Abstract
Objective: To evaluate the use of routine MRI in differentiating between patients with progressive supranuclear palsy (PSP), multiple system atrophy (MSA), corticobasal degeneration (CBD) and control subjects.
Methods: Two neuroradiologists rated blindly and independently axial T2-weighted and proton density MR images of 54 patients with MSA, 35 patients with PSP, 5 patients with CBD, and 44 control subjects.
Results: More than 70% of patients with PSP and more than 80% of patients with cerebellar predominant MSA could be classified correctly with 0.5-T or 1.5-T scans, and no patient in these groups was misclassified. In the remaining patients an unequivocal differentiation could not be made. However, only approximately 50% of patients with parkinsonism-predominant MSA could be classified correctly, and 19% of them (all of whom had had 0.5-T scans) were misclassified.
Conclusions: Characteristic findings on routine MRI, either 1.5 T or 0.5 T, can contribute to the identification of MSA and PSP. However, in a minority of patients no unequivocal diagnosis can be made using MRI findings alone.
The clinical differentiation of atypical parkinsonian syndromes—such as multiple system atrophy (MSA), progressive supranuclear palsy (PSP), and corticobasal degeneration (CBD)—from idiopathic Parkinson’s disease (IPD) and between each other is often difficult, leading to misdiagnosis even up to the time of death.1,2
We examined the value of routine MRI in differentiating patients with MSA from patients with IPD and from control subjects.3 The findings characteristically differentiating MSA from IPD patients and control subjects were: supratentorially, a hyperintense rim at the putaminal edge, putaminal hyperintensity (on 0.5-T scans only), and putaminal atrophy; and infratentorially, atrophy and signal change, corresponding to degeneration of pontine nuclei, transverse pontine fibers, Purkinje cells, and cell loss in the inferior olives. The majority of MSA patients showed these abnormalities. However, in a minority of patients with predominantly parkinsonian features (MSA-P) with none of these specific findings, MRI alone was not helpful diagnostically, even though a clinical diagnosis of probable MSA could be made. MR images in IPD patients did not differ in any of the investigated findings from those in control subjects. In PSP, MRI has often been considered to be of little help in making a diagnosis.4 Nevertheless, a number of findings suggestive of PSP—such as midbrain atrophy with enlargement of the third ventricle4 and pontine tegmental atrophy, signal increase in the midbrain, as well as frontal and temporal lobe atrophy later on5—have all been described. Occasionally, signal increase in the inferior olives was also seen.6 On the other hand, MRI in CBD can show cortical atrophy and hypointensity of putamen and globus pallidus, and asymmetry of global atrophy, when sought specifically.7 In a recent study comparing MR images of PSP and CBD patients, midbrain atrophy was found to be highly specific for PSP, and asymmetric frontoparietal atrophy for CBD.8 We studied the specificity and sensitivity of these findings in a direct and blinded evaluation of MR images in MSA compared with PSP and CBD patients.
Methods.
Patients.
We examined MR images of 35 patients with PSP, in four of whom the diagnosis was confirmed subsequently at autopsy. Twenty-three of these patients fulfilled the National Institute of Neurological Disorders and Stroke criteria9 for probable PSP, and eight patients met the criteria for possible PSP because they did not have falls within the first year of symptom onset. However, all of these patients fulfilled other proposed criteria for PSP.10,11 A total of 54 patients had clinically probable MSA,12 two cases of which were proved subsequently pathologically, and 30 of whom had MSA-P and 24 MSA-C (with predominantly cerebellar features). Five patients had CBD (autopsy confirmed in four patients) according to criteria derived from clinicopathologic correlation.13 A total of 44 control subjects had no clinical suspicion of parkinsonism. None of the diagnostic criteria for PSP, MSA, or CBD incorporate or are influenced by MRI findings, except when they point to an alternative diagnosis. Mean age was not different between the groups, but disease duration was shorter in the PSP group than in the MSA and CBD groups (both, p < 0.05). Table 1 provides the demographic data. In addition, we evaluated the MR images of nine patients with atypical parkinsonism in whom a probable clinical diagnosis could not be made according to the previously mentioned criteria.
