Journal of Neuroimaging Vol 11 No 2 April 2001Bakshi et al: MRI Rating of Brain Atrophy in MS

Regional Brain Atrophy IsAssociated With Physical Disabilityin Multiple Sclerosis:Semiquantitative MagneticResonance Imaging andRelationship to Clinical Findings

Rohit Bakshi, MD

Ralph H. B. Benedict, PhD

Robert A. Bermel, BA

Lawrence Jacobs, MD

A B S T R A C T

Objective. Brain atrophy may occur early in the course of multi-ple sclerosis (MS) and may be associated with disability. Brainmagnetic resonance imaging (MRI) of 114 MS patients (group A)were analyzed for regional atrophy (vs age-/gender-matchedcontrols) and T1 and T2 lesions using 4-point rating systems.Thirty-five separate patients (group B) were analyzed for corticalatrophy (ordinal scale), third ventricular width, and total T2

hyperintense lesion volume (computer assisted). In group A,regression modeling indicated that inferior frontal atrophy (P =.0003) and T2 lesions in the pons (P = .02) predicted physical dis-ability (Expanded Disability Status Scale [EDSS] score). Sec-ondary progressive (SP) versus relapsing patients were pre-dicted by inferior parietal (P = .002), superior parietal (P = .006),temporal (P = .008), inferior frontal (P = .01), superior frontal (P =.01), cerebellum (P = .01), occipital (P = .01), and midbrain (P =.02) atrophy. SP patients were also predicted by total atrophy(P = .01) and third ventricular enlargement (P = .03) but not T1 orT2 lesions. In group B, the regression model predicting EDSSscore included only superior frontal atrophy (r = 0.515, P = .002).Mean kappa coefficients of ordinal ratings were 0.9(intraobserver) and 0.8 (interobserver). Ordinal ratings corre-lated well with quantitative assessments. The authors concludethat brain atrophy is closely associated with physical disabilityand clinical course in MS patients and can be appreciated usinga semiquantitative MRI regional rating system.

Key words: Multiple sclerosis, magnetic resonance imaging(MRI), brain, atrophy, disability.

Bakshi R, Benedict RHB, Bermel RA, Jacobs L.Regional brain atrophy is associated with physical

disability in multiple sclerosis: semiquantitative magneticresonance imaging and relationship to clinical findings.

J Neuroimaging 2001;11:129–136.

Characteristic hyperintense lesions on T2-weightedimages in the brain1 and spinal cord2 are useful in diagnos-ing multiple sclerosis (MS) and can provide a longitudinalassessment of pathologic alterations in treatment trials.3

However, bright T2 lesions are nonspecific in defining MSpathologic changes extending from mild reversiblechanges to severe irreversible white matter damage.4 Fur-thermore, the absence of T2 hyperintensities does not nec-essarily equate with the absence of white matter pathol-ogy; areas that appear normal may contain microscopicdisease.5 Contrast enhancement on T1-weighted imagesindicates inflammation and blood-brain barrier disrup-tion, but it is transient, variable, and may not predict per-manent injury or disease progression.6 Both hyperintenseand enhancing lesion loads have shown poor correlationswith physical disability in MS and provide incompleteassessments of therapies.7

The destructive potential of the MS disease process hasbeen recognized by widespread axonal transection8 andmacroscopic brain atrophy9–12 that are common in MSbrains and can be seen early in the disease process. Thesefindings are in line with positron emission tomographydata showing widespread cortical and subcortical brainhypometabolism in MS patients13 that correlated withcognitive disability14 and atrophy.15

Methods for accurate measurement of brain atrophy inMS are currently being developed using computerizedtechniques.9–11 These approaches require equipment andspecially trained personnel. In the clinical setting of rou-tine care of MS patients, it would be desirable to be able torapidly identify and stratify brain atrophy. The goal of thecurrent study was to test an ordinal (visual) rating systemof regional brain atrophy using magnetic resonance imag-ing (MRI) film hard copies for its reliability and correla-

Copyright © 2001 by the American Society of Neuroimaging 129

Received September 13, 2000, and in revised form Octo-ber 30, 2000. Accepted for publication November 3,2000.

From Imaging Services, Kaleida Health (RB); and the De-partment of Neurology (RB, RHBB, RAB, LJ) and the De-partments of Psychiatry and Psychology (RHBB), StateUniversity of New York at Buffalo.

Address correspondence to Dr Bakshi, NeuroscienceCenter, E-2, Buffalo General Hospital, 100 High Street,Buffalo, NY 14203. E-mail: rbakshi@buffalo.edu.

tion with clinical parameters in MS patients. We also com-pared data obtained by the ordinal rating system ofatrophy to ordinal ratings of T1 and T2 lesions and moreestablished quantitative (computer-assisted) methods ofmeasuring MRI disease burden including T2 lesionvolume16 and central atrophy (third ventricular width)12

for their ability to predict clinical findings.

