Diffusion tensor MRI of early upper motor neuron involvement in amyotrophic lateral sclerosis

Abstract

Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative system disorder affecting both upper and lower motor neurons. Despite supportive electrophysiological investigations, the involvement of the upper motor neuron is often difficult to assess at an early stage of disease. Diffusion tensor MRI provides an estimate of the orientation of fibre bundles in white matter on the basis of the diffusion characteristics of water. Diffusivity is generally higher in directions along fibre tracts than perpendicular to them. This degree of directionality of diffusion can be measured as fractional anisotropy. Changes in tissue structure due to degeneration of the corticospinal fibres can lead to a modification of the degree of directionality which can be detected by diffusion tensor MRI. We investigated 15 patients with ALS, six of whom had no clinical signs of upper motor neuron involvement at the time of MRI investigation, but developed pyramidal tract symptoms later in the course of their disease. These patients met the El Escorial criteria as their disease progressed. We found a decrease in fractional anisotropy in the corticospinal tract, corpus callosum and thalamus in all 15 ALS patients, including the patients without clinical signs of upper motor neuron lesion, compared with healthy controls. Regression analysis showed a negative correlation between fractional anisotropy and central motor conduction time obtained by transcranial magnetic stimulation, allowing spatial differentiation between the degenerated corticospinal tract fibres that supply the upper and lower extremities. Thus, diffusion tensor MRI can be used to assess upper motor neuron involvement in ALS patients before clinical symptoms of corticospinal tract lesion become apparent, and it may therefore contribute to earlier diagnosis of motor neuron disease.

diffusion tensor MRI; amyotrophic lateral sclerosis; upper motor neuron; corticospinal tract; transcranial magnetic stimulation

ALS = amyotrophic lateral sclerosis; CMCT = central motor conduction time; DTI = diffusion tensor imaging; EPI = echoplanar imaging; LMN = lower motor neuron; MEP = motor evoked potential; MNI = Montreal Neurological Institute; ROI = region of interest; STEAM = stimulated echo acquisition mode; TMS = transcranial magnetic stimulation; SNR = signal‐to‐noise ratio; UMN = upper motor neuron

Introduction

The diagnosis of amyotrophic lateral sclerosis (ALS), a progressive neurodegenerative system disorder affecting both upper and lower motor neurons (LMN), is mainly based on clinical criteria with possible symptoms such as progressive weakness of voluntary muscles, muscular atrophy, spasticity, tendon jerks and Babinski signs. Upper motor neuron (UMN) pathology often starts in the primary motor and premotor cortex, with secondary degeneration of motor fibres and gliosis along the corticospinal tract (Davidoff, 1990). The lesions of the LMN include the brainstem and spinal cord. Diagnosis at an early stage of the disease is desirable because of the poor prognosis of ALS, with an average life expectancy of 25 months after onset of symptoms, and the necessity of excluding other curable diseases. It is often difficult to decide whether the UMN is involved. This is because of only discrete UMN signs in clinical investigation or severe simultaneous lesions of the LMN. In contrast, LMN involvement can be detected subclinically using supportive techniques, such as needle electromyography. Both neurophysiological and neuroimaging techniques have been used to evaluate UMN pathology, but there are currently no sensitive techniques available in clinical practice for objectively assessing UMN damage at an early stage of the disease.

Transcranial magnetic stimulation (TMS) measurements may contribute to the diagnosis of ALS by revealing a clinically undetectable UMN dysfunction. However, the diagnostic sensitivity in unravelling UMN lesions, in particular those of limb muscles compared with cranial muscles, is rather low (Urban et al., 2001), and published results are inconsistent (Eisen et al., 1990; Berardelli et al., 1991; Claus et al., 1995). The triple stimulation technique can increase the sensitivity to UMN lesions but it is not yet being used routinely as it is difficult to apply (Buhler et al., 2001).

In neuroimaging, MRI‐FLAIR (fluid‐attenuated inversion recovery) images and T2‐ and proton density‐weighted MRIs may show increased signal intensity in the white matter (Hecht et al., 2001). However, these findings are rather unspecific and are not yet quantifiable (Karitzky and Ludolph, 2001). Proton magnetic resonance spectroscopy is useful for assessing UMN involvement in ALS (Ellis et al., 2001), but it is not sensitive enough to reveal early changes in the subcortical white matter in the motor region. This is due to considerable overlap between the ranges of metabolic peak area ratios from patients in an early stage of the disease and those from healthy control subjects (Ellis et al., 1998). Functional MRI and magnetization transfer are also promising methods. Further investigations are needed to elucidate their significance as diagnostic and prognostic tools in patients with ALS (Tanabe et al., 1998; Brooks et al., 2000a).