Patient characteristics
MRI evaluation.
All patients had axial and one midsagittal T2-weighted and proton-density scans. A total of 23 PSP patients, 31 MSA patients, 3 CBD patients, 30 control subjects, and 3 patients with an unclear diagnosis were examined on a 0.5-T Vectra scanner (General Electric Medical Systems, Milwaukee, WI), and 12 PSP patients, 23 MSA patients, 2 CBD patients, 14 control subjects, and 6 patients with an unclear diagnosis were examined on a 1.5-T Signa scanner (General Electric Medical Systems, Milwaukee, WI). On the 0.5-T scanner, repetition time (TR) was 2,400 to 3,800 msec, echo time (TE) was 23 to 35 msec and 90 to 92 msec, slice thickness was 6 mm, and the interslice gap was 0.2 to 2 mm. On the 1.5-T scanner, TR was 2,400 to 4,000 msec, TE was 20 to 39 msec and 90 to 100 msec, slice thickness was 5 mm, and the interslice gap was 1.5 to 2 mm. Two experienced neuroradiologists rated all the scans independently and were blind to the diagnosis or clinical features except for age of the patient. The abnormal features evaluated were the presence of global atrophy, frontal or temporal lobe atrophy exceeding the global atrophy, a hyperintense lateral margination of the putamen (compared with cortical signal), putaminal atrophy, putaminal hyper- or hypointense signal relative to that of the globus pallidus, atrophy or abnormal signal in the globus pallidus, thinning or smudging of the substantia nigra, and abnormality of the red or subthalamic nucleus (STN). Other midbrain abnormalities evaluated were atrophy and signal change of tectum or tegmentum, and an anteroposterior midbrain diameter of less than 17 mm on axial scan. The infratentorial items evaluated were the presence of atrophy or signal change in the pons (diffuse or in the shape of a “hot-cross bun”3), middle cerebellar peduncles, cerebellum, medulla and inferior olives, and the width of the third and fourth ventricles. All paired abnormalities were evaluated for each side. Visual rating (other than measuring the midbrain diameter) was preferred to volumetric measurements because these were considered invalid in this retrospective study with varying slice positions, and moreover are not in routine use.
Statistical comparison was performed with the chi-square test and Fisher’s exact test. Interrater reliability was calculated with Cohen’s kappa (κ < 0.20 [slight agreement] to 0.81 to 1.00 [excellent agreement]. Comparison of means was performed with the t-test for independent samples. Statistical comparison of each item between all four groups was performed separately and combined for 0.5-T and 1.5-T scans, and a finding was defined as abnormal if at least one rater found an abnormality.
Results.
Interrater reliability was excellent (κ = 0.81 to 1.0) for atrophy and signal increase of middle cerebellar peduncles, pontine atrophy, and the “hot-cross bun sign.” It was slight for diffuse signal changes in the pons and medulla (both seen in two patients by only one rater), which were therefore excluded from additional analysis. Signal increase in globus pallidus (κ = 0.56) and putamen (κ = 0.66) were seen exclusively on 0.5-T scans , and signal decrease of globus pallidus and putamen was seen almost exclusively on 1.5-T scans (κ = 0.5 and κ = 0.46). Abnormalities of the STN were not seen, and hyperintensity of the red nucleus was seen only on 0.5-T scans. For the remaining findings the interrater reliability was good to fair (κ = 0.41 to 0.8), and similar for 0.5-T and 1.5-T scans.
