The combined use of conventional MRI and MR spectroscopic imaging increases the diagnostic accuracy in amyotrophic lateral sclerosis

Page 1

Ti

AVVa

b

c

d

a

ARRA

KCAMHHM

1

eamiu

vf

h0

European Journal of Radiology 84 (2015) 151–157

Contents lists available at ScienceDirect

European Journal of Radiology

journa l homepage: www.e lsev ier .com/ locate /e j rad

he combined use of conventional MRI and MR spectroscopic imagingncreases the diagnostic accuracy in amyotrophic lateral sclerosis

medeo Cervoa,1, Sirio Cocozzaa,∗,1, Francesco Saccàb, Sara M.d.A. Giorgioa,incenzo Brescia Morrab, Enrico Tedeschia, Angela Marsili b, Giovanni Vaccab,incenzo Palmac, Arturo Brunetti a, Mario Quarantelli d

Department of Advanced Biomedical Sciences, University “Federico II”, Naples, ItalyDepartment of Neurosciences, Reproductive Sciences and Odontostomatology, University “Federico II”, Naples, ItalyU.O.C. Neurofisiopatologia, PO S. Gennaro ASL Napoli 1, Naples, ItalyBiostructure and Bioimaging Institute, National Research Council, Naples, Italy

r t i c l e i n f o

rticle history:eceived 27 August 2014eceived in revised form 21 October 2014ccepted 28 October 2014

eywords:onventional MRImyotrophic lateral sclerosisotor cortexyperintensityypointensityagnetic Resonance Spectroscopy

a b s t r a c t

Purpose: We aimed to assess, in amyotrophic lateral sclerosis (ALS), the diagnostic accuracy of thecombined use of conventional MRI signal changes (namely, hypointensity of the precentral cortex andhyperintensity of the corticospinal tracts on T2-weighted images), and N-Acetyl-Aspartate (NAA) reduc-tion in the motor cortex at Magnetic Resonance Spectroscopy (MRS), which are affected by limiteddiagnostic accuracy when used separately.Methods: T2-hypointensity and NAA/(Choline + Creatine) ratio of the precentral gyrus and T2-hyperintensity of the corticospinal tracts were measured in 84 ALS patients and 28 healthy controls,using a Region-of-Interest approach.

Sensitivity and specificity values were calculated using Fisher stepwise discriminant analysis, andcross-validated using the leave-one-out method.Results: Precentral gyrus T2 signal intensity (p < 10−4) and NAA peak (p < 10−6) were significantly reducedin patients, and their values did not correlate significantly to each other both in patients and controls,while no significant differences were obtained in terms of T2-hyperintensity of the corticospinal tract.Sensitivity and specificity of the two discriminant variables, taken alone, were 71.4% and 75.0%, for

NAA peak, and 63.1% and 71.4% for T2-hypointensity, respectively. When using these two variables incombination, a significant increase in sensitivity (78.6%) and specificity (82.1%) was achieved.Conclusions: Precentral gyrus T2-hypointensity and NAA peak are not significantly correlated in ALSpatients, suggesting that they reflect relatively independent phenomena. The combined use of thesemeasures improves the diagnostic accuracy of MRI in ALS diagnosis.

. Introduction

Amyotrophic lateral sclerosis (ALS) is an idiopathic neurodegen-rative disorder characterized by selective degeneration of brainnd spinal motor neurons controlling voluntary muscle move-

ents. In ALS diagnosis, brain Magnetic Resonance Imaging (MRI)

s often performed to exclude other mimicking conditions, partic-larly when bulbar onset occurs.

∗ Corresponding author at: Department of Advanced Biomedical Sciences, Uni-ersity “Federico II”, Via Pansini, 5, 80131 Naples, Italy. Tel.: +39 081 2203187x216;ax: +39 081 2296117; mobile +39 333 6078796.

E-mail address: [email protected] (S. Cocozza).1 These authors contributed equally to this work.

ttp://dx.doi.org/10.1016/j.ejrad.2014.10.019720-048X/© 2014 Elsevier Ireland Ltd. All rights reserved.

© 2014 Elsevier Ireland Ltd. All rights reserved.

In the past years, several studies have shown different find-ings at conventional MRI in ALS patients. These include increasedsignal of corticospinal tracts (CST) on T2-weighted images [1–4](Fig. 1) and reduced signal intensity of motor cortex on T2-weightedimages [2,5–7] (Fig. 2).

