Magnetic Resonance Imaging–BasedVolumetry Differentiates Idiopathic
Parkinson’s Syndrome from Multiple SystemAtrophy and Progressive Supranuclear Palsy
Jorg B. Schulz, MD,* Martin Skalej, MD,† Dirk Wedekind,* Andreas R. Luft,† Michael Abele, MD,*Karsten Voigt, MD,† Johannes Dichgans, MD,* and Thomas Klockgether, MD*
By using three-dimensional magnetic resonance imaging–based volumetry, we studied atrophy of the caudate nucleus,putamen, brainstem, and cerebellum in patients with idiopathic Parkinson’s syndrome (IPS, n 5 11), progressive su-pranuclear palsy (PSP, n 5 6), and multiple system atrophy with predominant parkinsonism (MSA-P, n 5 12) or ataxia(MSA-C, n 5 17). Patients were compared with a total of 46 controls, of whom 16 were age matched. Mean striatal,cerebellar, and brainstem volumes were normal in patients with IPS. We found significant reductions in mean striataland brainstem volumes in patients with MSA-P, MSA-C, and PSP, whereas patients with MSA-C and MSA-P alsoshowed a reduction in cerebellar volume. On an individual basis, volumes of structures in patients with MSA and PSPshowed an extensive overlap with the normal range with the exception of brainstem volumes in patients with MSA-C.Therefore, groups could not be discriminated on the basis of individual structure volumetry. Application of stepwisediscriminant analysis, however, allowed discrimination of all 12 patients with MSA-P, 15 of 17 patients with MSA-C,and 5 of 6 patients with PSP from the normal and IPS cohorts. However, patients with IPS could not be separated fromcontrols and patients with MSA-P could not be separated from patients with PSP. In conclusion, total intracranialvolume–normalized magnetic resonance imaging–based volumetric measurements provide a sensitive marker to discrim-inate typical and atypical parkinsonism.
Schulz JB, Skalej M, Wedekind D, Luft AR, Abele M, Voigt K, Dichgans J, Klockgether T. Magnetic resonanceimaging–based volumetry differentiates idiopathic Parkinson’s syndrome from multiple system
atrophy and progressive supranuclear palsy. Ann Neurol 1999;45:65–74
Progressive supranuclear palsy (PSP) and multiple sys-tem atrophy (MSA) are two neurodegenerative disor-ders often confused with idiopathic Parkinson’s syn-drome (IPS). Although the definite diagnosis of thesedisorders can only be made by neuropathological ex-amination, clinical diagnostic criteria have been devel-oped for a possible or probable diagnosis.1
Modern imaging methods such as magnetic reso-nance imaging (MRI), single-photon emission com-puted tomography (SPECT), and positron emissiontomography (PET) are increasingly used in patientswith IPS and related disorders to study their morpho-logical and functional characteristics.2–10 Using two-dimensional planimetric MRI evaluation, we reportedearlier cerebellar and brainstem atrophy in patientswith a parkinsonian (MSA-P) or cerebellar (MSA-C)subtype of MSA.6 However, this technique was limitedby the necessity to subjectively choose slices to measure
regions of interest, did not allow measuring of cerebel-lar and brainstem volumes, and could not be used toquantify atrophy of the basal ganglia. Therefore, we ap-plied an MRI-based three-dimensional (3D) techniquethat allows measuring of the volumes of the cerebel-lum, brainstem, putamen, and caudate nucleus in pa-tients with IPS, PSP, and MSA in a semiautomatedand highly reliable way.11,12
By using discriminant analysis, we show that each ofthese disorders is associated with a characteristic pat-tern of volume loss in the brain structures under study.Knowledge of these changes may be useful to improvethe clinical diagnosis of IPS and related disorders.
Subjects and MethodsControl SubjectsTo study determinants of areal volumes of the brain, we ex-amined 46 healthy controls (Table 1). They had no history
From the Departments of *Neurology and †Neuroradiology, Uni-versity of Tubingen, Tubingen, Germany.
Received Apr 3, 1998, and in revised form Jun 15. Accepted forpublication Jul 19, 1998.
Address correspondence to Dr Schulz, Department of Neurology,University of Tubingen, Hoppe-Seyler-Str 3, D-72076 Tubingen,Germany.
of a disease of the central nervous system, their neurologicalexaminations were normal, and routine evaluations of theirMRIs gave no pathological findings. Of these, a subset of16 age-matched and sex-matched healthy persons served ascontrols for comparison with patients with parkinsoniansyndromes.