Subjects and Methods

Group A

We studied a consecutive series of clinically definite17 MSpatients aged 18 to 60 who were referred for MRI during a2-year period from a community-based university-affili-ated MS clinic. After excluding those with poor- qualityMR images, other major illnesses, or primary progressivedisease course,18 there were 114 patients included in thisgroup (89 women, mean age = 43 years [SD = ±9 years],range 19-60 years, mean disease duration = 9.6 years [SD= ±9 years], range 0.5-43 years). MS was relapsing-remit-ting (RR, n = 80) or secondary progressive (SP, n = 34).18

The patients underwent neurologic examination by anMS specialist to derive an Expanded Disability StatusScale (EDSS) rating of physical disability.7 EDSS scoreswere 0 to 9 (mode = 3.0). Age-matched normal controlsubjects (n = 114) in whom MS was excluded clinicallywere consecutively identified from our institution’s nor-mative MRI database.19 They were scanned by the sameMRI protocol as the MS patients; their original MRI re-ports were normal. The controls’ scans were reviewed toensure that they were normal.

Axial brain MRI was performed at 1.5-T usingT1-weighted (T1WI), fast spin echo T2-weighted (T2WI),proton-density-weighted (PDWI), and fast T2-weightedfluid-attenuated inversion recovery (FLAIR) 5-mm inter-leaved (contiguous) images. Fifty-four patients were alsostudied by postcontrast T1WI. One experienced observermasked to clinical details rated the MRIs using a previ-ously described method.20–22 All axial slices were ana-lyzed for lesions and atrophy in superior and inferior fron-tal lobes, superior and inferior parietal lobes, occipitaland temporal lobes, cerebellum, midbrain, and pons.Standard fissures (eg, central, lateral, parieto-occipitalsulci) were used to separate the lobes. The third ventricledemarcated superior versus inferior frontal/parietallobes. The severity of hyperintense lesions on T2WI andhypointense lesions on noncontrast T1WI was assessed.We used an ordinal rating system to assess severity oflesions in each region based on the percentage of involve-ment (total lesion volume as a proportion of the entire sizeof that region). Hyperintense T2 plaques were ordinally

rated by severity according to percentage of involvementof the entire region as normal, grade 1 (mild; < 25%),grade 2 (moderate; 25%-50%), or grade 3 (severe; >50%).The brightness of lesions compared with normal graymatter on T2WI was confirmed by PDWI and FLAIRimages. Hypointense T1 plaques were ordinally rated asnormal, grade 1 (mild; <10%), grade 2 (moderate; 10%-25%), or grade 3 (severe; >25%). Total T1 and T2 lesionloads were the means of all ordinal ratings. The degree ofcontrast-enhancement was assessed by counting the num-ber of enhancing lesions.

Atrophy was assessed on noncontrast T1WI as previ-ously detailed20–22 by ordinally rating the enlargement ofsubarachnoid spaces in the above-mentioned regions andrating the enlargement of the ventricular cavities (lateralventricles [body, frontal, temporal, and occipital horns]and third and fourth ventricles) and was rated against arandomly chosen age-/gender-matched control scan asnormal, grade 1 (mild; <10%), grade 2 (moderate; 10%-25%), or grade 3 (severe; >25%), according to the percent-age of parenchymal volume loss or percentage of ventric-ular size increase. Total atrophy for a patient was the meanof all regional ordinal ratings. The rating of atrophy, T1,T2, and enhancing lesions was easy to implement. All rat-ings were determined by assessment of hard copies ofMRI films and usually required 10 to 15 minutes perpatient. A single control subject’s scan was placed next toa patient’s scan for the atrophy ratings (same gender, agewithin 2 years). Thus, cases were matched to patients on aone-to-one basis (once a control subject was used, he orshe was removed from the pool).

To assess the validity of the ordinal ratings of lesionsand atrophy, an individual who was masked to clinicaldetails and ordinal MRI ratings performed a quantitativeanalysis in a subset of 20 randomly chosen patients using aproprietary Unix-based workstation (EasyVision, 1997release, Philips Medical Systems, Best, the Netherlands).We tested the validity of our ordinal rating method ofdetermining regional atrophy by measuring third ventric-ular width using a previously detailed method.12 We testedthe validity of our ordinal rating of hyperintense plaquesby manually tracing hyperintense lesions on FLAIRimages. The superior frontal, inferior frontal, superiorparietal, inferior parietal, temporal, occipital, cerebellum,pons, and midbrain regions were assessed separately asdefined above. Each plaque volume was determined bymultiplying its area by slice thickness (5 mm). Lesion vol-ume for each region was the sum of all lesions from thatregion. The regions were collapsed across both hemi-spheres. Total brain T2 lesion volume was the sum of alllesions seen on all slices.