Beside conventional MRI, advanced MR techniques have beenintroduced to investigate ALS patients, such as Magnetic ResonanceSpectroscopy (MRS) or Diffusion Tensor Imaging (DTI). MRS con-sistently showed a reduction of N-Acetyl-Aspartate (NAA) in theprimary motor cortex and along the CST [8–11].

More recently, 3D MRS acquisitions have allowed MRS measure-

ments across the entire brain during a single scan, showing thepossibility to follow neuronal loss along the CST [12].

On the other hand DTI measures have shown not only to beable to discriminate between controls and ALS patients, but even

Page 2

152 A. Cervo et al. / European Journal of Radiology 84 (2015) 151–157

Fig. 1. Corticospinal tract hyperintensities in ALS. Signal alterations along the corticospinal tracts in T2-weighted axial TSE images, in the ALS patient with the highestCST-Hyper value (bottom, 58yo male, 3 month disease duration), compared to a 55yo male HC (top). Sections through cerebral peduncles (left) and internal capsule (right)are shown. Hyperintensity of the corticospinal tract can be clearly appreciated at both levels in the patient. At the same levels, only a mild hyperintensity can be also seen inthe HC images.

Fig. 2. Precentral gyrus hypointensity in ALS. Hypointensities along the precentral cortex in T2-weighted axial TSE images, in two ALS patients, compared with an age/ands reciated and an3 the mo

ae

sia(s

d

ex/matched HC. A significant hypointensity of the motor cortex can be clearly appisease duration, ALSFRS-R 19), compared to both the HC (on the left, 68yo female)month disease duration, ALSFRS-R 36). Sections at the level of the hand knobs of

mong different phenotypes, showing also a correlation with dis-ase severity [13].

Alongside with brain imaging in ALS, few studies investigatedpinal cord abnormalities in ALS patients (reviewed in [14]), show-ng alteration in both conventional (reduced cross-sectional areand T1 hyperintensities in the antero-lateral columns) and DTI

lower mean fractional anisotropy) acquisitions in the cervicalpinal cord.

However, both conventional and advanced MR techniqueso not provide a satisfactory level of confidence about the

d in the patient more severely impaired (on the right, 67yo female with 17 monthearly stage patient with shorter disease duration (in the middle, 68yo female withtor cortex are shown.

diagnosis of ALS, and the few studies assessing the combineduse of these parameters were limited to small groups of patients[1,15].

In the present study, we apply a multiparametric MR approachto assess the relationship between these measures and the sen-sitivity and specificity of their combined use in ALS diagnosis.

Our approach includes the evaluation, in the precentral gyrus(PCG), of the presence of T2 hypointensity (PCG-hypo) and of theNAA/(Choline + Creatine) ratio (PCG-NAA), and, along the CST, ofthe T2 hyperintensity (CST-hyper).

Page 3

nal of

2

2

tOd

bp

ri

sM

miHa

2

3G

wo

sl

roavc“

h

2

sFS

2

a

TS

HAa

A. Cervo et al. / European Jour

. Methods

.1. Subjects

We retrospectively collected the MRI scans performed withinhe standard workup for ALS at our Institution from 2006 to 2013.nly patients fulfilling the criteria for a diagnosis of probable orefinite ALS according to El Escorial were included in the analysis.

Additionally, a group of 28 healthy controls (HC) of compara-le age and sex, who underwent an MRI study over the same timeeriod and with the same acquisition protocol, was analyzed.

Disease duration (DD) was measured based on the first occur-ence of motor symptoms clearly related to the disease, as reportedn clinical records.

The revised ALS functional rating scale (ALSFRS-R) score, a mea-ure of disease severity [16], collected within two weeks from theRI study, was available for all patients.Subjects with no history of neurological disorders or any other

edical condition that could affect the central nervous system werencluded in the HC group. Demographic information of patients andC data included in the study are listed in Table 1. This study waspproved by the local Ethics Committee.

.2. MRI data acquisition

All MRI studies were acquired with identical protocol on aTesla MRI scanner (Trio, Siemens Medical Systems, Erlangen,ermany).

The sequences used for the analysis included Fast Spin-Echo T2-eighted axial images and MRS imaging, acquired with the same

rientation.T2-weighted images had the following parameters: 25 axial

lices, slice thickness = 4 mm, TE = 105 ms, TR = 4500 ms, echo trainength = 13, acquisition matrix = 3842, FOV = 230 mm × 230 mm.

2D multivoxel 1H MRSI was performed using a spin-echo (point-esolved spectroscopy) sequence with water suppression by meansf selective excitation (TE = 270 ms, FOV = 160 mm × 160 mm,cquisition matrix = 16 × 16, thickness = 15 mm, zero-filled to aoxel size of 5 mm × 5 mm × 15 mm), acquiring a single axial slice,entered at the same level of the T2-weighted slice where handknobs” of the motor cortex was best seen.