Diffusion tensor imaging (DTI) is a relatively new method in structural neuroimaging. It allows the estimation of the orientation of fibre bundles in white matter on the basis of the diffusion characteristics of water. Diffusivity is generally higher in directions along fibre tracts than perpendicular to them (Chenevert et al., 1990). This can be described mathematically by a tensor, which is characterized by its three eigenvectors and the corresponding eigenvalues. The eigenvector associated with the largest eigenvalue (the first eigenvector) indicates the predominant orientation of fibres in the given voxel. The directionality of diffusion can be quantified by the fractional anisotropy index, which is a rotationally invariant property of the diffusion tensor (Basser and Pierpaoli, 1996). Fractional anisotropy values range from 0 (no directional dependence of diffusion coefficients) to 1 (diffusion along a single direction). Changes in tissue structure (in this case degeneration of the corticospinal fibres) can lead to a modification of the degree of directionality, which can be detected by diffusion tensor MRI. Therefore, in degenerated white matter tracts of patients with ALS one would expect to find changes in the anisotropy of diffusion in comparison with healthy subjects. DTI has already produced promising results in assessing UMN pathology in patients with ALS (Ellis et al., 1999). The authors demonstrated that fractional anisotropy correlates with UMN involvement in ALS patients. For the analysis of diffusion characteristics, Ellis and colleagues used six predefined regions of interest (ROIs) along white matter tracts descending through the posterior limb of the internal capsule.

In our study, we were interested in white matter changes in the motor system at different levels of the brain caused by ALS. Thus, we calculated the fractional anisotropy voxelwise in a larger area not confined to the internal capsule. This area covered fibre bundles descending from the motor and premotor cortex to the brainstem. We calculated voxel‐based statistics on the fractional anisotropy maps to demonstrate fractional anisotropy differences in patients with ALS compared with healthy controls, using SPM99 (Friston et al., 1995). As the fractional anisotropy correlates with UMN lesions, we additionally investigated the correlation between fractional anisotropy and central motor conduction time (CMCT), which was derived from TMS. To explore the usefulness of DTI for assessing early UMN lesions, we studied a subgroup of six patients without clinical signs of UMN involvement at the time of MRI investigation. These patients developed clinical evidence of corticospinal tract degeneration in the course of their disease. In accordance with our anatomical a priori hypothesis regarding white matter changes in ALS along the corticospinal tract, we corrected for multiple comparisons using the small volume correction method implemented in SPM99.

Methods

Subjects

We studied a total of 15 patients with TMS, conventional MRI and DTI. Among these, nine had definite, probable or possible ALS and six displayed no clinical signs of UMN involvement according to the revised El Escorial criteria (Brooks et al., 2000b) at the time of investigation, but developed clinical pyramidal tract symptoms later in the course of their disease. Patients were compared with 12 healthy, age‐matched controls. Clinical details of the patients are provided in Table 1. As suggested by the World Federation of Neurology (Brooks et al., 2000b), we identified signs of UMN degeneration as any of the following: hyper‐reflexia with pathological spread of reflexes; clonic tendon reflexes; preserved reflexes in weak wasted limbs; spasticity; emotional lability; loss of superficial abdominal reflexes; and Babinski sign. Although brisk deep tendon reflexes in the absence of muscle weakness and wasting are not considered as a UMN sign according to the El Escorial criteria, this clinical feature is also listed in Table 1. LMN signs included muscular weakness, wasting and fasciculation. Time after onset of symptoms ranged from 6 to 24 months (mean 11.9 months, SD 5.6). Age ranged from 27 to 63 years (mean 52.2 years, SD 11.8). The healthy controls were free of neurological or other diseases and were age‐matched, with an age range from 28 to 63 years (mean 52.8 years, SD 10.9). Only subjects without contraindications for MRI were included. Written informed consent to participation in the study was obtained from all patients and healthy controls. The study was approved by the ethics committee of the University Hospital Eppendorf.

Transcranial magnetic stimulation

TMS was delivered by a Magstim magnetic stimulator 200 HP (Magstim Company, Dyfed, UK) using a 90‐mm circular coil. Initially, we determined the optimal coil position. This was defined as the position in which the largest motor evoked potential (MEP) was produced in the target muscle. We recorded EMG signals from the contralateral first dorsal interosseus and anterior tibial muscle via surface electrodes (Viking IV; Nicolet, Kleinostheim, Germany) and analysed them off‐line. Threshold was defined as the lowest stimulus intensity capable of inducing an MEP with an amplitude of at least 0.05 mV in five of 10 trials, the muscle being tested in the resting state. Then the stimulus intensity was increased to 20% above threshold, and stimuli were applied during slight voluntary contraction. We also obtained MEPs by stimulating the cervical and lumbar spinal roots to calculate the CMCT. F‐waves of the ulnar and peroneal nerve were measured to calculate peripheral motor conduction times. MEP amplitudes were expressed as percentages of the maximal M response amplitude (MEP amplitude ratio) and compared with values of healthy, age‐matched subjects (Kloten et al., 1992).