130 Journal of Neuroimaging Vol 11 No 2 April 2001

Group B

Thirty-five clinically definite or laboratory-supported MSpatients17 were recruited from a tertiary care comprehen-sive MS clinic to compare the MRI ordinal rating systemwith quantitative computer-assisted measures of MRI le-sions and atrophy. Patients underwent MRI and neuro-logic examination to derive an EDSS score. Twenty-onepatients had SP MS, and 14 had RR MS. Their EDSSscores were 1.0 to 8.5 (mode = 3.0), and none had otherneurologic diseases, incidences of substance abuse, pri-mary psychiatric disorders, or clinical relapses (orcorticosteroids) within the prior 3 weeks. Brain MRI wasperformed at 1.5 T. The protocol included a 5 mm thick(2.5 mm interslice gap) axial T1-weighted series with anaxial 5-mm (2.5 mm interslice gap) dual-echo, conven-tional spin echo series with intermediate and T2 weight-ing. Individuals masked to the clinical details analyzedthe MRIs. The images were processed on a PC using soft-ware written in Image Display Language (Image DisplayLanguage, Boulder, CO). Hyperintense T2-lesion volumewas obtained by an experienced technician using a man-ual tracing procedure.16 The total brain T2 lesion volume(obtained by multiplying lesion area by the sum of slicethickness and slice gap)16 and third ventricular width12

were obtained. Cortical atrophy was assessed in superiorfrontal, inferior frontal, superior parietal, inferior parietal,temporal, and occipital regions by an experienced ob-server masked to clinical details using the same ordinalrating method described for group A.

Statistical Methods

Associations between MRI ratings and clinical data(EDSS score; RR vs SP subtype) were analyzed by linearor logistic regression modeling. To limit the ratio of pre-dictor variables to participants, only MRI variables with aSpearman rank correlation P < .01 with EDSS or Mann-Whitney U Test P < .01 comparing RR vs SP patientswere tested in the univariate models. Those variables sig-nificant at P < .05 in the univariate models were includedin the forward selection models. Regression models weredesigned to determine whether visual ratings of atrophy,T1 lesion loads, and T2 lesion loads predict EDSS score orSP vs RR clinical course (group A patients) and whethervisual ratings of atrophy account for more variance in theprediction of EDSS score or SP vs RR clinical course thancomputer-assisted measures of T2 lesion load and thirdventricular width (group B patients). Reported P valuesfor regression coefficients in the final models were not ad-justed for the number of models considered because ofthe exploratory nature of the analysis.

Results

Reliability of Ratings

For the ordinal rating method, 27 randomly chosen pa-tients from group A had repeat ratings of lesions and atro-phy without knowledge of the original data. This includedrepeat ratings by the same experienced observer at least 1month after the initial scoring and independent ratings bya second trained observer. The intra- and interobserverreliabilities of the T2 lesion, T1 lesion, and atrophy ordinalrating method was moderate to excellent (mean kappa =0.9, range 0.8-1.0; mean kappa = 0.8, range 0.6-1.0). Ex-amples of kappa scores for key MRI ratings are given inTable 1. In a randomly chosen subset of 10 MS patientsfrom group A, the intraobserver mean coefficient of varia-tion for the computer-assisted quantitation of total brainT2 lesion volume was 1.2%, and for third ventricularwidth, 5.7%. The intraobserver mean coefficient of varia-tion for the measurement of third ventricular width waspreviously shown to be 3% for the computer-assistedmethod used in group B.12

Validity of Rating Scales

Spearman rank correlations between ordinal and quanti-tative determinations of third ventricular size showed ex-cellent agreement (r = 0.7, P = .002). Spearman rankcorrelations between ordinal and volumetric assessmentsof bright T2 lesions showed excellent agreement fornearly all of the ratings: superior frontal (r = 0.7, P =.002), inferior frontal (r = 0.57, P = .002), superior parietal(r = 0.83, P = .0003), inferior parietal (r = 0.7, P = .002),

Bakshi et al: MRI Rating of Brain Atrophy in MS 131

Table 1. Reliability of Key Regional Ordinal (visual) Measure-ments in a Subset of Patients (n = 27)

Intraobserver Interobserver(kappa (kappa

coefficient) coefficient)

Superior frontal atrophy 0.85 0.75Inferior frontal atrophy 0.80 0.69Superior parietal atrophy 0.95 0.85Inferior parietal atrophy 0.80 0.77Temporal atrophy 0.91 0.82Occipital atrophy 0.88 0.73Lateral ventricular bodiesenlargement 1.0 0.77

Third ventricular enlargement 1.0 0.74Fourth ventricular enlargement 1.0 0.73Mean of all ratings(atrophy, T1, T2) 0.9 0.8

Range of all ratings(atrophy, T1, T2) 0.8-1.0 0.6-1.0

temporal (r = 0.72, P = .001), midbrain (r = 0.44, P =.0021), pons (r = 0.61, P = .0002), cerebellum (r = 0.974,P < .0001), and total lesion volume (r = 0.728, P = .001).The only exception was that the agreement for occipitallesions was less robust (r = 0.3, P = .06). The capability tovalidate cortical atrophy ordinal ratings with computerMRI analysis is being developed in the NeuroscienceCenter at Buffalo General Hospital.