During the MRI study the subjects were laying supine with theead fixed by straps and foam pads to minimize head movement.

.3. Conventional MRI analysis

T2-weighted images were analyzed to assess PCG hypointen-ity and CST hyperintensity by using a set of pre-defined ROIs (seeigs. 3 and 4), placed in consensus by three trained raters (SC, AC,MDAG), blinded to the clinical diagnosis.

.3.1. PCG-hypointensity evaluationPre- and post-central cortices were sampled on the T2-weighted

xial slice in which the hand knob of the motor cortex was

able 1ubjects demographics and clinical variables.

HC ALS

Age (mean ± SD) 57.5 ± 13.9 (range 29–80) 61.1 ± 11.2 (range 28–81)Sex (M/F) 12/16 52/32ALSFRS-R (median) n/a 38 (range 19–47)DD (mean ± SD) n/a 1.5 ± 1.4AAO (mean ± SD) n/a 59.6 ± 11.3 (range 27–80)

C, healthy controls; ALS, amyotrophic lateral sclerosis; SD, standard deviation;LSFRS-R, amyotrophic lateral sclerosis functional rating scale revised; n/a, notpplicable; DD, disease duration; AAO, age at onset. Ages and DD are in years.

Radiology 84 (2015) 151–157 153

best represented. For each exam, two linear ROIs were bilaterallyhand-drawn on the PCG and the postcentral gyrus of the selectedslice using a 2-mm brush tool of a commercial biomedical imageprocessing software (Osirix 5.6 Pixmeo, Geneva, Switzerland;www.osirix-viewer.com) (Fig. 3).

These ROIs were placed within the cortical rim along the entirelength of the central sulcus, avoiding CSF.

As a measure of PCG hypointensity, the minimum pixel valuein the corresponding ROI, normalized by the mean value of thepost-central ROI, was used. The use of ROI’s minimum value, ratherthan the average value over the PCG ROI, has been chosen to assessthe grade of PCG hypointensity, in order to avoid the confoundingeffect of both CSF and subcortical white matter hyperintensities(e.g. gliosis) that could affect ROI mean values.

2.3.2. CST-hyperintensity evaluationTo sample CST signal intensity, three circular ROIs (6 mm of

diameter each) were bilaterally placed along the course of each CST(one at the level of the cerebral peduncles, one in the posterior limbof the internal capsule and one in the subcortical white matter adja-cent to the PCG), centered on the relative local maxima. For eachROI, corresponding mean signal intensity was normalized by themean value of a control ROI placed in the same slice on a structureapparently unaffected by the disease [17] (Fig. 4). For the evalu-ation of CST hyperintensity, we could not use a similar approach(i.e. measuring the maximum value of the ROI) to the one usedfor the PCG-hypo analysis, due to the presence of CST-neighboringstructures with high-intensity signal, such as cisternal CSF, dilatedVirchow–Robin spaces, and gliosis.

2.4. MRS analysis

Spectroscopy data were analyzed using the software availableon the scanner (Syngo MR B17, Siemens, Erlangen, Germany). Pro-cessing included Fourier transformation, Gaussian filtering in thetime domain, and phase and baseline correction, followed by peakidentification of Choline (Cho, centered at 3.22 ppm), creatine andphosphocreatine (Crea, 3.02 ppm), and NAA (2.02 ppm), and fittingof the corresponding integrals under the curve.

To assess upper motor neuron (UMN) density/integrity, spec-tra from voxels immediately anterior to the bilateral central sulci(Fig. 5) were averaged. As the ratios of NAA to Crea and/or Chohave been proposed as markers of UMN loss/dysfunction, in thepresent work we selected the ratio that best discriminated the twopopulations for comparison with other MRI measures.

2.5. Statistics

Student’s t-test was used to assess differences in age betweenHC and patients. Between-group differences in gender were testedby the Chi-squared test.

As an influence of age has been previously suggested on CST-hyper, PCG-hypo and PCG-NAA in normal subjects [18–20], thesignificance of these correlations was tested in the HC group bySpearman’s correlation coefficient.

Correlation between the tested MR variables and the clinicaldata (ALSFRS, DD and age at onset – AAO) was assessed in patientsby Spearman’s correlation coefficient.

Sensitivity and specificity values, as well as the mean classi-fication accuracy, were calculated to show the predictive powerof the combined MRI measures during the classification process,

using Fisher stepwise discriminant analysis, and cross-validatedusing the leave-one-out method. During the leave-one-out cross-validation, each case is extracted once and treated as test data,while the remaining cases are treated as a new dataset.