MRI protocol and data analysis

MRIs were acquired on a 1.5 Tesla MR system (Magnetom Vision, Siemens, Erlangen, Germany) with 25 mT/m maximum gradient strength. Head movement was limited by a vacuum fixation cushion. Structural T1‐weighted images with a resolution of 1 × 1 × 1 mm³ were acquired using three‐dimensional fast low angle shot (FLASH) imaging [flip angle 30°, repetition time (TR) 15 ms, echo time (TE) 5 ms, matrix 256 × 256 × 196]. For DTI, we used a single‐shot stimulated echo acquisition mode (STEAM) sequence (flip angle 15°, TR 8872 ms, TE 65 ms) (Nolte et al., 2000). The STEAM sequence is an MRI sequence which yields single‐shot images within a measuring time of ∼600 ms. We chose STEAM rather than echoplanar imaging (EPI) sequences, which are widely used for diffusion‐weighted imaging, because of the crucial advantages of STEAM for our study. EPI sequences characteristically display signal loss and geometric distortions caused by susceptibility gradients in orbitofrontal and inferior temporal brain areas (Johnson and Hutchison, 1985). In particular, in the vicinity of air–tissue interfaces (e.g. brainstem), signal alteration and distortions become a problem when using EPI in whole‐brain studies to cover regions from the cortex to the brainstem, as in our study (Nolte et al., 2000). At the level of the internal capsule, the ferrous basal ganglia also lead to signal loss, resulting in a decreased signal‐to‐noise ratio (SNR) when using EPI sequences. The susceptibility artefacts of EPI may be reduced by using segmented instead of single‐shot EPI, although the susceptibility to motion due to segmentation cannot be removed completely by correction methods (Atkinson et al., 2000). Additionally, eddy current effects arising from switching strong diffusion gradients (Nolte et al., 2000) contribute to image artefacts in EPI. These eddy currents depend on the direction and amplitude of the diffusion gradients and can be reduced only to a limited degree (Koch and Norris, 2000).

In comparison with EPI, the STEAM sequence exhibits a lower SNR. To increase the SNR in the tensor maps, data acquisition was repeated 20 times (scan time of 25 min for DTI). The number of averages was not increased further, so as not to impair patient compliance and comfort. We measured the SNR on the averaged images obtained with a b value of 0 s/mm² (where b = γ²G²δ² (Δ – δ/3), γ = gyromagnetic ratio; G = gradient amplitude; δ = gradient pulse duration; Δ = delay between the leading edges of the gradient pulses) in an ROI in the posterior limb of the internal capsule at the expected location of the pyramidal tract fibres. The SNR was defined as the ratio of the mean signal intensity in the internal capsule to the standard deviation of the signal in a background ROI outside the brain (Hunsche et al., 2001). This calculation provided an SNR of 96.5. Previous studies recommended an SNR of 20 in images obtained with a b value of 0 s/mm² for a reliable estimate of parameters derived from DTI (Basser et al., 1996; Hunsche et al., 2001).

Because of our interest in imaging the corticospinal tract from the primary motor and premotor cortex to the internal capsule and the brainstem, we chose STEAM rather than EPI because of its insensitivity to eddy currents and susceptibility gradients, especially in the vicinity of air–tissue interfaces, such as the brainstem (Nolte et al., 2000), to guarantee anatomically reliable diffusion tensor maps.

For diffusion‐weighted imaging, the matrix size was 56 × 64 and the field of view was 168 × 192 mm². The imaging volume of 20 coronal slices comprised pyramidal tract fibres descending from the primary motor and premotor cortex to the brainstem [y = 32 mm to y = –49 mm; coordinates in the Montreal Neurological Institute (MNI) space], slice thickness 3 mm without a gap between slices, 20 coronal slices (maximum of possible number of slices with the software) and voxel size 3 × 3 × 3 mm. Diffusion weighting was obtained with a Stejskal–Tanner spin echo preparation (Stejskal and Tanner, 1965) with b = 750 s/mm² and six gradient directions (gradient coordinate system, x, y, z): (0, 1, 1); (0, 1, –1); (1, 0, 1); (1, 0, –1); (1, 1, 0); and (1, –1, 0). In addition, we acquired a reference image without diffusion weighting.