PatientsWe studied 46 consecutive parkinsonian patients who ful-filled the diagnostic criteria for one of the following diseases:IPS, MSA-P, MSA-C, or PSP. A summary of their clinicalcharacteristics is given in Table 1.
IPS. IPS patients showed at least two of the cardinal symp-toms (akinesia, rigidity, and tremor) and at least two of thefollowing additional criteria: asymmetrical onset of disease,clear-cut response to levodopa and/or apomorphine, and ap-pearance of levodopa-associated responsive fluctuations anddyskinesias during the course of the disease.13,14 Other focalneurological signs or symptoms of autonomic failure wereabsent.
MSA. Twenty-nine patients fulfilled the following diagnos-tic criteria of Quinn15 for probable MSA: (1) sporadic adult-onset nonresponsive or poorly levodopa-responsive parkin-sonism and/or sporadic progressive cerebellar ataxia; (2)severe symptomatic autonomic failure with at least posturalsyncope or presyncope and/or pronounced urinary inconti-nence or retention not due to other causes; and (3) absenceof dementia according to criteria of the Diagnostic and Sta-tistic Manual of Mental Disorders, third edition, revised, gen-eralized tendon areflexia, or predominant down-gaze su-pranuclear palsy. Based on their clinical presentation,patients were subdivided into two groups—one with pre-dominant parkinsonism (MSA-P) and another with predom-inant ataxia (MSA-C).
PSP. The PSP group comprised 6 patients who fulfilled thefollowing mandatory inclusionary criteria of Litvan and col-leagues16 for the diagnosis of probable PSP: (1) graduallyprogressive parkinsonism with onset at age 40 years or later;(2) vertical supranuclear palsy with downward gaze abnor-malities; and (3) severe postural instability with unexplained
falls. Mandatory exclusionary criteria were history of enceph-alitis, hallucinations, early or prominent cerebellar symp-toms, prominent unexplained dysautonomia, unilateral dys-tonia, an alien hand syndrome, cortical dementia ofAlzheimer’s type, or focal lesions on neurological examina-tion or computed tomography or MRI scans.
MRIMRI SEQUENCES. MRI-based volumetry was performed byusing the method described by Luft and associates.12 In sum-mary, all measurements were performed on a Siemens Mag-netom Vision 1.5-T scanner (Siemens AG, Erlangen, Ger-many), using the standard head coil. Two MRI series wereacquired. A 3D Fourier transform (3D-FT) fast low-angleshot (FLASH) sequence producing isotropic T1-contrastedimage sets in high resolution (repetition time [TR] 5 15msec; echo time [TE] 5 5 msec; flip angle 5 30°; numberof excitations [NEX] 5 1; slice thickness 5 0.9 mm; pixelsize 5 0.9 3 0.9 mm) was scanned. A double-contrast2D-FT turbo spin echo (TurboSE) was acquired twice ininterleaved slice positions to obtain a gapless set of images(TR 5 5,800 msec; TE 5 15/75 msec; 2 NEX; slice thick-ness 5 2.0 mm; gap 5 2 mm; pixel size 5 0.9 3 0.9 mm).As the TurboSE sequence included two echoes, two imagesets of different contrast (TE 5 15 msec: proton densitycontrast; TE 5 75 msec: T2-weighted contrast) but equalslice position were obtained. For both sequences, slice thick-ness, orientation, and signal-to-noise ratio were optimized byusing phantom objects that simulated typical contrasts andforms of brain structures.11 FLASH images were used forcerebellar and brainstem volumetry. The total intracranialand basal ganglia volumes were measured by using TurboSEimages.
For postprocessing, data were transmitted to a graphicsworkstation (SGI Indigo R4400, Silicon Graphics, MountainView, CA). MRI-based volumetry was performed by usingthe method described in detail elsewhere.11,12,17 In brief,volumetry consisted of manual, landmark-defined preseg-mentation followed by automated region growing–based de-tailed segmentation and calculation of volume consideringpartial volume effects. Interactive presegmentation was neces-sary wherever the boundaries between structures were not con-trasted and therefore were not segmented by region growing.
Bladder dysfunction 5 retention or incontinence; orthostatic hypotension 5 orthostatic reduction in systolic blood flow by 30 mm Hg ormore; IPS 5 idiopathic Parkinson’s syndrome; MSA-C 5 multiple system atrophy with cerebellar subtype; MSA-P 5 multiple system atrophywith parkinsonism; PSP 5 progressive supranuclear palsy.
66 Annals of Neurology Vol 45 No 1 January 1999
DELINEATION OF STRUCTURES OF INTEREST. Manualpresegmentation of brainstem and cerebellum included defi-nition of the posterior, superior, and inferior borders byplanes adjusted for landmarks as shown in Figure 1. Afterinteractive definition of all boundaries, automated segmenta-tion was applied and the volumes were calculated by addingthe volumes of all segmented voxels.
