hickman et al. (2005) found such an association but differed in that their patients were all examined at the same 1 year time point following optic neuritis when one would expect a similar pattern of recovery processes taking place in each patient, whereas the present cohort was heterogeneous in terms of disease duration and therefore may have differing degrees of recovery processes. Perhaps more importantly, the present study surveyed only 4 mm of optic nerve compared to the 16 mm of nerve surveyed in the study of Hickman et al. (2005); the less comprehensive sampling of the lesions may have reduced the opportunity to detect clinically relevant DT changes. The VEP amplitude is believed to reflect the number of optic nerve fibres in functional continuity (Jones and Brusa, 2003) and a
504
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505
reduction in amplitude in optic neuritis could be due to two mechanisms. In the acute phase of optic neuritis associated with substantial visual impairment, there is a marked loss of VEP amplitude (Jones, 1993), often abolishing the response completely, which usually resolves over a period of a few weeks concurrent with the period when gadolinium enhancement on MRI disappears (Youl et al., 1991; Hickman et al., 2004). This suggests that the VEP amplitude reduction in the acute phase, characterised by blood – brain barrier leakage, inflammation, oedema, and demyelination, is caused by temporary conduction block in optic nerve axons. Conduction block from inflammation and oedema is unlikely to be the cause of the reduced VEP amplitude in the present cohort as they were all at least 1 year from onset of optic neuritis. The reduced VEP amplitude arising from the chronic lesions that we studied could reflect either: (i) conduction block from persistent demyelination or (ii) axonal loss. The latter will likely have resulted from acute inflammation-induced axonal transection (a well described finding in acute inflammatory lesion in multiple sclerosis; Trapp et al., 1998), although post-inflammatory mechanisms for axonal loss may also contribute (Compston and Coles, 2002). The significant correlation of VEP amplitude reduction with MD increase suggests that the increased diffusivity is of functional relevance and may reflect either axonal loss or conduction block in surviving axons. This association was found in the clinically affected eyes (whole field only) and also in the clinically unaffected eyes (whole and central fields). The latter finding is of interest as two previous follow-up studies have found nonsignificant marginal reductions in the VEP amplitude of unaffected fellow eyes over time which has been postulated to be due to an insidious process of axonal degeneration (Brusa et al., 1999, 2001). The findings in the asymptomatic nerves in the present study would support this hypothesis. The lack of correlation with central field VEP amplitude in affected eyes may be due to the lower SNR of this measure. As the PERG N95 response is believed to be of retinal ganglion cell origin and shows less consistent changes than the VEP in optic neuritis (Holder, 1991), it is not unexpected that there were no convincing correlations with DT-MRI, and the correlation demonstrated with the unaffected eye k – is more likely to be due to type 1 error. The mean of the orthogonal eigenvalues k –, which reflects diffusion orthogonal to the axis of the optic nerve, behaved in a similar way to MD in that it was significantly increased in affected nerves and correlated with whole field VEP amplitude to a similar extent. The principal eigenvalue k \\, however, was not significantly different between the three groups of optic nerves. A possible explanation for this is that although some axons are lost following optic neuritis, the remaining axons are still aligned in the same direction (along the optic nerve) thus maintaining a relatively normal k \\. The spaces between the remaining axons may allow increased diffusion of water molecules in the parallel direction which could counteract the effects of axonal disruption reducing anisotropy. At the same time, increased spaces between axons would be expected to contribute to an increase in k – and MD. Changes in FA values may occur secondary to changes in any or all of the three eigenvalues. In inflammatory conditions such as optic neuritis, all three eigenvalues could increase and therefore any percentage change in FA is likely to be less than the change in any given eigenvalue. If all three eigenvalues increased together to the same extent, then it would be possible for FA to be unchanged thus making a case that eigenvalues may be more sensitive than FA in
quantifying pathology. It may be of interest in future studies to look at both FA and eigenvalues. Future studies will need to concentrate on overcoming the methodological difficulties specific to DT-MRI of the optic nerve. Higher resolution imaging would help to decrease partial volume; improved SNR should result from increases in gradient strength, the use of multichannel receiver coils, and higher field scanners. Parallel imaging techniques were not widely available at the time this study was started. The main advantage of parallel imaging is that it is implemented by the manufacturer and therefore does not require special set up. It is not limited, like ZOOM-EPI, to noncontiguous slices/multiple acquisitions. The disadvantage is that it will reduce the already very low SNR still further and although this can, in theory, be made up by further averaging (albeit at the expense of longer scan times), it is not obvious what the effect of such low SNR will be on the algorithms used for the image reconstruction, nor how Rayleigh noise correction will interact with the parallel imaging reconstruction. More work would be needed to determine at what point the noise correction was best applied and to devise algorithms that can cope with position dependent sensitivity and ‘‘geometry factor’’. The challenge will be to produce a sequence with good coverage of the whole optic nerve with clinically acceptable scanning times; development of such methods could provide a powerful tool to study the effect of novel therapies that protect the optic nerve and its axons from damage in optic neuritis and other optic neuropathies.