Data processing was performed in Matlab 5.3 (MathWorks, Natick, MA, USA). As the STEAM sequence is insensitive to eddy currents, correction of geometrical distortion (shift, shear and scale artefacts) was not necessary. The images of each subject were realigned and the T1‐weighted data set was spatially normalized to the template using the software package SPM99 (Friston et al., 1995). Individual brain masks were used to exclude the scalp from normalization. The resulting affine and non‐linear transformations were applied to the diffusion‐weighted data. The diffusion gradient directions were rotated according to every rigid rotation in the process of normalization. The normalization allows a voxelwise group comparison based on the common stereotactic MNI space (Friston et al., 1995). To ensure correct normalization of the MRIs, we used voxel‐based morphometry on the white matter segment of the structural T1‐weighted images. Diffusion tensor maps were computed voxelwise on the normalized data with multivariate linear regression (Basser et al., 1994) using SPM99. Fractional anisotropy was derived from the diffusion tensor and the fractional anisotropy maps were smoothed with a Gaussian filter of 6 mm FWHM (full‐width half‐maximum) to validate the statistical inference for calculation of the corrected P values. The residuals were tested for normal distribution, an assumption for valid statistical inference through the probabilistic behaviour of Gaussian random fields (Worsley, 1994) as implemented in SPM99. We performed two voxel‐based statistics on the fractional anisotropy maps using SPM99. First, fractional anisotropy in patients was compared with that in healthy controls in a two‐sample t‐test. Secondly, a linear regression was calculated to determine the correlation between fractional anisotropy and CMCT in ALS patients separately for arms and legs. Patients with absent MEPs due to paresis and muscular atrophy or prolongation of peripheral motor conduction time were excluded from the regression analysis. Because of our a priori hypothesis regarding the pyramidal tract, we also performed small volume correction for multiple comparisons. Ellis et al. (1999) previously reported reduced fractional anisotropy in the posterior limb of the internal capsule in ALS patients compared with healthy controls. We centred the correction volume (sphere of 8 mm) at voxels with local fractional anisotropy maxima within the pyramidal tract from the primary motor and premotor cortex to the brainstem (P < 0.05, corrected). The images presented are thresholded at P < 0.01, uncorrected. To visualize the eigenvector, we show the projection of the first eigenvector onto the image planes.

Results

The voxel‐based morphometry on the white matter segments of the structural T1‐weighted images did not show any differences between patient and control groups, elucidating the accurate normalization of the images. Thus, any differences in fractional anisotropy can be attributed to differences in the diffusion‐weighted images.

We found a bilateral decrease in fractional anisotropy in the posterior half of the posterior limb of the internal capsule (right, P < 0.00001; left, P < 0.012, corrected) and in the corona radiata underneath the motor cortex (right, P < 0.03; left, P < 0.006, corrected) and premotor cortex (right, P < 0.018; left, P < 0.001, corrected) in the 15 patients with ALS compared with healthy controls (Figs 1 and 2, Table 2). The global maximum was located in the internal capsule (21, –15, –6) with P < 0.046, corrected for entire volume. The decrease in fractional anisotropy in the pyramidal tract extended to the brainstem (right, P < 0.022; left, P < 0.042, corrected) (Fig. 3, Table 2). Fractional anisotropy was also reduced in the corpus callosum (1, –12, 24) and right thalamus (9, –12, 4) in ALS patients compared with healthy subjects (P < 0.01, uncorrected) (Figs 1 and 3A).

Seven out of 15 patients had a pathological CMCT obtained by TMS to at least one extremity (Table 3). Two patients without any signs of UMN involvement on clinical investigation displayed pathological prolongation of the CMCT (Table 3).

The regression analysis revealed a negative correlation between fractional anisotropy and CMCT to the upper and lower extremities. Figure 4 shows regions with negative correlation between fractional anisotropy and CMCT separately for the arms and legs. They were located in the internal capsule (right arm, P < 0.033; left arm, P < 0.020, corrected; left leg, P < 0.053, corrected) and in the corona radiata underneath the motor and premotor cortex (right arm, P < 0.034; right leg, P < 0.035, corrected) (Fig. 4, Table 4). The diagrams in Fig. 4 show the regression lines for each extremity with the corresponding correlation coefficient r and the statistical P value.

The six patients without any clinical signs of UMN lesion at the time of MRI investigation also showed a bilateral reduction in fractional anisotropy in the corona radiata underneath the motor cortex (right, P < 0.021, corrected) and premotor cortex (right, P < 0.016; left, P < 0.02, corrected) and in the posterior half of the posterior limb of the internal capsule (right, P < 0.001; left, P < 0.034, corrected) (Fig. 5, Table 5). The decrease in fractional anisotropy in the pyramidal tract involved fibre bundles leading to the brainstem (right, P < 0.019; left, P < 0.045, corrected) (Fig. 6, Table 5). Additionally, fractional anisotropy was reduced in the corpus callosum (not shown) and in the right thalamus (10, –12, 7, P < 0.01, uncorrected) (Fig. 5B) in comparison with healthy subjects.

Discussion

We investigated the fibre integrity of the cerebral white matter in ALS patients compared with healthy controls using DTI and TMS. First, our voxel‐based statistics showed decreased fibre integrity in the corticospinal tract fibres descending from the motor and premotor cortex to the internal capsule and brainstem. Secondly, we found a negative correlation between fractional anisotropy and CMCT in the corticospinal tract fibres. The correlation analysis enables an anatomical subdivision of the corticospinal tract between fibres supplying the upper and lower extremities. Thirdly, a subgroup of patients with no signs of UMN lesion at the time of MRI investigation, who developed pyramidal tract signs later in the course of their disease, already showed reduced fibre integrity in the corticospinal tract underneath the premotor and motor cortex, in the internal capsule and in the brainstem in DTI. Additionally, we demonstrated reduced fractional anisotropy in the thalamus and in the corpus callosum in all patients investigated.