The basal ganglia were measured by using the first andsecond echoes of the TurboSE dataset. The additional infor-mation from the second contrast allowed better identificationof each nucleus (multispectral analysis). After manual preseg-mentation, in which the putamen was separated from theinsular cortex, region growing was applied and the volume ofthe caudate nucleus and putamen was calculated as describedabove (Fig 2). By using this technique, we were not able toconsistently separate the pallidum from the capsula interna,especially in diseased patients with reduced basal ganglia vol-umes. Therefore, we do not report on pallidal volumes inthis study. The total intracranial volume (TICV) was esti-mated by using the proton density contrast (TE 5 15 msec)of the TurboSE sequence.
Statistical AnalysisThe reliability and precision of the volumetric technique wasreported previously.11,12 Data are expressed as mean 6 SEMvalues. For statistical analysis, JMP software (SAS Institute,Cary, NC) was used. Statistical analysis with a paired t testdid not reveal significant right/left differences of the caudatenucleus and putamen volumes. For further analysis, the re-spective right and left volumes of a structure were added toyield total volumes. Individual volumes were normalized forintersubject variation in head size by dividing structure vol-ume by the TICV of that particular subject. Associations be-tween area volumes, TICV, age, and sex in normal subjectswere evaluated by using stepwise regression. Statistical anal-ysis was performed by analysis of variance followed by Dun-nett’s post hoc test to compare group means.18 To discrim-inate between groups, we used stepwise linear discriminantanalysis.
ResultsNinety-two subjects are included in this report—46healthy controls and 46 consecutive patients with asporadic parkinsonian syndrome (see Table 1). Fromthe control group we extracted a subgroup of 16 age-matched and sex-matched individuals for statisticalcomparisons of group means. Age differences betweengroups were not significantly different by analysis ofvariance (F4,57 5 2.12). Disease duration of patientswith IPS was significantly longer than that of patientswith MSA-P, MSA-C, and PSP.
Control SubjectsThe structure volumes of men were generally largerthan those of women (Fig 3, left column). Further-more, TICV in men was significantly larger than thatof women (1,543 6 29 cm3 vs 1,349 6 32 cm3, p ,0.0001). Associations between structure volumes,TICV, age, and sex in controls were evaluated by using
stepwise regression (Table 2). The results allow to pre-dict the volumes of brain structures examined depend-ing on TICV, age, and sex. The stepwise regressionrevealed that TICV most significantly contributed tothe variance of control data. The volumes of the cere-bellum, caudate nucleus, and putamen were negativelycorrelated with age. When calculated separately formen and women, only the age-associated decrease inputamen volume was significant in both sexes. Themean decline of putamen volume with age was 40mm3/yr.
Patients with Parkinsonian SyndromesAs TICV and age contributed significantly to the vari-ance of control data, we selected a subgroup of 16 age-matched control subjects and calculated the individualnormalized structural volumes as percentages of TICV.These were used for further statistical analysis. All nor-malized structure volumes were significantly differentby analysis of variance between the patient and controlgroups (Table 3).
Mean absolute brainstem volume in the subset ofage-matched control subjects was 32.8 6 1.2 cm3.Compared with controls, normalized brainstem vol-umes were significantly reduced in patients withMSA-C, MSA-P, and PSP, but not in patients withIPS (Fig 4A). There was almost no overlap of the in-dividual normalized brainstem data of MSA-C patientsand controls. Although the mean of normalized brain-stem volumes of MSA-P and PSP patients was signifi-cantly different from controls and IPS patients, therewas substantial overlap between data of those groups.
The mean absolute cerebellar volume in the subsetof age-matched control subjects was 128.6 6 4.6 cm3.Compared with control subjects, normalized cerebellarvolumes were significantly reduced in patients withMSA-C and MSA-P but not in patients with IPS andPSP (see Fig 4B). However, there was substantial over-lap of individual cerebellar data of MSA-C and MSA-Ppatients with both controls and IPS patients.
The mean caudate volume in normal subjects was5.94 6 0.32 cm3. Normalized caudate data of PSP andMSA-P patients were significantly reduced comparedwith control subjects and IPS patients (see Fig 4C).Again, there was substantial overlap between the indi-vidual data of groups.
The mean absolute putamen volume in the subset ofage-matched control subjects was 7.72 6 0.43 cm3.Compared with controls and IPS patients, normalizedputamen data of PSP and MSA-P patients were signif-icantly reduced (see Fig 4D). However, there again wassubstantial overlap between individual MSA-P and IPSdata that did not allow a straightforward classification ofpatients according to their individual putamen volumes.