Page 4

154 A. Cervo et al. / European Journal of Radiology 84 (2015) 151–157

Fig. 3. Pre- and post-central gyrus ROIs for conventional MRI analysis. T2-weighted axial image at the level of the hand knob of the motor cortex with (A) and without (B)the two superimposed hand-drawn ROIs used for the assessment of PCG-hypo (red) and of the mean post-central gyrus signal intensity (green). ROIs were drawn using a2 mm-brush tool along the entire length of the central sulcus, avoiding CSF. Relative T2-hypointensity of the PCG can be appreciated.

F of the cerebral peduncles (A), posterior limb of the internal capsule (B) and the hand knobo e corresponding ROIs used for data normalization (in green), placed within the midbraint yrus (C).

W

f1

3

g

dtra

tt

p

i

t

Table 2PCG-Hypo, PCG-Hyper and PCG-NAA values.

Variable HC ALS

PCG-NAA 1.35 ± 0.13 1.19 ± 0.14 p < 10−6

PCG-hypo 0.65 ± 0.7 0.56 ± 0.10 p < 10−4

PCG-hyper MC 0.99 ± 0.6 1.02 ± 0.15 n.s.PCG-hyper IC 1.25 ± 0.15 1.26 ± 0.14 n.s.PCG-hyper CP 0.85 ± 0.09 0.81 ± 0.09 n.s.

PCG-NAA, NAA/Cho+Crea ratio in the precentral gyrus; PCG-hypo, signal intensityof the precentral gyrus; PCG-hyper MC, signal intensity of the white matter at thelevel of the motor cortex; PCG-hyper IC, signal intensity of the white matter at thelevel of the posterior limb of the internal capsule; PCG-hyper CP, signal intensity ofthe white matter at the level of the cerebral peduncles.For each measure, mean ± standard deviation is reported, along with the significance

ig. 4. Corticospinal tract ROI positioning. Selected T2-weighted images at the levelf the motor cortex (C), with superimposed ROIs placed on the CST (in red), and thegmentum (A), in the splenium of the corpus callosum (B) and in the postcentral g

Fisher stepwise discriminant analysis was performed usingilk’s Lambda as the chosen selection criterion.All statistical analyses were performed using Statistical Package

or Social Science (SPSS) package (SPSS Inc. Released 2007, Version6.0. Chicago, SPSS Inc.).

. Results

ALS and HC groups were not significantly different for age andender.

The NAA/(Cho + Crea) ratio proved to be the best measure toiscriminate between the two populations (p < 10−6 at Student’s-test), compared to NAA/Cho (p < 10−3) and NAA/Crea (p < 10−4)atios. For this reason, this measure was used for the subsequentnalysis as spectroscopic marker of neuronal loss.

No significant correlation emerged between any pair of the threeested variables neither in the HC nor in the ALS group, suggestingheir relative independence.

PCG-NAA, PCG-hypo and CST-hyper measures in HC and ALSatients are reported in Table 2.

PCG-NAA alone correctly classified 60/84 ALS patients (sensitiv-ty 71.4%), with a misclassification of 7/28 HC (specificity 75.0%).

PCG-hypo alone correctly classified 53/84 ALS patients (sensi-ivity 63.1%), with a misclassification of 8/28 HC (specificity 71.4%).

of the difference between HC and ALS patients at Mann/Whitney test.n.s., not significant.

CST-hyper alone correctly classified 35/84 ALS patients (sen-sitivity 43.8%), with a misclassification of 11/28 HC (specificity60.7%).

At Fisher stepwise discriminant analysis, only PCG-NAA andPCG-hypo were selected as able to discriminate between SLA andHC subjects. Leave-one-out cross validation showed that when

PCG-NAA and PCG-hypo were used in combination, 66/84 sub-jects could be correctly classified as ALS patients, providing asubstantial increase in sensitivity (78.6%), with 5/28 HC showing

Page 5

A. Cervo et al. / European Journal of Radiology 84 (2015) 151–157 155

Fig. 5. Precentral gyri spectrum with corresponding MRS slab. Representative spectrum from the precentral regions at the level of the hand knobs in both a 50yo femaleH multc rs forp

a8

a

al

C (top) and a 48yo female ALS patient (bottom). On the right, for each subject, theorresponding to the center of the MRS slab, along with coronal and sagittal localizeatient.

n abnormal value in at least one of the two measures (specificity2.1%).

Sensitivity and specificity for each measure, and the overall

ccuracy of the method, are reported in Table 3.