Corticospinal tract changes in ALS patients versus controls

Our voxel‐based statistic showed a significant bilateral decrease in fractional anisotropy in the posterior half of the posterior limb of the internal capsule of patients with ALS compared with healthy controls. This result agrees well with the known pathology of the combined motor neuron disease and the localization of the corticospinal tract fibres in the internal capsule (Hanaway and Young, 1977). A recent study using predefined ROIs for analysis of diffusion characteristics along white matter tracts descending through the internal capsule also demonstrated a reduction in the fractional anisotropy in the posterior limb of the internal capsule (Ellis et al., 1999). Because our field of view was not confined to the internal capsule but also included the primary motor and premotor cortex and parts of the brainstem, we were able to assess fibre disintegration at different levels of the motor system. Thus, we also detected a decrease in fractional anisotropy in more caudal parts of the corticospinal tract, comprising the mesencephalon and pons. Additionally, we found a decrease in fractional anisotropy underneath the motor and premotor cortex in ALS patients compared with controls.

The decrease in fractional anisotropy underneath the motor cortex corresponds to degeneration of motor fibres emerging from the primary motor cortex (Davidoff, 1990). About 60% of pyramidal tract axons originate from the primary motor cortex (Brodmann area 4) while the remaining fibres arise from the premotor cortex (Brodmann area 6) and the parietal lobe (Davidoff, 1990). Thus, the reduced fractional anisotropy underneath the premotor cortex can be explained by disintegration of pyramidal tract fibres emerging from the premotor cortex. Anatomical connections between the premotor and motor cortex could also contribute to the decrease in fractional anisotropy. Clinically, ALS patients may display degenerative changes in the premotor cortex (Lawyer and Netsky 1953; Pioro et al., 1994). This can result in degeneration of the fibres emerging from the premotor cortex, resulting in a decrease in fractional anisotropy underneath it. Hence, we demonstrated white matter disintegration along the corticospinal tract.

Extramotor involvement in ALS and white matter changes beyond the corticospinal tract

The decrease in fractional anisotropy in the corpus callosum reflects the neuropathological findings of degenerated pyramidal tract bundles running across the middle of the corpus callosum in patients with ALS (Brownell et al., 1970). It also corresponds to the atrophy of the corpus callosum that is often found in ALS and to other pyramidal tract degeneration processes (e.g. in familiar spastic paralysis), which are often observed in conventional MRI (Yamauchi et al., 1995; Krabbe et al., 1997). Severe atrophy in the anterior fourth of the corpus callosum is also associated with cognitive decline and psychiatric symptoms in ALS (Yamauchi et al., 1995).

Many pyramidal tract axons terminate in, or send collateral branches to, a number of supraspinal structures (e.g. the striatum, sensory and motor nuclei of the thalamus, red nucleus, pontine nuclei, midbrain and bulbar reticular formation, inferior olive, dorsal column nuclei, trigeminal nuclei) (Davidoff, 1990). Regarding the thalamus, necropsy studies have shown degeneration in the thalamus in patients with ALS (Iwanaga et al., 1997). The decrease in fractional anisotropy in the thalamus in this study agrees well with this result. In addition, ALS patients with impaired verbal fluency showed significantly attenuated rCBF responses in the anterior thalamic nuclear complex in PET studies (Kew et al., 1993). Patients with familiar ALS also revealed bilateral thalamic hypoperfusion as well as parietal and frontal hypoperfusion in single photon emission computed tomography (SPECT) (Kumar and Abdel‐Dayem, 1999).

In addition to the areas reported above, we found decreased fractional anisotropy in extramotor regions in the frontal lobe [underneath the left inferior frontal gyrus (Brodmann areas 44, 45 and 47) and underneath the medial frontal gyrus (Brodmann areas 8 and 9)]. Similar areas have been reported in previous ALS studies (Lloyd et al., 2000; Ellis et al., 2001).

Correlation between fractional anisotropy and CMCT

A regression analysis was calculated to determine the correlation between fractional anisotropy and CMCT obtained by TMS. The displayed regions exhibited a negative correlation between fractional anisotropy and CMCT to all extremities in patients with ALS. In these voxels, an increase in CMCT was associated with a decrease in fractional anisotropy. The separation of the CMCTs to all extremities into four groups (one extremity per group) and the correlation with fractional anisotropy for each extremity allows the anatomical differentiation of the corticospinal tract between fibres to the arms and legs. The observed topology of the corticospinal tract fibres to the arms and legs in the internal capsule corresponds to the known anatomical topology. In the internal capsule the motor fibres to the legs run posterior and lateral to the fibres to the arm. In contrast, the white matter fibres to the legs directly underneath the premotor and motor cortex are located medial to the fibres supplying the upper extremity, as we showed in the coronal section. These results suggest an influence of white matter disintegration in the corticospinal tract on the prolongation of the CMCT in ALS patients.