Abnormal findings that occurred significantly more frequently in MSA and PSP patients than control subjects are shown in tables 2 and 3⇓. The most frequently seen abnormalities in PSP were signal increase and atrophy of the midbrain, thinning or smudging of the substantia nigra, atrophy of the putamen, atrophy and signal increase of globus pallidus (on 0.5-T scans), and atrophy of the red nucleus. Supratentorial atrophy was seen in the temporal and frontal lobes, and infratentorial atrophy was seen in the cerebellum, pons, and middle cerebellar peduncles. Infratentorial signal increase or atrophy of other infratentorial structures, hyperintense putaminal rim, or putaminal hyperintensity were seen rarely, and were seen only by one rater on scans otherwise showing the previously mentioned findings typical for PSP. Typical for MSA-P were a hyperintense putaminal rim, putaminal atrophy and hyperintensity, atrophy and signal decrease of the globus pallidus (mainly on 1.5-T scans), thinning or smudging of the substantia nigra, and infratentorial signal increase and atrophy. However, midbrain atrophy and signal increase and dilatation of the third ventricle was also seen. On the other hand, no patient had temporal or frontal lobe atrophy, a midbrain diameter less than 17 mm, or signal increase of the globus pallidus. Scans of patients with MSA-C all showed infratentorial atrophy and signal change. The majority also had some atrophy of the midbrain, but the midbrain diameter was smaller than 17 mm in only two of them. Putaminal atrophy was found in 50%, and a hyperintense putaminal rim in a third, and substantia nigra thinning or smudging in more than half of all MSA patients. All CBD patients had global atrophy, but there was no marked asymmetry. Most patients in this group had thinning or smudging of the substantia nigra, signal change in the globus pallidus, and dilatation of the third ventricle, and some had midbrain atrophy (marked only in one patient) or signal increase, and infratentorial atrophy. However, no scan showed infratentorial signal increase, temporal or frontal lobe atrophy exceeding the global atrophy, a putaminal hyperintense rim or hyperintensity, or globus pallidus atrophy.
Abnormalities differentiating progressive supranuclear palsy from multiple system atrophy
Abnormalities differentiating multiple system atrophy from progressive supranuclear palsy
Abnormalities discriminating between PSP, MSA, and CBD.
Factors that differentiated PSP from MSA significantly were a midbrain diameter of less than 17 mm, dilatation of the third ventricle and signal increase of the midbrain (figure), frontal and temporal lobe atrophy, signal increase in the globus pallidus (on 0.5-T scans), and atrophy of the red nucleus. Signal increase of the red nucleus was seen exclusively in PSP and only on 0.5-T scans, but this was not significant, probably due to small numbers (see table 2). Abnormalities of the STN, which is pathologically involved in PSP, were not seen, but we only evaluated axial scans, on which this nucleus is not well visualized. The features that were seen significantly more frequently in MSA-P than in PSP patients were the presence of signal increase in cerebellum, middle cerebellar peduncles and pons (hot-cross bun sign), dilatation of the fourth ventricle, atrophy of the dentate nucleus, signal decrease in the globus pallidus (mainly on 1.5-T scans), and a hyperintense rim or hyperintensity of the whole putamen (the latter only on 0.5-T scans). MSA-C differed significantly from PSP in all infratentorial abnormalities (signal increase and atrophy). Moderate midbrain atrophy (with a midbrain diameter of more than 17 mm), atrophy of globus pallidus or putamen, putaminal hypointensity, and smudging of the substantia nigra did not help to differentiate between these conditions. The only finding seen more frequently in CBD than in MSA-P patients was global atrophy (100% versus 36%, p < 0.05) and hyperintensity of the globus pallidus on 0.5-T scans (p < 0.05). Hyperintensity of the globus pallidus was also more frequent than in PSP patients (p < 0.05). No other finding was significantly different between CBD and PSP or MSA patients. This was most likely due to the small number of patients with this rare disorder.
Figure. Moderate midbrain atrophy and hyperintensity in a patient with progressive supranuclear palsy on a 0.5-T scan.