Linear regression analysis failed to reveal a significantinverse correlation between the disease durations of
Schulz et al: MRI-Based Volumetry in Parkinsonian Syndromes 67
68 Annals of Neurology Vol 45 No 1 January 1999
IPS, MSA-P, MSA-C, PSP, and normalized structurevolumes of respective brain areas.
Discrimination Between Control Subjects andPatients with IPS, MSA-P, MSA-C, and PSPBy using stepwise linear discriminant analysis (includ-ing normalized volumes of brainstem, cerebellum, cau-date nucleus, putamen, age, and sex as independentvariables) to predict the clinical diagnosis, the onlyvariables that appeared in the final model were TICV-normalized volumes of brainstem, caudate nucleus, pu-tamen, and cerebellum (see Table 3; p , 0.003, foreach term). Age and sex did not add more significantinformation to the final model. The sensitivity of thismathematical model, using a linear function includingnormalized volumes of brainstem, caudate nucleus, pu-tamen, and cerebellum as variables to distinguish pa-tients with a non-IPS parkinsonian syndrome from pa-tients with IPS and control subjects, was assessed bycalculating probability scores. In general, patients withMSA-C, MSA-P, and PSP were well separated frompatients with IPS and control subjects (Fig 5). Espe-cially, no IPS patient was classified as having MSA-Pand no MSA-P patient was classified as having IPS(Table 4). Two patients with MSA-C were diagnosedas control subjects. Both patients had a disease dura-tion of less than 1 year. Further, patients with MSA-Pwere well separated from patients with MSA-C. IPSpatients could not be discriminated from control sub-jects. The overall correct classification was 65%.
DiscussionIn the present study, including patients with the clin-ical probable diagnosis of a parkinsonian syndrome andage-matched control subjects, we have demonstratedthat MRI-based volumetry of the brainstem, cerebel-
lum, caudate nucleus, and putamen is useful in dis-criminating between IPS and other parkinsonian syn-dromes. As the diagnostic classification of MSA-P andIPS is the major difficulty for the clinician,19 thesemeasurements may prove to be helpful in classifyingpatients with parkinsonian syndromes.
ImagingWe earlier established and reported on volumetric mea-surements of the brainstem and cerebellum in normalindividuals by using MRI.11,12 Here we applied thesetechniques to patients with parkinsonian syndromes.We also performed volumetric measurements of thecaudate nucleus and putamen. Basal ganglia volumetrymost commonly uses manual segmentation of a fewrepresentative sections.20–22 We applied automated re-gion growing–based segmentation with minimal inter-active and standardized steps (manual presegmentationusing landmarks), which can be considered more pre-cise and reliable.23
Control SubjectsThe absolute volumes of the cerebellum, caudate nu-cleus, and putamen and the intracranial volume mea-sured in our study are in agreement with results pub-lished in the literature.22,24,25 Decrease of volume wasnegatively correlated with age in the cerebellum, cau-date nucleus, and putamen. This is consistent withobservations of age-related atrophy in other imagingor autopsy studies.26–32 The decrease of the cerebellarvolume was significantly more pronounced in menthan in women (see Fig 3, right column). The age-dependent volume decrease of brain structures may inpart explain the age dependency of neurodegenerativediseases. Further, our results allow to predict individualvolumes of the caudate nucleus, putamen, brainstem,
Fig 1. Region growing–based segmentation of the cerebellum and brainstem. On the midsagittal (top panels) and the parasagittal(bottom panels) section on the left side, a successfully segmented cerebellum is shown. The four images on the right side display asegmented brainstem from medial (top middle) to lateral (bottom right) position. The brainstem is presegmented by using fourplanes (black) that are orthogonal to the sagittal plane. Each plane is adjusted for two landmarks (superior: mamillary body, poste-rior commissure, displaced downward for one-third of the height of the midbrain; inferior: posterior rim of the foramen magnum,parallel to the superior plane; posterior: posterior commissure, obex, displaced backward as far as the inferior colliculus; the anteriorplane was required by the software but is not needed to separate the brainstem from the cerebrospinal fluid. Therefore, anteriordelineation was done by region growing). Subsequently, the region growing algorithm is able to define the exact outline of thebrainstem in three dimensions. The segmented brainstem is then subtracted from the original images, resulting in a blackened area(images on the left). Subtraction avoided overlap between the volumes of brainstem and cerebellum. (Sequence type: three-dimensional, fast low-angle shot (FLASH); TE/TR 5 5 msec/15 msec; slice thickness 5 0.9 mm; resolution 5 0.9 3 0.9 mm2.)