Patients without signs of UMN involvement at the time of MRI investigation

This study demonstrated that ALS patients without any clinical evidence of UMN lesion at the time of MRI investigation also showed a bilateral reduction in fractional anisotropy in the posterior limb of the internal capsule, underneath the motor and premotor cortex and in the brainstem, in comparison with healthy controls. Additionally, the fractional anisotropy of the corpus callosum, right thalamus and frontal lobe in these patients was reduced when compared with that in normal subjects. Hence, DTI can be used to detect early lesions of the corticospinal tract in patients with ALS and may therefore contribute to earlier diagnosis of the disease in the future. This result matches findings from post‐mortem cytochemistry analyses in patients with the progressive muscular atrophy variant of ALS, who frequently have undetected corticospinal tract pathology (Ince et al., 2003). We are planning single‐subject analyses of ALS patients without clinical signs of UMN involvement in order to explore the clinical impact of DTI on the assessment of UMN lesions. This may help to differentiate between upper and LMN involvement at an early stage of the disease.

Acknowledgements

We wish to thank all the patients and test subjects for their cooperation in the study. This work was supported by a grant from the Karberg Stiftung.

Fig. 1 Voxels with a significant decrease in fractional anisotropy the pyramidal tract, corpus callosum and right thalamus (shown in green) in 15 ALS patients compared with healthy controls. Coordinates are in MNI space (mm). (A) Pyramidal tract (21, –15, –6; global maximum with P < 0.05, corrected for entire volume; –24 –12, 12), corpus callosum (1, –12, 24) and right thalamus (9, –12, 4) in this coronal section. (B) Transverse view of pyramidal tract in the posterior half of the posterior limb of the internal capsule (21, –15, –6; –24, –12, 12). Statistical fractional anisotropy results (thresholded at P < 0.01, uncorrected) of ALS patients versus controls are superimposed on a spatially normalized T1‐weighted MRI. Corrected P values (small volume correction) are given in Table 2.

Fig. 2 Voxels with significantly decreased fractional anisotropy underneath the motor and premotor cortex (shown in green) in 15 ALS patients compared with controls. Coordinates are in MNI space (mm). (A) Coronal section at y = –18. Pyramidal tract descending from the motor cortex (24, –18, 39; –21, –21, 42) and premotor cortex (–18, –18, 60). (B) Coronal section at y = –3. Pyramidal tract bilaterally underneath the premotor cortex (21, –3, 42; –42, –3, 33). Statistical fractional anisotropy results (thresholded at P < 0.01, uncorrected) are superimposed on a spatially normalized T1‐weighted MRI. Main fibre direction is shown by the dashed yellow lines. For corrected P values after small volume correction, see Table 2.

Fig. 3 Voxels with a significant decrease in fractional anisotropy in the caudal pyramidal tract fibres to the brainstem (shown in red) in 15 ALS patients compared with healthy subjects. Coordinates are in MNI space (mm). (A) Coronal view of pyramidal tract fibres in the brainstem (9, –18, –21; –6, –15, –21). (B) Transverse view of pyramidal tract fibres descending through the brainstem (9, –18, –21; –6, –15, –21). Statistical fractional anisotropy results (thresholded at P < 0.01, uncorrected) are superimposed on a spatially normalized T1‐weighted MRI. The dashed yellow lines show the projection of the first eigenvector, which corresponds to the main fibre direction. Corrected P values (small volume correction) are listed in Table 2.

Fig. 4 Voxels with negative correlation between fractional anisotropy and CMCT to the extremities are shown in green (hands) and red (legs) in patients with ALS. Overlapping voxels are shown in yellow. Coordinates are in MNI space (mm). (A) Coronal section at y = –12. Negative correlation between fractional anisotropy and CMCT to hands (24, –12, 12; –27, –12, 15) and to leg (–27, –18, 15) in the pyramidal tract at level of the internal capsule. (B) Coronal section at y = –18. Negative correlation between fractional anisotropy and CMCT to hand (21, –18, 42) and leg (12, –15, 57) in pyramidal tract fibres descending from the motor and premotor cortex. (C) Transverse view of regions with negative correlation between fractional anisotropy and CMCT in the posterior half of the posterior limb of the internal capsule with pyramidal tract fibres supplying the hands (24, –12, 12; –27, –12, 15) and leg (–27, –18, 15). A–C show statistical fractional anisotropy results (thresholded at P < 0.01, uncorrected) superimposed on a spatially normalized T1‐weighted MRI. Corrected P values (small volume correction) are given in Table 4. The surrounding diagrams display the regression line for each extremity with the corresponding correlation coefficient r and statistical P value.