Group A

Atrophy (cerebrospinal fluid and ventricular enlargementas compared with age- and gender-matched controls) was

common in RR and SP MS patients. Atrophy was wide-spread, involving cortical, central, and infratentorial re-gions. Frontal, parietal, temporal, and occipital corticalsulci were enlarged in MS patients. The superior frontalareas were most commonly involved and occipital sulcileast commonly involved. Atrophy also commonly af-fected the midbrain, pons, and cerebellum (midbrainmost commonly). All of the ventricular regions were com-monly enlarged in the MS patients (Tables 2 and 3). For il-lustrative purposes, patients are grouped into mild (0-3),moderate (3.5-6.0), and severe (>6.0) EDSS disability inFigure 1 and Tables 2 and 3. Spearman rank correlation

132 Journal of Neuroimaging Vol 11 No 2 April 2001

Table 2. Enlargement of Subarachnoid Spaces in Group A Multiple Sclerosis (MS) Patients (number and percentage of MS patientsper group)

EDSSa

Region All RR SP ≤ 3.0 3.5-6.0 ≥6.5and Grade (N = 114) (n = 80) (n = 34) (n = 66) (n = 37) (n = 11)

Superior frontalGrade 0 59 (52%) 44 (55%) 15 (44%) 38 (57%) 19 (52%) 2 (18%)Grade 1 28 (24%) 22 (28%) 6 (18%) 17 (26%) 9 (24%) 2 (18%)Grade 2 or 3 27 (24%) 14 (17%) 13 (38%) 11 (17%) 9 (24%) 7 (64%)

Inferior frontalGrade 0 76 (66%) 57 (71%) 19 (56%) 48 (73%) 24 (65%) 4 (36%)Grade 1 27 (24%) 21 (26%) 6 (18%) 15 (23%) 11 (30%) 1 (10%)Grade 2 or 3 11 (10%) 2 (3%) 9 (26%) 3 (4%) 2 (5%) 6 (54%)

Superior parietalGrade 0 65 (57%) 51 (64%) 14 (41%) 44 (66%) 18 (49%) 3 (27%)Grade 1 21 (18%) 14 (17%) 7 (21%) 9 (14%) 10 (27%) 2 (18%)Grade 2 or 3 28 (25%) 15 (19%) 13 (38%) 13 (20%) 9 (24%) 6 (55%)

Inferior parietalGrade 0 76 (66%) 61 (76%) 15 (44%) 48 (73%) 23 (62%) 5 (46%)Grade 1 27 (24%) 15 (19%) 12 (35%) 13 (20%) 11 (30%) 3 (27%)Grade 2 or 3 11 (10%) 4 (5%) 7 (21%) 5 (7%) 3 (8%) 3 (27%)

TemporalGrade 0 63 (55%) 48 (60%) 15 (44%) 42 (64%) 18 (48%) 3 (27%)Grade 1 24 (21%) 19 (24%) 5 (15%) 12 (18%) 11 (30%) 1 (9%)Grade 2 or 3 27 (24%) 13 (16%) 14 (41%) 12 (18%) 8 (22%) 7 (64%)

OccipitalGrade 0 76 (66%) 59 (74%) 17 (50%) 48 (73%) 23 (62%) 5 (46%)Grade 1 26 (23%) 16 (20%) 10 (29%) 15 (23%) 9 (24%) 2 (18%)Grade 2 or 3 12 (11%) 5 (6%) 7 (21%) 3 (4%) 5 (14%) 4 (36%)

MidbrainGrade 0 51 (45%) 42 (53%) 9 (26%) 33 (50%) 16 (43%) 2 (19%)Grade 1 34 (30%) 21 (26%) 13 (38%) 21 (32%) 10 (27%) 3 (27%)Grade 2 or 3 29 (25%) 17 (21%) 12 (35%) 12 (18%) 11 (30%) 6 (54%)

PonsGrade 0 66 (58%) 54 (68%) 12 (35%) 46 (70%) 17 (46%) 3 (28%)Grade 1 18 (16%) 8 (10%) 10 (29%) 8 (12%) 8 (22%) 2 (18%)Grade 2 or 3 30 (26%) 18 (22%) 12 (35%) 12 (18%) 12 (32%) 6 (54%)

CerebellumGrade 0 66 (58%) 52 (63%) 14 (41%) 45 (68%) 17 (46%) 4 (36%)Grade 1 23 (20%) 15 (19%) 8 (23%) 9 (14%) 11 (30%) 3 (27%)Grade 2 or 3 25 (22%) 13 (18%) 12 (36%) 12 (18%) 9 (24%) 4 (36%)

RR = relapsing-remitting MS patients, SP = secondary progressive MS patients.a. EDSS = Expanded Disability Status Scale, divided into categories for illustrative purposes only.

results between cortical atrophy ratings and physical dis-ability (EDSS) scores are shown in Table 4.