Š
Fig 2. Region growing–based segmentation of caudate nucleus and putamen. The left part of each image shows the segmented struc-ture at a higher slice position than the right part. A two-dimensional histogram (bottom right) is used to specify the pixel intensitiesof the caudate nucleus and putamen. In this histogram, the intensity values of two magnetic resonance imaging sequences (protondensity weighted and T2 weighted) are plotted. A structure is defined by an area in the histogram corresponding to its proton den-sity and T2 intensities. This allows a better specification of the properties of a structure and thus ameliorates the segmentation. Theboundary between the putamen and insula was not segmented automatically but had to be redrawn by hand (top right). (Sequencetype: turbo spin echo (TurboSE); TE/TR 5 15 msec/5,800 msec; slice thickness 5 2 mm; resolution 5 0.9 3 0.9 mm2.)
Š
Schulz et al: MRI-Based Volumetry in Parkinsonian Syndromes 69
Fig 3. Correlation of unnormal-ized structure volumes and totalintracranial volume (TICV)-normalized volumes. (Left) Totalvolumes of the brainstem, cere-bellum, caudate nucleus, andputamen for males (‘) and fe-males (▫). (Right) Individualvolumes were normalized bydividing structure volume bythe TICV.
70 Annals of Neurology Vol 45 No 1 January 1999
and cerebellum depending on TICV, age, and sex (seeTable 2).
Patients with Parkinsonian SyndromesMean striatal, cerebellar, and brainstem volumes werenormal in patients with IPS. This shows that there isno atrophy in the caudate nucleus and putamen afternigrostriatal degeneration. In contrast, loss of striatalneurons is an integral part of the degenerative processin MSA in addition to degeneration and gliosis in thesubstantia nigra, Purkinje cells, pontine nuclei, inferiorolives, and intermediolateral cell column.33–35 Weshow that volumetric measurements of the caudate nu-cleus and putamen in MSA-P, MSA-C, and PSP pa-tients are reduced compared with those of control sub-jects and IPS patients. These results are consistent withearlier reports, showing reduced dopamine D2 receptordensity, using [123I]iodobenzamide SPECT6–9 orraclopride PET.5 Loss of these receptors reflects the de-generation of striatal neurons. [18F]FluorodeoxyglucosePET showed a decrease of glucose metabolism in stri-atum and pallidum.36–38 Furthermore, magnetic reso-nance spectroscopy in patients with MSA-P but notwith IPS showed a decrease of N-acetylaspartate-to-creatine ratios in the putamen and pallidum comparedwith the control group, reflecting neuronal loss in thesestructures.10 In our earlier study,6 only 63% of MSA
patients showed a reduced [123I]iodobenzamide bind-ing in the basal ganglia of more than 2 SD below thenormal values. These results closely correspond to ourpresent results of reduced basal ganglia volumes insome but not all patients with MSA (see Fig 4). Inagreement with the clinical presentation of MSA pa-tients, the decrease of caudate nucleus and putamenvolumes was more advanced in MSA-P than in MSA-Cpatients.
The decrease of brainstem, caudate nucleus, and pu-tamen volumes reflects the widespread pathology inPSP patients. Further, there also was a trend of de-creased cerebellar volume. Pathologically, PSP is char-acterized by neuronal loss, gliosis, and neurofibrillarytangle formation in the striatum, pallidum, subtha-lamic nucleus, substantia nigra, locus ceruleus, tectum,tegmentum, basis pontis, certain cranial nerve nuclei,and the dentate nucleus.39,40 Within the basal ganglia,the pallidosubthalamic complex and the substantianigra are particularly affected. As our volumetric tech-nique did not reproducibly segment the pallidum, wecould not evaluate changes in pallidal volume.
Discriminating Between Control Subjects andPatients with Parkinsonian SyndromesMRI-based volumetry has recently been used to exam-ine age-related changes in the brain,29–32 hippocampalor medial temporal atrophy or the rate of atrophy inAlzheimer’s disease,32,41,42 or basal ganglia volumes inHuntington’s disease.20–22 In this study, we have dem-onstrated that MRI volumetric measurements of thebrainstem, cerebellum, caudate nucleus, and putamenare useful in discriminating between sporadic parkinso-nian syndromes.