Fig. 5 Voxels with significantly decreased fractional anisotropy in the pyramidal tract underneath the motor and premotor cortex, in the internal capsule and in the right thalamus (shown in green) in a subgroup of ALS patients without upper motor neuron signs at the time of MRI investigation, in comparison with healthy controls. Coordinates are in MNI space (mm). (A) Coronal view at y = –3: pyramidal tract underneath the premotor cortex (39, –3, –39; –42, –3, 33). (B) Coronal view at y = –12: pyramidal tract underneath the motor cortex (30, –9, 39); pyramidal tract in the internal capsule (21, –15, –6; –24, 12, 12); right thalamus (10, –12, 7; P < 0.01, uncorrected). (C) Transverse view of pyramidal tract fibres in the posterior limb of the internal capsule (21, –15, –6; –24, –12, 12). Statistical fractional anisotropy results (thresholded at P < 0.01, uncorrected) are superimposed on a spatially normalized T1‐weighted MRI. Corrected P‐values (small volume correction) are provided in Table 5.

Fig. 6 Voxels with a significant decrease in fractional anisotropy in the caudal pyramidal tract fibres (shown in red) in ALS patients without signs of upper motor neuron lesion, compared with healthy controls. Coordinates are in MNI space (mm). (A) Coronal view of pyramidal tract fibres descending to the brainstem (12, –18, –18; –9, –18, –21). (B) Transverse view of pyramidal tract fibres in the mesencephalon (12, –18, –18; –9, –18, –21). Statistical fractional anisotropy results (thresholded at P < 0.01, uncorrected) superimposed on a spatially normalized T1‐weighted MRI. Main fibre orientation is shown by the dashed yellow lines. Corrected P values (small volume correction) are given in Table 5.

Table 1

Clinical details of patients with ALS

Patient,sex	Diseaseonset (yr)	Disease duration(months)	El Escorial criteria	Clinicalonset	UMN involvement	LMN involvement
01, F	59	11	(Probable ALSlaboratory‐supported)	Bulbar	(Emotional lability, brisk DTR in all limbs)	B, dysphagia, loss of masseter reflex,resp. insuff., C, paresis of velum/facial muscles,general fasciculation
02, F	62	6	Definite ALS	Limb	Emotional lability, Babinski sign, brisk DTRin all weak wasted limbs	B, dysphagia, loss of masseter reflex, tetraparesis,muscle atrophy, general fasciculation
03, M	58	6	(Definite ALS)	Bulbar	(Emotional lability, brisk DTR in all weakwasted limbs)	B, dysphagia, paresis of facial muscles (all limbs),atrophy of arm, trunk (all limbs)
04, F	27	10	Possible ALS	Bulbar	Emotional lability, hyperactive gag reflex,attenuated abdominal reflexes	Paresis of tongue, velum, lips and arm
05, F	59	6	(Probable ALS)	Bulbar	(Emotional lability, Babinski sign, leg spasticity)	B, dysphagia, paresis of right arm, muscle atrophyof arms, general fasciculation
06, M	63	9	Probable ALS,laboratory‐supported	Limb	Hyper‐reflexia right leg, Babinski sign	Paresis of arms/left foot, muscle atrophy,general fasciculation
07, M	56	6	Probable ALS	Bulbar	Hyper‐reflexia both arms, loss ofabdominal reflexes	B, dysphagia, paresis of arms and right leg,general fasciculation
08, M	58	14	Probable ALS	Limb	Leg spasticity, Babinski sign, abdominal reflexesattenuated, brisk DTR in all weak wasted limbs	Tetraparesis, muscle atrophy, C, general fasciculation
09, M	57	21	Probable ALS	Limb	Loss of abdominal reflexes, foot clonus	Tetraparesis with atrophy, C, general fasciculation,attenuation of DTR left arm
10, M	56	13	Probable ALS,laboratory‐supported	Limb	Attenuation of abdominal reflexes, Babinski sign	Paresis of legs/trunk, atrophy of legs, loss ofDTR right leg, general fasciculation
11, M	31	24	Probable ALS	Limb	Hyper‐reflexia right arm, legs, abdominalreflexes attenuated	Paresis and muscle atrophy of arms, generalfasciculation, C
12, M	37	15	(Possible ALS)	Limb	(Hyper‐reflexia and spasticity of legs)	Tetraparesis, atrophy, DTR of arms reduced
13, M	61	17	(Possible ALS)	Limb	(Brisk DTR in the weak wasted right leg)	Paresis and muscle atrophy of the legs,general fasciculation
14, M	42	8	(Probable ALSlaboratory‐supported)	Limb	(Hyper‐reflexia left limbs, Babinski sign)	Dysarthria, fibrillation, paresis of both arms,fasciculation both arms, and chest
15, F	57	12	Definite ALS	Bulbar	Emotional lability, loss of abdominal reflexes,Babinski sign, spasticity of legs	B, dysphagia, facial muscle atrophy, loss of masseterreflex, tetraparesis, atrophy both arms, generalfasciculation, C

F = female; M = male; B = bulbar paralysis with dysarthria, paresis and atrophy of the tongue; DTR = deep tendon reflexes; C = cramps; () = clinical upper and lower motor neuron signs which the six ALS patients who had no symptoms of UMN lesion at the time of the MRI scan developed as their disease progressed; brackets are also used for the resulting El Escorial diagnosis that affirmed the diagnosis of ALS in this subgroup. In the course of their disease, all patients could be categorized according to the revised El Escorial criteria (Brooks et al., 2000b); ‘general fasciculation’ and ‘muscle atrophy’ appear in the table in the case of LMN signs in all limbs as well as the trunk.