Finally, when testing the relationship between MR measuresnd clinical variables, only PCG-hypo showed a significant corre-ation with ALSFRS-R (p < 0.01; R = 0.413).

ivoxel MRS sampling grid is shown superimposed on the T2-weighted axial imagereference. A clear reduction in the NAA/(Cho + Crea) ratio can be appreciated in the

4. Discussion

ALS diagnosis can be challenging, especially in

patients without clear symptoms and signs related tothe disease, or when other diseases/conditions (e.g.medullary compression due to spinal pathology) arepresent.

Page 6

156 A. Cervo et al. / European Journal of

Table 3Diagnostic accuracies of PCG-NAA, PCG-hypo and their combined use in ALS patients.

Variable Sensitivity Specificity Overall accuracy

PCG-NAA 60/84 (71.4%) 21/28 (75.0%) 81/112 (72.3%)PCG-hypo 53/84 (63.1%) 20/28 (71.4%) 73/112 (65.1%)CST-hyper 35/84 (43.8%) 17/28 (60.7%) 52/112 (46.4%)Combined 66/84 (78.6%) 23/28 (82.1%) 89/112 (79.4%)

PCG-NAA, NAA/Cho+Crea ratio in the precentral gyrus; PCG-hypo, signal intensityo

cics

vs

dtrdddi

ism[raiad

trdw

aanaats

pwA

itcep

absl

Secondly, other advanced techniques proposed as promisinginstruments for ALS diagnosis have not been included here.

f the precentral gyrus; CST-hyper, signal intensity along the corticospinal tract.

MRI is usually performed to exclude the possibility of otheronditions known as “Motor neuron disease mimic” syndromes,ncluding, but not limited to, cerebral or skull base lesions,ervical spondylotic myelopathy, foramen magnum lesions, andyringomyelia [21].

It has been proposed that MRI could provide additional rele-ant information, including findings characteristic of ALS, althoughomewhat conflicting results have been reported.

Hyperintensity of CST in T2-weighted images has beenescribed in several studies [1–4]. However, this signal modifica-ion has been shown to be not sensitive for ALS, with a frequencyanging from 15% to 76% when compared to HC [22] and has beenescribed not only in HC [4], but also in other neurological disor-ers, such as X-linked Charcot-Marie Tooth neuropathies, Krabbeisease, and adrenomyeloneuropathy [23–25]. Moreover, this find-

ng does not correlate with the clinical impairment [3].Cortical hypointensities located in the PCG in T2-weighted

mages have also been described in several studies in ALS [2,6,7,26],howing a correlation with the severity of the disease [5,6,15]. Thisodification, taken alone, is neither sensitive nor specific for ALS

22], due to its presence both in normal aging [19] and in other neu-ological disorders, like Alzheimer’s disease, Parkinson’s diseasend multiple infarctions [26,27]. The cause of this T2 shortenings still unclear. Several studies have suggested a pathological ironccumulation in motor cortex microglia [6,7], while other authorso not agree with this hypothesis [28].

MRS has been also proposed as a useful adjunct to conven-ional MR sequences, to detect UMN loss in ALS patients’ brains,esulting in a reduction of NAA. Several studies have detected aecreased NAA concentration in ALS, especially in PCG [8–11,29],hich appears to be correlated with the disease severity [9,30].

In the present study MRS imaging data were analyzed. Thispproach was chosen, compared to single-voxel MRS, as the shapend the extension of the motor cortex makes it suitable for a tech-ique that allows for convoluted sampling across the plane. Inddition, given the bilaterality of the disease, as multivoxel MRSllows to assess neuronal loss in different brain subregions simul-aneously, the possibility to average left and right values allowed aubstantial increase in the S/N ratio.

Recently, MRS alterations in cervical cord have been shown inresymptomatic SOD1-positive people, at risk for familial ALS [31],itnessing the potential of this approach to derive a biomarker forLS diagnosis and progression.

Combination of different MR techniques is a promising way toncrease confidence in ALS diagnosis. To our knowledge, this ishe first study that combines T2-weighted hypointensity of motorortex and MRS data, with the aim of evaluating their diagnosticfficiency, both alone and in combination, in a large group of ALSatients.

Few studies have assessed the combined use of conventional anddvanced MR measures in ALS diagnosis. Sarchielli et al. assessedoth PCG-hypo and NAA in the motor cortex [15], although the

pecificity of PCG-hypo was not tested. Instead, Charil and col-eagues reported a 100% sensitivity and specificity when combining

Radiology 84 (2015) 151–157

CST hyperintensity on FLAIR images and NAA/Crea ratio, althoughthese results were obtained in a group of only 11 patients [1].