The sensitivity (proportion of patients with a disease who have the abnormal finding), specificity (proportion of patients who do not have the disease who do not have the abnormality), and positive predictive value (likelihood of a person with the abnormal finding to have the disease) for the abnormalities are presented in tables 2 and 3⇑. Using the abnormalities that differentiated MSA and PSP with a high positive predictive value and good specificity as criteria for classification (table 4), a correct classification could be made in the great majority of PSP and MSA-C patients with either 1.5-T or 0.5-T scans (table 5). No patient in these groups was misclassified. However, a substantial minority could not be classified unequivocally (nine patients [26%] with PSP [two definite, four probable, and three possible] and four patients [17%] with MSA-C), because either none of the characteristic abnormalities was found or abnormalities characteristic for both PSP and MSA were seen. Moreover, only half of the patients classified clinically as having MSA-P could be classified correctly by MRI findings, and 19% were classified incorrectly because of the presence of at least one finding characteristic for PSP. We found no abnormalities specific for CBD. Overall, 1.5-T scans were slightly more sensitive than 0.5-T scans because only one patient with possible PSP (8%) had no abnormalities on 1.5-T scans, whereas one patient with possible PSP (4%) and four patients with MSA (13%) had no abnormalities on 0.5-T scans.
Criteria for the differentiation of progressive supranuclear palsy, MSA-P, and MSA-C on routine MRI in a patient with atypical parkinsonism
Correct classification using the proposed MRI criteria to differentiate PSP and MSA patients
Clinical correlation.
Patients with parkinsonism in the MSA group had basal ganglia abnormalities more frequently than those without parkinsonism (p < 0.01), and patients with cerebellar dysfunction more often had infratentorial abnormalities (p < 0.05) than those without. In the PSP group, those who were later proved pathologically to have PSP, or fulfilled criteria for probable PSP,9 had midbrain abnormalities more frequently than those who only fulfilled possible criteria (p < 0.05). Disease duration was longer in PSP patients with midbrain abnormalities (p < 0.01). However, disease duration, age, and sex were not otherwise different between patients with the presence or absence of basal ganglia, midbrain, or supra- or infratentorial findings in the PSP or MSA groups. The numbers were too small to show any correlations in the CBD group.
Patients with uncertain diagnosis.
In these nine patients with atypical parkinsonism in whom a differentiation between MSA and PSP could not be made according to published criteria, the MR image revealed findings typical for PSP in three patients (33%) and typical for MSA in two patients (22%). In the four remaining patients (44%), no specific abnormality was found.
Discussion.
In this study routine MRI revealed characteristic findings that can be used in clinical practice to help differentiate between PSP and MSA. Using the proposed criteria (see table 4) most patients in the PSP and MSA-C groups could be classified unequivocally with either 1.5-T or 0.5-T scans (74% and 83%). However, in the MSA-P group only slightly more than half of the patients could be classified using 1.5-T scans, and less than 50% could be classified using 0.5-T scans. Moreover, 0.5-T scans misclassified 19% of MSA patients as having PSP. Although in all patients this was due to one or two abnormalities seen mostly by one rater, 0.5-T scans seem unsuitable for differentiating between MSA-P and PSP. On the 1.5-T scans, 47% of patients still could not be classified unequivocally, even though a clinical diagnosis was already possible. The only clinical factors that discriminated patients who had characteristic findings on MRI from those who did not were disease duration and diagnostic certainty in the PSP group, which correlated with the presence of midbrain atrophy. This comparatively low sensitivity of MRI abnormalities in MSA-P is in accordance with our previous finding3 of a high specificity, but lower sensitivity, of these MRI findings in comparison with IPD and control subjects. In the current study, specific changes were seen on all MR images of patients with MSA who had a 1.5-T scan, and in 87% of those who had 0.5-T scans, which is more than in the previous study.3 This improved sensitivity may be due to the learning curve, a change of investigators, and a slightly different patient group.