Forward selection linear regression modeling usingEDSS score as a dependent variable and all MRI vari-ables that were significant in univariate analyses indicatedthat inferior frontal atrophy (P = .0003) (Fig 1) and T2

lesions in the pons (P = .02) predicted physical disability(EDSS score). The association between inferior frontalatrophy and EDSS scores is expressed against severity ofdisability in Figure 1. In a second model using logisticregression, SP (vs RR) patients were predicted by atrophyin inferior parietal (P = .002), superior parietal (P = .006),temporal (P = .008), inferior frontal (P = .01), superiorfrontal (P = .01), cerebellum (P = .01), occipital (P = .01),and midbrain (P = .02) areas. SP patients were also pre-dicted by total atrophy (P = .01) and third ventricular atro-phy (P = .03) scores but not T1 or T2 lesions. Thus, atrophyshowed a stronger association with clinical parametersthan did T1 or T2 lesions.

The ordinal rating method detected atrophy early aswell as later in the disease process (in patients with onlymild or more severe physical disability) (Tables 2 and 3).Of the RR patients, 45% had superior frontal atrophy

Bakshi et al: MRI Rating of Brain Atrophy in MS 133

Table 3. Ventricular Enlargement in Group A Multiple Sclerosis (MS) Patients (number and percentage of MS patients per group)

EDSSa

Region All RR SP ≤ 3.0 3.5-6.0 ≥ 6.5and Grade (N = 114) (n = 80) (n = 34) (n = 66) (n = 37) (n = 11)

LV bodiesGrade 0 46 (40%) 36 (45%) 10 (29%) 30 (46%) 13 (35%) 3 (27%)Grade 1 26 (23%) 17 (21%) 9 (27%) 14 (21%) 11 (30%) 1 (9%)Grade 2 or 3 42 (37%) 27 (34%) 15 (44%) 22 (33%) 13 (35%) 7 (64%)

LV frontal hornsGrade 0 56 (49%) 43 (54%) 13 (38%) 38 (57%) 16 (43%) 2 (18%)Grade 1 24 (21%) 16 (20%) 8 (24%) 10 (15%) 10 (27%) 4 (36%)Grade 2 or 3 34 (30%) 21 (26%) 13 (38%) 18 (28%) 11 (30%) 5 (46%)

LV temporal hornsGrade 0 74 (65%) 55 (69%) 19 (56%) 47 (71%) 23 (63%) 4 (36%)Grade 1 30 (26%) 21 (26%) 9 (26%) 16 (24%) 12 (32%) 2 (18%)Grade 2 or 3 10 (9%) 4 (5%) 6 (18%) 3 (5%) 2 (5%) 5 (46%)

LV occipital hornsGrade 0 74 (65%) 55 (69%) 19 (55%) 43 (65%) 27 (74%) 4 (36%)Grade 1 24 (21%) 16 (20%) 8 (24%) 16 (24%) 5 (13%) 3 (28%)Grade 2 or 3 16 (14%) 9 (11%) 7 (21%) 7 (11%) 5 (13%) 4 (36%)

Third ventricleGrade 0 56 (50%) 42 (52%) 14 (41%) 36 (55%) 16 (43%) 4 (36%)Grade 1 29 (25%) 22 (28%) 7 (21%) 18 (27%) 10 (27%) 1 (9%)Grade 2 or 3 29 (25%) 16 (20%) 13 (38%) 12 (18%) 11 (30%) 6 (54%)

Fourth ventricleGrade 0 64 (56%) 49 (61%) 15 (44%) 41 (62%) 18 (48%) 5 (46%)Grade 1 33 (29%) 22 (28%) 11 (32%) 17 (26%) 13 (35%) 3 (27%)Grade 2 or 3 17 (15%) 9 (11%) 8 (24%) 8 (12%) 6 (16%) 3 (27%)

LV = lateral ventricles, RR = relapsing-remitting MS patients, SP = secondary progressive MS patients.a. EDSS = Expanded Disability Status Scale, divided into categories for illustrative purposes only.