Table 2

Regions of decreased fractional anisotropy in 15 ALS patients versus healthy controls within the pyramidal tract at the small volume correction level (sphere of 8 mm) centred at voxels with local maxima (Figs 1–3)

Region	15 ALS patients compared with healthy controls
	x, y, z	Z score	P corrected
Pyramidal tract in the internal capsule (Fig. 1A, B)	21, –15, –6,	4.54	0.000
	–24, –12, 12	3.61	0.012
Pyramidal tract in the brainstem (Fig. 3A, B)	9, –18, –21	3.45	0.022
	–6, –15, –21	3.27	0.042
Pyramidal tract underneath motor cortex, BA 4 (Fig. 2A, B)	24, –18, 39	3.37	0.030
	–21, –21, 42	3.79	0.006
Pyramidal tract underneath premotor cortex, BA 6 (Fig. 2B)	21, –3, 42	3.51	0.018
	–42, –3, 33	4.29	0.001
	–18, –18, 60	3.31	0.037

Coordinates (x, y, z) are in MNI space. BA = Brodmann area; P corrected = P value corrected for multiple comparisons.

Table 3

Central motor conduction time (ms) of patients with ALS

Patient, sex	M. interosseus dorsalis DI			M. tibialis anterior
	Right	Left	Normal	Right	Left	Normal
01, F	8.1	8.3	8.7	16.2	14.7	19.9
02, F	8.7	7.7	8.7	26.2	22.1	19.9
03, M	6.4	6.8	7.8	15.1	14.5	17.7
04, F	7.9	7.5	7.8	12.9	12.5	17.2
05, F	8.7	7.9	7.8	16.5	16.8	17.7
06, M	7.1	7.0	8.7	15.3	17.0	19.9
07, M	7.9	7.8	7.8	16.1	15.8	17.7
08, M	0M	0M	7.8	14.9	15.1	17.7
09, M	8.4	6.2	8.7	16.4	16.4	19.9
10, M	9.7	8.7	7.8	26.3	24.8	17.7
11, M	0M	9.2	7.8	19.7	14.4	17.7
12, M	0M	0M	7.8	0M	0M	17.7
13, M	5.8	exc.	8.7	19.5	0M	19.9
14, M	0M	9.0	7.8	19.3	16.9	17.7
15, F	0M	0M	7.8	0M	0M	17.7

F = female; M = male; 0M = absent MEP due to paresis and muscular atrophy; normal = upper limit of normal values for age of the patient (Kloten et al., 1992); exc. = excluded due to prolongation of peripheral motor conduction time. Bold face indicates prolonged central motor conduction time.

Table 4

Regions of decreased fractional anisotropy (FA) correlated with central motor conduction time (CMCT) separately for arms and legs in patients with ALS within the pyramidal tract at the small volume correction level (sphere of 8 mm), centred at voxels with local maxima (Fig. 4)

Region	Correlation FA and CMCT to arms			Correlation FA and CMCT to legs
	x, y, z	Z score	P corrected	x, y, z	Z score	P corrected
Pyramidal tract in internal capsule (Fig. 4A, C)	24, –12, 12	3.34	0.033
	–27, –12, 15	3.48	0.020	–27, –18, 15	3.21	0.053
Pyramidal tract underneath motor cortex,BA 4/premotor cortex, BA 6 (Fig. 4B)	21, –18, 42	3.33	0.034	12, –15, 57	3.32	0.035

Coordinates (x, y, z) are in MNI space. BA = Brodmann area; P corrected = P value corrected for multiple comparisons.

Table 5

Regions of decreased fractional anisotropy in ALS patients without upper motor neuron (UMN) involvement at the time of investigation within the pyramidal tract at the small volume correction level (sphere of 8 mm), centred at voxels with local maxima (Figs 5 and 6)

Region	ALS patients with no UMN signs versus controls
	x, y, z	Z–score	P corrected
Pyramidal tract underneath premotor cortex, BA 6 (Fig. 5A)	39, –3, 39	3.49	0.016
	–42, –3, 33	3.42	0.020
Pyramidal tract underneath motor cortex, BA 4 (Fig. 5B)	30, –9, 39	3.41	0.021
Pyramidal tract in the internal capsule (Fig. 5B, C)	21, –15, –6	4.32	0.001
	–24, –12, 12	3.22	0.034
Pyramidal tract in the brainstem (Fig. 6A, B)	12, –18, –18	3.44	0.019
	–9, –18, –21	2.85	0.045

Coordinates (x, y, z) are in MNI space; BA = Brodmann area; P corrected = P value corrected for multiple comparisons.