We chose to combine PCG-hypo and PCG-NAA because, in ourdata sets, these two measures proved to be relatively independentand capable of discriminating the two groups. It is noteworthy thatthe lack of correlation between these two measures has been incon-stantly reported. Previous works in smaller patient groups haveeither shown [15] or failed to detect [5] significant differences inNAA in the motor cortex between patients with and without corti-cal hypointensities in the PCG.

As previous studies have linked PCG-hypo to iron accumulationmainly in the microglia [6], we speculate that this lack of correla-tion between PCG-hypo and PCG-NAA could be explained by thepresence of a different time course of PCG-NAA decrease (an earlyphenomenon, reflecting UMN loss or dysfunction), compared toiron accumulation (possibly related to inflammatory phenomenasecondary to microglial migration and activation), which leads todetectable PCG-hypo at a later time point.

To fully elucidate the temporal relationship between PCG-hypoand PCG-NAA over time, longitudinal studies with both techniquesshould be carried out.

The sensitivity and specificity of PCG-hypo and PCG-NAA, whichwas low when the two variables were taken alone, significantlyincreased when they were used in combination. Therefore, inpatients undergoing an MRI study for possible ALS, beside theassessment of T2-hypointensity in the motor cortex, the use of MRScan ameliorate sensitivity, at the expenses of a limited increase inscanning time.

In agreement with the high variability of CST-hyperintensityreported in the literature [22], we found no significant differencesbetween ALS patients and HC; this could be explained both by thelow specificity of this finding, known to be present also in nor-mal aging [20], and by the use of FSE-T2-weighted images, whosesensitivity has been reported to vary from 14% to 63%[1].

To reduce the reportedly high variability of conventional MRIfindings, probably due to the use of qualitative evaluation ofabsence/presence or grading of a specific sign based on subjectiveassessment [22], we chose a ROI-based semiquantitative approach,slightly more time-consuming, but more reproducible.

The use of ROI’s minimum value for the analysis of PCG-hypo,although in general noisier than mean values, permitted to avoidthe confounding effect of both CSF and subcortical white matterhyperintensities (e.g. gliosis) that could affect ROI mean values. Apossible future expansion of the current approach may be the useof a measure of the skewness of the distribution of the values of thevoxels in the ROI, which may provide a less noisy measure of thePCG hypointensity.

Previous studies showed a correlation between the severity ofthe disease and PCG-NAA [9,30]. Conversely, we did not find asignificant relationship between clinical data and this variable, pos-sibly due to methodological differences, such as the correction ofour values for age (as it correlates both with clinical data and MRIfindings), or the use of ALSFRS, which has been reported to be unre-lated to MRS data [15,32].

Finally, some limitations should be considered in the presentreport. First, the retrospective nature of this study prevented theuse of axial FLAIR images, which were not included in the standardworkup for ALS patients at our institution. FLAIR images may bemore sensitive in showing CST hyperintensity in ALS [3], althoughit is noteworthy that direct comparison of FSE-T2-weighted andFLAIR images has shown this to be at least partly compensated bya lower specificity of FLAIR sequence [5].

In particular, DTI analysis has demonstrated microstructuralalterations not only along the CST, specifically at the level of the

Page 7

nal of

p[ttdHdwmiM

5

Nttnd

C

si

R

p

A

t–

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

A. Cervo et al. / European Jour

osterior limb of the internal capsule, but also in other WM regions33,34], including the cervical spinal cord, with significant correla-ion with the clinical scores [13,35]. Moreover, using magnetizationransfer imaging, further microstructural modifications have beenescribed both in the precentral gyrus and in extramotor areas [36].owever, as also these techniques appear to lack an acceptableiagnostic accuracy when taken alone [37], it remains to be testedhether they are relatively independent from the more diffuse MRIeasures used in the present work, and to evaluate whether their

nclusion in the proposed diagnostic setup could further increaseRI accuracy in ALS diagnosis.

. Conclusions

In conclusion, our findings suggest that T2-hypointensity andAA decrease in the motor cortex, assessed by a semiquantita-

ive analysis, reflect two independent phenomena in ALS. Thesewo variables, which taken alone do not provide acceptable diag-ostic accuracy in ALS diagnosis, can be combined to improve theiagnostic accuracy of MRI in ALS.

onflict of interest

All authors do not have any financial and personal relation-hips with other people or organizations that could inappropriatelynfluence (bias) our work.

ole of the funding source

The funding source had no involvement in study design, inter-retation of data and in the writing of the article.

cknowledgement

This study was partly supported by the Italian Ministry of Educa-ion, University, and Research (MIUR) with a grant (PRIN 2010-20112010XE5L2R 001) to E.T.

eferences

[1] Charil A, Corbo M, Filippi M, et al. Structural and metabolic changes in the brainof patients with upper motor neuron disorders: a multiparametric MRI study.Amyotroph Lateral Scler 2009;10(5/6):269–79.