Table 4. Spearman Rank Correlation of Magnetic ResonanceImaging (MRI) Findings and Physical Disability (Ex-panded Disability Status Scale [EDSS]) in MultipleSclerosis Patients

Group A Group B(n = 114) (n = 35)

MRI Variable r P Value r P Value

Total T2 lesion volume (cm3)a NP NP 0.44 .008Third ventricular width NP NP 0.51 .002Superior frontal atrophyb* 0.26 <.001 0.55 <.001Inferior frontal atrophyb** 0.28 <.0001 0.38 .02Superior parietal atrophyb 0.30 <.0001 0.23 .17Inferior parietal atrophyb 0.21 <.001 0.15 .39Temporal atrophyb 0.27 <.001 0.53 .001Occipital atrophyb 0.22 <.001 0.38 .02

NP = not performed.a. Computerized measurement.b. Ordinal (visual) rating.*Predictor of EDSS at P = .0003 in the regression model (group B).**Predictor of EDSS at P = .002 in the regression model (group A).

(≥grade 1) and 17% had moderate or severe atrophy(grade 2 or 3). Forty percent of patients with an EDSSscore of 3 or less had some degree of midbrain atrophy;18% had moderate or severe atrophy. Fifty-five percent ofRR patients had enlarged lateral ventricular bodies; in34%, it was moderate to severe. Forty-three percent of

mildly disabled patients had enlargement of the frontalhorns, which was moderate to severe in >25% of patients.

Group B

Ordinal measures of cortical atrophy (superior frontal, in-ferior frontal, temporal, and occipital) correlated signifi-

134 Journal of Neuroimaging Vol 11 No 2 April 2001

0

10

20

30

40

50

0-3.0 3.5-6.0 6.5-9.0

Physical disability (EDSS score)

Inferior frontal atrophy vs. EDSS

Atrophyscore

>0(%

ofpatien

ts)

Fig 1. Inferior frontal atrophy compared with Expanded Disability Status Scale (EDSS) scores in multiple sclerosis patientsfrom group A (n = 114). For illustrative purposes, patients are divided into mild, moderate, and severe physical disability EDSSranges. Note the occurrence of atrophy in the mildly disabled patients. The percentage of patients with an atrophy rating greaterthan 0 rises with higher EDSS ranges. Forward selection linear regression modeling using EDSS score as a dependent variableand all MRI variables that were significant in univariate analyses indicated that inferior frontal atrophy (P = .0003) and T2 lesions inthe pons (P = .02) predicted EDSS score.

0

10

20

30

40

0-3.0 3.5-6.0 6.5-9.0

Physical disability (EDSS score)

Atrophyscore

>2.0

(%ofpatien

ts)

Superior frontal atrophy vs. EDSS

Fig 2. Superior frontal atrophy compared with Expanded Disability Status Scale (EDSS) scores in multiple sclerosis patientsfrom group B (n = 35). For illustrative purposes, patients are divided into mild, moderate, and severe physical disability EDSSranges. Note the occurrence of atrophy in the mildly disabled patients. The percentage of patients with an atrophy rating greaterthan 2 rises with higher EDSS ranges.Regression modeling showed that of all the MRI variables, only superior frontal atrophy sig-nificantly predicted EDSS (P = .002).

cantly with EDSS score (Table 4, Fig 2). Total T2 lesionvolume and third ventricular width also correlated signifi-cantly with EDSS score (Table 4). However, regressionmodeling indicated that when all MRI predictors were ac-counted for, only superior frontal atrophy independentlypredicted EDSS score (Fig 2). Correlations between MRIand EDSS were as follows: total T2 lesion volume r = 0.44(P = .008), third ventricular width r = 0.51 (P = .002), supe-rior frontal atrophy r = 0.55 (P = .001), inferior frontal at-rophy r = 0.38 (P = .02), superior parietal atrophy r = 0.23(P = .17), inferior parietal atrophy r = 0.15 (P = .39), tempo-ral atrophy r = 0.53 (P = .001), and occipital atrophy r =0.38 (P = .02). As noted above, all of the MRI measureswere positively intercorrelated, and regression analysiswas pursued to determine which MRI variables moststrongly predicted EDSS score. The regression model re-tained only superior frontal atrophy.

Group differences between SP and RR patients werereliable for total T2 lesion volume (P = .02) and superiorfrontal (P = .003), inferior frontal (P = .002), temporal (P =.007), and occipital (P = .003) atrophy. In each case, moresevere atrophy was seen in the SP group. Trends wereapparent for third ventricle width (P = .06) and superiorparietal atrophy (P = .08). Discriminant function analysiswas used to control for multicollinearity. In the finalmodel, only occipital atrophy was retained (Wilks’slambda = 14.8, P = .001).