Acknowledgments The NMR Research Unit is supported by the Multiple Sclerosis Society of Great Britain and Northern Ireland. We thank Dr. Simon Hickman for helpful comments.
References Bammer, R., Augustin, M., Strasser-Fuchs, S., et al., 2000. Magnetic resonance diffusion tensor imaging for characterizing diffuse and focal white matter abnormalities in multiple sclerosis. Magn. Reson. Med. 44, 583 – 591. Barker, G.J., 2000. Technical issues for the study of the optic nerve with MRI. J. Neurol. Sci. 172 (Suppl. 1), S13 – S16. Barker, G.J., 2001. Diffusion-weighted imaging of the spinal cord and optic nerve. J. Neurol. Sci. 186 (Suppl. 1), S45 – S49. Barkhof, F., Karas, G.B., van Walderveen, M.A., 2000. T1 hypointensities and axonal loss. Neuroimaging Clin. N Am. 10 (4), 739 – 752 (Nov). Basser, P.J., Mattiello, J., LeBihan, D., 1994. MR diffusion tensor spectroscopy and imaging. Biophys. J. 66, 259 – 267. Brusa, A., Jones, S.J., Kapoor, R., Miller, D.H., Plant, G.T., 1999. Long-term recovery and fellow eye deterioration after optic neuritis, determined by serial visual evoked potentials. J. Neurol. 246 (9), 776 – 782 (Sep). Brusa, A., Jones, S.J., Plant, G.T., 2001. Long-term remyelination after optic neuritis: a 2-year visual evoked potential and psychophysical serial study. Brain 124, 468 – 479. Chabert, S., Molko, N., Cointepas, Y., Le Roux, P., Le Bihan, D., 2002. Functional diffusion tensor imaging of the optic nerve using a non CPMG fast spin echo sequence. Proc. Intl. Soc. Mag. Reson. Med., 1115. Compston, A., Coles, A., 2002. Multiple sclerosis. Lancet 359, 1221 – 1231.
S.A. Trip et al. / NeuroImage 30 (2006) 498 – 505 Dong, Q., Welsh, R.C., Chenevert, T.L., et al., 2004. Clinical applications of diffusion tensor imaging. J. Magn. Reson. Imaging 19, 6 – 18. Farnsworth, D., 1943. The Farnsworth – Munsell 100-hue and dichotomous tests for color vision. J. Opt. Soc. Am. 33, 568 – 578. Ferris, F.L. III, Kassoff, A., Bresnick, G.H., Bailey, I., 1982. New visual acuity charts for clinical research. Am. J. Ophthalmol. 94, 91 – 96. Filippi, M., Inglese, M., 2001. Overview of diffusion-weighted magnetic resonance studies in multiple sclerosis. J. Neurol. Sci. 186 (Suppl. 1), S37 – S43. Ford, J.C., Hackney, D.B., Lavi, E., Phillips, M., Patel, U., 1998. Dependence of apparent diffusion coefficients on axonal spacing, membrane permeability, and diffusion time in spinal cord white matter. J. Magn. Reson. Imaging 8 (4), 775 – 782 (Jul – Aug). Freeman, A.J., Ballinger, R., Werner, M., Guy, J., 1998. Diffusion weighted EPI of the human optic nerve at 3 T. Proc. Int. Soc. Mag. Reson. Med. 6, 1265. Gass, A., Moseley, I.F., Barker, G.J., et al., 1996. Lesion discrimination in optic neuritis using high-resolution fat-suppressed fast spin-echo MRI. Neuroradiology 38, 317 – 321. Hickman, S.J., Toosy, A.T., Miszkiel, K.A., et al., 2004. Visual recovery following acute optic neuritis—A clinical, electrophysiological and magnetic resonance imaging study. J. Neurol. 251, 996 – 1005. Hickman, S.J., Wheeler-Kingshott, C.A., Jones, S.J., Miszkiel, K.A., Barker, G.J., Plant, G.T., Miller, D.H., 2005. Optic nerve diffusion measurement from diffusion weighted imaging in optic neuritis. AJNR Am. J. Neuroradiol. 