[2] Cheung G, Gawel MJ, Cooper PW, Farb RI, Ang LC, Gawal MJ. Amyotrophiclateral sclerosis: correlation of clinical and MR imaging findings. Radiology1995;194(1):263–70.

[3] Hecht MJ, Fellner F, Fellner C, Hilz MJ, Heuss D, Neundorfer B. MRI-FLAIRimages of the head show corticospinal tract alterations in ALS patients morefrequently than T2-, T1- and proton-density-weighted images. J Neurol Sci2001;186(1/2):37–44.

[4] Mirowitz S, Sartor K, Gado M, Torack R. Focal signal-intensity variations in theposterior internal capsule: normal MR findings and distinction from pathologicfindings. Radiology 1989;172(2):535–9.

[5] Bowen BC, Pattany PM, Bradley WG, et al. MR imaging and localized protonspectroscopy of the precentral gyrus in amyotrophic lateral sclerosis. AJNR AmJ Neuroradiol 2000;21(4):647–58.

[6] Kwan JY, Jeong SY, Van Gelderen P, et al. Iron accumulation in deep corticallayers accounts for MRI signal abnormalities in ALS: correlating 7 Tesla MRIand pathology. PLoS ONE 2012;7(4):e35241.

[7] Oba H, Araki T, Ohtomo K, et al. Amyotrophic lateral sclerosis: T2 shortening inmotor cortex at MR imaging. Radiology 1993;189(3):843–6.

[8] Gredal O, Rosenbaum S, Topp S, Karlsborg M, Strange P, Werdelin L. Quantifi-cation of brain metabolites in amyotrophic lateral sclerosis by localized protonmagnetic resonance spectroscopy. Neurology 1997;48(4):878–81.

[9] Mitsumoto H, Ulug AM, Pullman SL, et al. Quantitative objective mark-

ers for upper and lower motor neuron dysfunction in ALS. Neurology2007;68(17):1402–10.

10] Pioro EP, Antel JP, Cashman NR, Arnold DL. Detection of cortical neuron loss inmotor neuron disease by proton magnetic resonance spectroscopic imaging invivo. Neurology 1994;44(10):1933–8.

[

Radiology 84 (2015) 151–157 157

11] Kalra S, Hanstock CC, Martin WR, Allen PS, Johnston WS. Detection of cere-bral degeneration in amyotrophic lateral sclerosis using high-field magneticresonance spectroscopy. Arch Neurol 2006;63(8):1144–8.

12] Stagg CJ, Knight S, Talbot K, Jenkinson M, Maudsley AA, Turner MR. Whole-brainmagnetic resonance spectroscopic imaging measures are related to disabilityin ALS. Neurology 2013;80(7):610–5.

13] Turner MR, Kiernan MC, Leigh PN, Talbot K. Biomarkers in amyotrophic lateralsclerosis. Lancet Neurol 2009;8(1):94–109.

14] Bede P, Hardiman O. Lessons of ALS imaging: pitfalls and future directions – acritical review. Neuroimage Clin 2014;4:436–43.

15] Sarchielli P, Pelliccioli GP, Tarducci R, et al. Magnetic resonance imaging and1H-magnetic resonance spectroscopy in amyotrophic lateral sclerosis. Neuro-radiology 2001;43(3):189–97.

16] The Amyotrophic Lateral Sclerosis Functional Rating Scale. Assessment of activ-ities of daily living in patients with amyotrophic lateral sclerosis. The ALS CNTFtreatment study (ACTS) phase I-II Study Group. Arch Neurol 1996;53(2:):141–7.

17] Filippini N, Douaud G, Mackay CE, Knight S, Talbot K, Turner MR. Corpus cal-losum involvement is a consistent feature of amyotrophic lateral sclerosis.Neurology 2010;75(18):1645–52.

18] Charles HC, Lazeyras F, Krishnan KR, et al. Proton spectroscopy of humanbrain: effects of age and sex. Prog Neuropsychopharmacol Biol Psychiatry1994;18(6):995–1004.

19] Imon Y, Yamaguchi S, Katayama S, et al. A decrease in cerebral cortex inten-sity on T2-weighted with ageing images of normal subjects. Neuroradiology1998;40(2):76–80.