Discussion

The semiquantitative methods for measuring atrophyused in this study were found to be reliable, were associ-ated with physical disability and disease course, and wererecently shown to correlate significantly with depres-sion,22 impaired quality of life,21 cognitive dysfunction,23

and brain hypometabolism detected by positron emissiontomography in MS patients.15 The visual MRI rating waseasy to perform in the clinical setting, and data obtainedcould be used to quantitate brain volume loss in MS pa-tients. This regional method could reveal selective areasof atrophy that cannot be derived from general measure-ments such as the brain parenchymal fraction9 and otherglobal measures.10 Quantitative methods of assessing re-gional and global atrophy, as well as volumetric T1 and T2

lesion loads, require computer equipment, training, andpersonnel costs that limit their use in the clinical setting. Itwas reassuring to see that our ordinal rating of bright T2 le-sions and central atrophy (third ventricular enlargement)showed validity when compared with computer-assistedmeasurements. Thus, our relatively simple rating meth-ods can provide useful information about the MS diseaseprocess and can be used clinically.

Atrophy measured by the ordinal rating methodshowed a better relationship to clinical parameters thaneither T1 or T2 plaques. There is growing recognition ofthe destructive potential of MS, including axonaltransection, Wallerian degeneration, and atrophy in MSbrains, even early in the disease.8–10,12,24 Both depression22

and quality-of-life changes21 associated with MS wereassociated with atrophy but not T2 hyperintense lesions.Atrophy in MS was associated with physical9,10,12 and cog-nitive dysfunction23,25and brain hypometabolism15 andcould predict future levels of disability.10 These previousstudies combined with the current study indicate thatatrophy is a clinically relevant phenomenon in MS thatprovides unique information not given by measures offocal inflammation (T1 or T2 plaques). Little is knownabout the pathophysiology of brain atrophy such as therole of demyelination, inflammation, axonal loss, orWallerian degeneration. The closer coupling of atrophywith clinical findings than between plaques and clinicalfindings indicates that atrophy may not be simply a sec-ondary effect of focal inflammation in white matter. Atro-phy may result from involvement of both gray and whitematter including direct axonal or neuronal injury.24 Thus,it is possible that neuroprotective agents may comple-ment the therapeutic effects of immunomodulatoryagents in MS patients.8

Our study showed that atrophy commonly occurredearly in the disease course in MS patients. A significantproportion of MS patients with RR disease or mild physi-cal disability had some degree of atrophy, both centraland cortical, indicating a relatively widespread tissue lossor degeneration. Other investigators also reported atro-phy in patients with only mild or early clinical involve-ment.9,12,26 Widespread axonal transection occurs com-monly in MS patients, including patients with early ormild clinical involvement.8 In this study, we found promi-nence of cerebrospinal fluid spaces across a wide range ofMS patients that may indicate a subset of patients with aparticularly destructive disease process early in the clini-cal course. Our ordinal rating scale can be readily used toidentify patients who are developing early brain atrophy.However, our ordinal rating method is probably toocoarse for monitoring patients in clinical trials. We plan tofurther test the validity of this system against state-of-the-art computer-assisted quantitative methods for measuringregional cortical atrophy.27 It would be of interest to deter-mine whether our ordinal rating measurements of atro-phy correlate with those obtained by transcranial Dopp-ler ultrasound. A recent report showed that transcranialDoppler ultrasound can accurately measure enlargementof the third ventricle and frontal horns that correlate withphysical and cognitive disability in MS.28

Bakshi et al: MRI Rating of Brain Atrophy in MS 135

This work was supported by NIH/National Institute of NeurologicalDisorders and Stroke Grant No 1 K23 NS02210-01 to the first author. Wethank Jack H. Simon for performing the quantitative magnetic reso-nance imaging analysis of group B patients at the University of ColoradoHealth Sciences Center, Denver. We are grateful to Anna Davidson andRoger Priore of the State University of New York at Buffalo for statisticalconsultation.

References

1. Fazekas F, Barkhof F, Filippi M, et al. The contribution ofmagnetic resonance imaging to the diagnosis of multiplesclerosis. Neurology 1999;53:448–456.

2. Bakshi R, Kinkel PR, Mechtler LL, et al. Magnetic reso-nance imaging findings in 22 cases of myelitis: comparisonbetween patients with and without multiple sclerosis. Eur JNeurol 1998;5:35–48.

3. Filippi M, Horsfield MA, Ader HJ, et al. Guidelines for usingquantitative measures of brain magnetic resonance imagingabnormalities in monitoring the treatment of multiple scle-rosis. Ann Neurol 1998;43:499–506.

4. van Walderveen MAA, Kamphorst W, Scheltens P, et al.Histopathologic correlate of hypointense lesions onT1-weighted spin-echo MRI in multiple sclerosis. Neurology1998;50:1282–1288.

5. Loevner LA, Grossman RI, Cohen JA, Lexa FJ, Kessler D,Kolson DL. Microscopic disease in normal appearing whitematter on conventional MR images in patients with multiplesclerosis: assessment with magnetization-transfer measure-ments. Radiology 1995;196:511–515.