26, 951 – 956. Holder, G.E., 1991. The incidence of abnormal pattern electroretinography in optic nerve demyelination. Electroencephalogr. Clin. Neurophysiol. 78, 18 – 26. Horsfield, M.A., Jones, D.K., 2002. Applications of diffusion-weighted and diffusion tensor MRI to white matter diseases—A review. NMR Biomed. 15 (78), 570 – 577. Iwasawa, T., Matoba, H., Ogi, A., et al., 1997. Diffusion-weighted imaging of the human optic nerve: a new approach to evaluate optic neuritis in multiple sclerosis. Magn. Reson. Med. 38, 484 – 491. Jones, S.J., 1993. Visual evoked potentials after optic neuritis. Effect of time interval, age and disease dissemination. J. Neurol. 240, 489 – 494. Jones, S.J., Brusa, A., 2003. Neurophysiological evidence for long-term repair of MS lesions: implications for axon protection. J. Neurol. Sci. 206, 193 – 198. Miller, A.J., Joseph, P.M., 1993. The use of power images to perform quantitative analysis on low SNR MR images. Magn. Reson. Imaging 11, 1051 – 1056.
505
Parker, G.J., Schnabel, J.A., Symms, M.R., Werring, D.J., Barker, G.J., 2000. Nonlinear smoothing for reduction of systematic and random errors in diffusion tensor imaging. J. Magn. Reson. Imaging 11, 702 – 710. Plummer, D.L., 1992. DispImage: a display and analysis tool for medical images. Rivista di Neuroradiologia 5, 489 – 495. Poser, C.M., Paty, D.W., Scheinberg, L., et al., 1983. New diagnostic criteria for multiple sclerosis: guidelines for research protocols. Ann. Neurol. 13, 227 – 231. Roychowdhury, S., Maldjian, J.A., Grossman, R.I., 2000. Multiple sclerosis: comparison of trace apparent diffusion coefficients with MR enhancement pattern of lesions. AJNR Am. J. Neuroradiol. 21, 869 – 874. Takahashi, M., Ono, J., Harada, K., Maeda, M., Hackney, D.B., 2000. Diffusional anisotropy in cranial nerves with maturation: quantitative evaluation with diffusion MR imaging in rats. Radiology 216 (3), 881 – 885 (Sep). Trapp, B.D., Peterson, J., Ransohoff, R.M., et al., 1998. Axonal transection in the lesions of multiple sclerosis. N. Engl. J. Med. 338, 278 – 285. Von Meerwall, E., Fergusson, R.D., 1981. Interpreting pulsed-gradient spinecho diffusion experiments with permeable membranes. J. Chem. Phys. 74, 6956 – 6959. Werring, D.J., Clark, C.A., Barker, G.J., Thompson, A.J., Miller, D.H., 1999. Diffusion tensor imaging of lesions and normal-appearing white matter in multiple sclerosis. Neurology 52, 1626 – 1632. Wheeler-Kingshott, C.A.M., Ciccarelli, O., Parker, G.J.M., Barker, G.J., 2000. Quantification of the motion of the optic nerve. Proc. ESMRMB 17th Annual, p. 443. Wheeler-Kingshott, C.A., Parker, G.J., Symms, M.R., et al., 2002. ADC mapping of the human optic nerve: increased resolution, coverage, and reliability with CSF-suppressed ZOOM-EPI. Magn. Reson. Med. 47, 24 – 31. Wheeler-Kingshott, C.A., Hickman, S.J., et al., 2002. Investigating cervical spinal cord structure using axial diffusion tensor imaging. NeuroImage 16 (1), 93 – 102. Wheeler-Kingshott, C.A., Trip, A., Symms, M.R., Parker, G.J.M., Barker, G.J., Miller, D.H., 2004. DTI of the optic nerve. Proc. Int. Soc. Mag. Reson. Med. 12, 2471. Youl, B.D., Turano, G., Miller, D.H., et al., 1991. The pathophysiology of acute optic neuritis. An association of gadolinium leakage with clinical and electrophysiological deficits. Brain 114 (Pt. 6), 2437 – 2450.