20] Ngai S, Tang YM, Du L, Stuckey S. Hyperintensity of the precentral gyralsubcortical white matter and hypointensity of the precentral gyrus onfluid-attenuated inversion recovery: variation with age and implicationsfor the diagnosis of amyotrophic lateral sclerosis. AJNR Am J Neuroradiol2007;28(2):250–4.

21] Kato S, Shaw P, Wood-Allum C, Leigh P, Shaw C. Amyotrophic lateral sclerosis.In: Dickson W, editor. Neurodegeneration: the molecular pathology of demen-tia and movement disorders. Basel, Switzerland: ISN Neuropath Press; 2003. p.350–71.

22] Agosta F, Chio A, Cosottini M, et al. The present and the future of neuroimagingin amyotrophic lateral sclerosis. AJNR Am J Neuroradiol 2010;31(10):1769–77.

23] Kassubek J, Bretschneider V, Sperfeld AD. Corticospinal tract MRI hyper-intensity in X-linked Charcot-Marie-Tooth Disease. J Clin Neurosci2005;12(5):588–9.

24] Satoh JI, Tokumoto H, Kurohara K, et al. Adult-onset Krabbe disease withhomozygous T1853C mutation in the galactocerebrosidase gene: unusual MRIfindings of corticospinal tract demyelination. Neurology 1997;49(5):1392–9.

25] Teriitehau C, Adamsbaum C, Merzoug V, Kalifa G, Tourbah A, Aubourg P. Sub-tle brain abnormalities in adrenomyeloneuropathy. J Radiol 2007;88(7–8 Pt1):957–61.

26] Waragai M. MRI and clinical features in amyotrophic lateral sclerosis. Neuro-radiology 1997;39(12):847–51.

27] Imon Y, Yamaguchi S, Yamamura Y, et al. Low intensity areas observed onT2-weighted magnetic resonance imaging of the cerebral cortex in variousneurological diseases. J Neurol Sci 1995;134(Suppl.):27–32.

28] Hecht MJ, Fellner C, Schmid A, Neundorfer B, Fellner FA. Cortical T2 signalshortening in amyotrophic lateral sclerosis is not due to iron deposits. Neu-roradiology 2005;47(11):805–8.

29] Rule RR, Suhy J, Schuff N, Gelinas DF, Miller RG, Weiner MW. Reduced NAA inmotor and non-motor brain regions in amyotrophic lateral sclerosis: a cross-sectional and longitudinal study. Amyotroph Lateral Scler Other Motor NeuronDisord 2004;5(3):141–9.

30] Ellis CM, Simmons A, Andrews C, Dawson JM, Williams SC, Leigh PN. A pro-ton magnetic resonance spectroscopic study in ALS: correlation with clinicalfindings. Neurology 1998;51(4):1104–9.

31] Carew JD, Nair G, Andersen PM, et al. Presymptomatic spinal cord neu-rometabolic findings in SOD1-positive people at risk for familial ALS. Neurology2011;77(14):1370–5.

32] Han J, Ma L. Study of the features of proton MR spectroscopy ((1)H-MRS) onamyotrophic lateral sclerosis. J Magn Reson Imaging 2010;31(2):305–8.

33] Senda J, Ito M, Watanabe H, et al. Correlation between pyramidal tract degen-eration and widespread white matter involvement in amyotrophic lateralsclerosis: a study with tractography and diffusion-tensor imaging. AmyotrophLateral Scler 2009;10(5-6):288–94.

34] Cirillo M, Esposito F, Tedeschi G, et al. Widespread microstructural white matterinvolvement in amyotrophic lateral sclerosis: a whole-brain DTI study. AJNRAm J Neuroradiol 2012;33(6):1102–8.

35] Nair G, Carew JD, Usher S, Lu D, Hu XP, Benatar M. Diffusion tensor imagingreveals regional differences in the cervical spinal cord in amyotrophic lateralsclerosis. Neuroimage 2010;53(2):576–83.

36] Cosottini M, Pesaresi I, Piazza S, et al. Magnetization transfer imaging demon-strates a distributed pattern of microstructural changes of the cerebral

cortex in amyotrophic lateral sclerosis. AJNR Am J Neuroradiol 2011;32(4):704–8.

37] Foerster BR, Dwamena BA, Petrou M, Carlos RC, Callaghan BC, Pomper MG.Diagnostic accuracy using diffusion tensor imaging in the diagnosis of ALS: ameta-analysis. Acad Radiol 2012;19(9):1075–86.