6. Kappos L, Moeri D, Radue EW, et al. Predictive value ofgadolinium-enhanced magnetic resonance imaging forrelapse rate and changes in disability or impairment in mul-tiple sclerosis: a meta-analysis. Lancet 1999;353:964–969.

7. Barkhof F. MRI in multiple sclerosis: correlation withexpanded disability status scale (EDSS). MS 1999;5:283–286.

8. Trapp BD, Peterson J, Ransohoff RM, Rudick R, Mort S, BoL. Axonal transection in the lesions of multiple sclerosis. NEngl J Med 1998;338:278–285.

9. Rudick RA, Fisher E, Lee JC, Simon J, Jacobs L. Use of thebrain parenchymal fraction to measure whole brain atrophyin relapsing-remitting MS. Neurology 1999;53:1698–1704.

10. Losseff NA, Wang L, Lai HM, et al. Progressive cerebralatrophy in multiple sclerosis: a serial MRI study. Brain1996;119:2009–2019.

11. Ge Y, Grossman RI, Udupa JK, et al. Brain atrophy inrelapsing-remitting multiple sclerosis and secondary pro-gressive multiple sclerosis: longitudinal quantitative analy-sis. Radiology 2000;214:665–670.

12. Simon JH, Jacobs LD, Campion MK, et al. A longitudinalstudy of brain atrophy in relapsing multiple sclerosis. Neurol-ogy 1999;53:139–148.

13. Bakshi R, Miletich RS, Kinkel PR, Emmet ML, Kinkel WR.High-resolution fluorodeoxyglucose positron emissiontomography shows both global and regional cerebral

hypometabolism in multiple sclerosis. J Neuroimaging 1998;8:228–234.

14. Blinkenberg M, Rune K, Jensen CV, et al. Cortical cerebralmetabolism correlates with MRI lesion load and cognitivedysfunction in MS. Neurology 2000;54:558–564.

15. Bakshi R, Miletich RS, Kinkel PR, Janardhan V, KinkelWR. MRI hemispheric atrophy is associated withFDG-PET hypometabolism and MRI white matter lesionsin multiple sclerosis. Neurology 1999;52(suppl 2):A361.Abstract.

16. Simon JH, Jacobs LD, Campion M, et al. Magnetic reso-nance studies of intramuscular interferon beta-1a for relaps-ing multiple sclerosis. Ann Neurol 1998;43:79–87.

17. Poser CM, Paty DW, Scheinberg L, et al. New diagnostic cri-teria for multiple sclerosis: guidelines for research protocols.Ann Neurol 1983;13:227–231.

18. Lublin FD, Reingold SC. Defining the clinical course ofmultiple sclerosis: results of an international survey. Neurol-ogy 1996;46:907–911.

19. Bakshi R, Kamran S, Kinkel PR, et al. Fluid-attenuatedinversion-recovery MR imaging in acute and subacute cere-bral intraventricular hemorrhage. Am J Neuroradiol 1999;20:629–636.

20. Bakshi R, Miletich RS, Henschel K, et al. Fatigue in multiplesclerosis: cross-sectional correlation with brain MRI find-ings in 71 patients. Neurology 1999;53:1151–1153.

21. Janardhan V, Bakshi R. Quality of life and its relationship tobrain lesions and atrophy on magnetic resonance images in60 patients with multiple sclerosis . Arch Neurol2000;57:1485–1491.

22. Bakshi R, Czarnecki D, Shaikh ZA, et al. Brain MRI lesionsand atrophy are related to depression in multiple sclerosis.NeuroReport 2000;11:1153–1158.

23. Benedict RHB, Bakshi R, Simon JH, Priore R, Miller C,Munschauer F. Frontal cortex atrophy predicts cognitiveimpairment in multiple sclerosis. J Neuropsychiatr ClinNeurosci (in press).

24. Simon JH. From enhancing lesions to atrophy in relapsingMS. J Neuroimmunol 1999;98:7–15.

25. Rao SM, Leo GJ, Haughton VM, St Aubin-Faubert P,Bernardin L. Correlation of magnetic resonance imagingwith neuropsychological testing in multiple sclerosis. Neurol-ogy 1989;39:161–166.

26. Brex PA, Jenkins R, Fox NC, et al. Detection of ventricularenlargement in patients at the earliest clinical stage of MS.Neurology 2000;54:1689–1691.

27. Goldstein JM, Goodman JM, Seidman LJ, et al. Corticalabnormalities in schizophrenia identified by structural mag-netic resonance imaging. Arch Gen Psychiatry 1999;56:537–547.

28. Berg D, Mäurer M, Warmuth-Metz M, Rieckmann P,Becker G. The correlation between ventricular diametermeasured by transcranial sonography and clinical disabilityand cognitive dysfunction in patients with multiple sclerosis.Arch Neurol 2000;57:1289–1292.

136 Journal of Neuroimaging Vol 11 No 2 April 2001