Our findings of high specificity, but limited sensitivity, of the described MRI abnormalities in patients with MSA are in accordance with those of Savoiardo et al.4 On the other hand, unlike some other authors,4 we have also found MRI useful in supporting a diagnosis of PSP in the great majority of patients. PSP could be differentiated from MSA by the presence of hyperintensity and marked atrophy of the midbrain (<17 mm diameter) as well as atrophy of the frontal or temporal lobes, signal increase in the globus pallidus (on 0.5-T scans only), and atrophy of the red nucleus.
The only abnormality seen significantly more frequently in CBD patients than in MSA patients in this study was global atrophy. However, the presence of infratentorial signal change, a hyperintense putaminal rim, or putaminal hyperintensity, which were characteristic for MSA, and atrophy of the globus pallidus as well as frontal or temporal lobe atrophy exceeding the global atrophy, which were characteristic for PSP and were not seen in any patient with CBD, appear to make a diagnosis of CBD unlikely. There was no obvious asymmetry in any of the paired abnormalities, but subtle changes may have been missed because they were not sought specifically, and coronal scans, which demonstrate asymmetric atrophy best,8 were not performed.
Like others, we found considerable overlap in some findings between the disease groups. This included thinning of the substantia nigra pars compacta (which has been found in MSA and other types of parkinsonism including PD, but not in control subjects15), atrophy of the globus pallidus or putamen, and midbrain atrophy (apart from marked midbrain atrophy with a diameter of less than 17 mm on axial scans, which was characteristic for PSP). Similar to our results in patients with MSA, others also found not only atrophy of the cerebellum and pons, but also of the midbrain.16 These abnormalities, although differentiating atypical parkinsonism from control subjects, are therefore not helpful in differentiating between atypical parkinsonian syndromes with routine MRI. In addition, a higher putaminal iron content than in IPD, as reflected in (mainly posterior) putaminal hypointensity, has been shown in quantitative MRI studies in MSA,17,18 but on visual evaluation of routine scans in this and other studies3,15,19 this was not a discriminating factor between MSA and IPD or control subjects, or between different atypical parkinsonian syndromes.
There are some limitations to this study. Most importantly, the same group of patients was used to determine the abnormalities characteristic for the diseases and to test the value of these findings in differentiating them. Thus, the criteria may be less effective in a different group of patients than shown in this sample in table 5. Therefore, the proposed criteria should be used to support diagnostic classification but should be interpreted with caution. Because none of the findings had optimal sensitivity, specificity, positive and negative predictive values, and may not be equally effective in differentiating other diseases (e.g., Pick’s disease), they should always be interpreted in a clinical context. Second, the size of the disease groups probably does not reflect the proportions of patient groups in a typical movement disorder clinic, where PSP is as frequent as MSA. CBD is rare, and the small size of this group precluded valid statistical comparisons. Furthermore, cortical asymmetry in CBD could have been missed as a discriminating factor for this condition because it was not sought specifically, and coronal scans were not performed. Lastly, the sensitivity of MRI findings in MSA and PSP might have been slightly higher if only 1.5-T scans had been used, particularly in patients with MSA-P. However, sensitivity and specificity differed only slightly between 0.5- and 1.5-T scans.
- Received May 18, 1999.
- Accepted September 16, 1999.
References
Letters: Rapid online correspondence
REQUIREMENTS
You must ensure that your Disclosures have been updated within the previous six months. Please go to our Submission Site to add or update your Disclosure information.
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
You May Also be Interested in
Alert Me
Recommended articles
Limb apraxia in corticobasal degeneration and progressive supranuclear palsy
Sonographic discrimination of corticobasal degeneration vs progressive supranuclear palsy
Direct genetic evidence for involvement of tau in progressive supranuclear palsy
Electrophysiologic patterns of oral-pharyngeal swallowing in parkinsonian syndromes