A linear discriminant function revealed that thebrainstem, caudate nucleus, putamen, and cerebellumprovided additive discriminatory power. With a math-ematical model including these four structures it waspossible to discriminate between groups. The normal-ized volumes of the caudate nucleus and putamen pro-vided the best information for discrimination betweengroups. We did not observe any mismatch betweenMSA-P and IPS, the separation that is clinically the
Table 2. Relationship Between Volume of Brain Structures, TICV, Age, and Sex in Control Subjects
Normalized Structure Volume Intercept TICV ( t ) Age (a) Sex (g)
The values in the table may be used to predict individual volumes depending on total intracranial volume (TICV), age (a), and sex (g). Thepredicted normalized structure volume Y of an individual (i) may be calculated as follows: Y(i) 5 Intercept 1 t p TICV(i) 1 a p Age(i) 1g p Sex(i) (sex 5 21, if male; sex 5 11, if female).ap , 0.05; bp , 0.01; and cp , 0.001 (other associations did not reach the level of significance).
Table 3. Analysis of Variance and Discriminant Analysis
F values of the univariate analysis of variance indicate mean differ-ences for each variable. The total intracranial volume–normalizedvolumes of each anatomical brain structure were used for analysis(ap , 0.002; bp , 0.0001). Relative weight of the variables indi-cates their contribution to the discrimination among groups as de-tected by stepwise discriminant analysis.
— 5 variable not included in the stepwise discriminant analysis.
Schulz et al: MRI-Based Volumetry in Parkinsonian Syndromes 71
Fig 4. Comparison of normalized structure volumes between parkinsonian syndromes and controls. Sizes of anatomical structureswere determined by magnetic resonance imaging–based volumetry and normalized for total intracranial volume (TICV). *p ,0.05, compared with control subjects and patients with multiple system atrophy with cerebellar subtype (MSA-C); **p , 0.01,compared with control subjects; 1p , 0.05, compared with control subjects, patients with multiple system atrophy with parkinson-ism (MSA-P), and patients with progressive supranuclear palsy (PSP); 11p , 0.01, compared with control subjects (analysis ofvariance followed by Dunnett’s post hoc test).
major diagnostic difficulty. The ability of discriminantfunction analysis to distinguish the pattern of atrophyin patients with IPS versus patients with MSA and PSPreflects the differing underlying pathologies in thebrainstem, cerebellum, caudate nucleus, and putamenof patients with these conditions. Because IPS patientsnormally do not exhibit neuronal loss or atrophy inposterior fossa structures, the caudate nucleus, or theputamen, the model did not discriminate between pa-tients with IPS and control subjects. We were not ableto separate patients with PSP from patients withMSA-P. As we only measured the total brainstem, wedid not differentiate between mesencephalic atrophy,typical for patients with PSP, and atrophy of pons andolivary nuclei, which may occur in patients withMSA-P.
Discriminant function analysis has previously beenused to distinguish IPS, PSP, and MSA on the basis ofthe pattern of the caudate and putamen [18F]dopa up-take.2 Compared with our study, the advantage of thismethod was the separation of IPS patients from controlsubjects, which was possible in 100% of normal sub-
jects due to the abnormalities in the nigrostriatal sys-tem. However, 36% of patients with IPS were classi-fied as having MSA or PSP. Patients with MSA wereassigned to MSA, Parkinson’s disease, and PSP groupswith roughly equal frequency. Compared with these re-sults, our method was superior in separating IPS fromother syndromes with parkinsonism.
Ideally, imaging criteria should complement clinicalcriteria that help to differentiate parkinsonian syn-dromes. By the neuropathological definition of MSAand PSP, these criteria should include imaging abnor-malities within both the basal ganglia and the ponto-cerebellar system in various degrees. The present studyshows that single criteria, such as the volume of a sin-gle structure, are not sufficient because they lack bothsensitivity and specificity. However, multivariate statis-tical analysis of different structures involved may en-able to discriminate between groups. Therefore, MRI-based volumetry of basal ganglia and posterior fossastructures may prove to be useful for discriminationbetween parkinsonian syndromes and may help to es-tablish the correct diagnoses of diseased patients duringtheir lifetimes.
References1. Koller WC, Montgomery EB. Issues in the early diagnosis of
kinson’s disease, multiple system atrophy, and Steele-Richardson-Olszewski syndrome: discriminant analysis of stria-tal 18F-dopa PET data. J Neurol Neurosurg Psychiatry 1994;57:278–284
3. Brooks DJ. Functional imaging of movement disorders. In:Marsden CD, Fahn S, eds. Movement disorders 3. Oxford:Butterworth-Heinemann, 1994:65–87
4. Leenders KL, Frackowiak RS, Lees AJ. Steele-Richardson-Olszewski syndrome. Brain energy metabolism, blood flow andfluorodopa uptake measured by positron emission tomography.Brain 1988;111:615–630
Table 4. Diagnostic Classification MatrixBased on MRI Volumetry
Summary of the results of the discriminant analysis. Shown is theclassification of subjects due to their MRI-based volumetry with re-spect to their clinical diagnosis. Rows represent the clinical diag-noses and columns the diagnoses predicted by MRI volumetry.Boldface indicates correct classifications.
MRI 5 magnetic resonance imaging; IPS 5 idiopathic Parkinson’ssyndrome; MSA-P 5 multiple system atrophy with parkinsonism;MSA-C 5 multiple system atrophy with cerebellar subtype; PSP 5progressive supranuclear palsy.
Fig 5. Centroid plot on the first two canonical variables. Thecentroid points for each group appear with a circle correspond-ing to the 95% confidence region. The mean values of theidiopathic Parkinson’s syndrome (IPS), multiple system atrophywith cerebellar subtype (MSA-C), multiple system atrophy withparkinsonism (MSA-P), and progressive supranuclear palsy(PSP) groups are well discriminated from each other. Biplotrays show the directions or the original variables measured(normalized volumes for the brainstem, cerebellum, caudatenucleus, and putamen).
Schulz et al: MRI-Based Volumetry in Parkinsonian Syndromes 73
5. Brooks DJ, Ibanez V, Sawle GV, et al. Striatal D2 receptorstatus in patients with Parkinson’s disease, striatonigral degen-eration, and progressive supranuclear palsy, measured with 11C-raclopride and positron emission tomography. Ann Neurol1992;31:184–192
6. Schulz JB, Klockgether T, Petersen D, et al. Multiple systematrophy: natural history, MRI morphology, and dopamine re-ceptor imaging with 123IBZM-SPECT. J Neurol NeurosurgPsychiatry 1994;57:1047–1056
7. van Royen E, Verhoeff NFLG, Speelman JD, et al. Multiplesystem atrophy and progressive supranuclear palsy: diminishedstriatal D2 dopamine receptor activity demonstrated by 123I-IBZM single photon emission computed tomography. ArchNeurol 1993;50:513–516
8. Schwarz J, Tatsch K, Arnold G, et al. 123I-Iodobenzamide-SPECT predicts dopaminergic responsiveness in patients withde novo parkinsonism. Neurology 1992;42:556–561
9. Brucke T, Podreka I, Angelberger P, et al. Dopamine D2 re-ceptor imaging with SPECT: studies in different neuropsychi-atric disorders. J Cereb Blood Flow Metab 1991;11:220–228
10. Davie CA, Wenning GK, Barker GJ, et al. Differentiation ofmultiple system atrophy from idiopathic Parkinson’s disease us-ing proton magnetic resonance spectroscopy. Ann Neurol 1995;37:204–210
11. Luft AR, Skalej M, Welte D, Kolb R. Reliability and exactnessof MRI-based volumetry: a phantom study. J Magn Reson Im-aging 1996;6:700–704
12. Luft AR, Skalej M, Welte D, et al. A semi-automated, threedimensional technique allowing quantification of cerebellar vol-ume and its substructure using MRI. Magn Reson Med 1998;40:143–151
13. Gibb WRG, Lees AJ. The relevance of the Lewy body to thepathogenesis of idiopathic Parkinson’s disease. J Neurol Neuro-surg Psychiatry 1988;51:745–752
14. Ward CD, Gibb WR. Research diagnostic criteria for Parkin-son’s disease. Adv Neurol 1990;53:245–249
15. Quinn N. Multiple system atrophy—the nature of the beast.J Neurol Neurosurg Psychiatry 1989;52(Suppl):78–89
16. Litvan I, Agid Y, Jankovic J, et al. Accuracy of clinical criteriafor the diagnosis of progressive supranuclear palsy (Steele-Richardson-Olszewski syndrome). Neurology 1996;46:922–930
17. Welte D, UK, Skalej M, Kolb R, et al. Quantifizierung desFehlers von Volumenbestimmungen in der Kernspintomogra-phie mit Hilfe von Agarose-Phantomen. Z Med Phys 1995;5:208–215
18. Dunnett CW. A multiple comparison procedure for comparingseveral treatments with a control. J Am Stat Assoc 1955;50:1096–1121
19. Hughes AJ, Daniel SE, Kilford L, Lees AJ. Accuracy of clinicaldiagnosis of idiopathic Parkinson’s disease: a clinico-patho-logical study of 100 cases. J Neurol Neurosurg Psychiatry 1992;55:181–184
20. Aylward EH, Brandt J, Codori AM, et al. Reduced basal gan-glia volume associated with the gene for Huntington’s disease inasymptomatic at-risk persons. Neurology 1994;44:823–828
21. Aylward EH, Codori A-M, Barta PE, et al. Basal ganglia vol-ume and proximity to onset in presymptomatic Huntingtondisease. Arch Neurol 1996;53:1293–1296
22. Aylward EH, Li Q, Stine OC, et al. Longitudinal change inbasal ganglia volume in patients with Huntington’s disease.Neurology 1997;48:394–399
23. Hillman GR, Kent TA, Kaye A, et al. Measurement of braincompartment volumes in MR using voxel composition calcula-tions. J Comp Assist Tomogr 1991;15:640–646
24. Luef G, Burtscher J, Kremser C, et al. Magnetic resonancevolumetry of the cerebellum in epileptic patients after phenyt-oin overdosages. Eur Neurol 1996;36:273–277
25. Aylward EH, Habbak R, Warren AC, et al. Cerebellar volumein adults with Down syndrome. Arch Neurol 1997;54:209–212
26. Raz N. Neuroanatomy of the aging brain observed in vivo. In:Bigler ED, ed. Neuroimaging II: clinical applications. NewYork: Plenum Press, 1996:153–182
27. Raz N, Dupuis JH, Briggs SD, et al. Differential effects of ageand sex on the cerebellar hemispheres and the vermis: a pro-spective MR study. Am J Neuroradiol 1998;19:65–71
28. Ellis RS. Norms of structural in human cerebellum from birthto old age. J Comp Neurol 1920;32:1–33
29. DeCarli C, Murphy DGM, Gillette JA, et al. Lack of age-related differences in temporal lobe volume of very healthyadults. Am J Neuroradiol 1994;15:689–696
30. Blatter DD, Bigler ED, Gale SD, et al. Quantitative volumetricanalysis of brain MR: normative database spanning 5 decades oflife. Am J Neuroradiol 1993;16:241–251
31. Matsumae M, Kikinis R, Morocz IA, et al. Age-related changesin intracranial compartment volumes in normal adults assessedby magnetic resonance imaging. J Neurosurg 1996;84:982–991
32. Jack CR Jr, Petersen RC, Xu YC, et al. Medial temporal atro-phy on MRI in normal aging and very mild Alzheimer’s disease.Neurology 1997;49:786–794
33. Graham JG, Oppenheimer DR. Orthostatic hypotension andnicotine sensitivity in a case of multiple system atrophy. J Neu-rol Neurosurg Psychiatry 1969;32:28–34
34. Fearnley JM, Lees AJ. Striatonigral degeneration. A clinicopath-ological study. Brain 1990;113:1823–1842
35. Quinn N. Multiple system atrophy. In: Marsden CD, Fahn S,eds. Movement disorders 3. Oxford: Butterworth-Heinemann,1994:262–281
36. Otsuka M, Ichiya Y, Kuwabara Y, et al. Glucose metabolism inthe cortical and subcortical brain structures in multiple systematrophy and Parkinson’s disease: a positron emission tomo-graphic study. J Neurol Sci 1996;144:77–83
37. Gilman S, St Laurent RT, Koeppe RA, et al. A comparison ofcerebral blood flow and glucose metabolism in olivopontocer-ebellar atrophy using PET. Neurology 1995;45:1345–1352
38. Perani D, Bressi S, Testa D, et al. Clinical/metabolic correla-tions in multiple system atrophy. Arch Neurol 1995;52:179–185
39. Daniel SE, de Bruin VMS, Lees AJ. The clinical and patholog-ical spectrum of Steele-Richardson-Olszewski syndrome (pro-gressive supranuclear palsy): a reappraisal. Brain 1995;118:759–770
40. Probst A, Luginbuhl M, Langui D, et al. Pathology of the stri-atum in progressive supranuclear palsy: abnormal tau proteinsin astrocytes and cholinergic interneurons. Neurodegeneration1993;2:183–193
41. Fox NC, Warrington EK, Freeborough PA, et al. Presymptom-atic hippocampal atrophy in Alzheimer’s disease. Brain 1996;119:2001–2007
42. Fox NC, Freeborough PA, Rossor MN. Visualisation and quan-tification of rates of atrophy in Alzheimer’s disease. Lancet1996;348:94–97
![[Left and right ventricular volumetry and ejection fraction with MRI: segmentation criteria and interobserver reproducibility]](https://static.documents.pub/doc/80x56/634c8a72155e68b74b05cf15/left-and-right-ventricular-volumetry-and-ejection-fraction-with-mri-segmentation.jpg)
