MRI Volumetric Analysis in Rasmussen Encephalitis: A Longitudinal Study


Address correspondence and reprint requests to Dr. M. Takeoka at Child Neurology Service, Department of Pediatrics, School of Medicine, Keio University, 35 Shinanomachi, Shinjuku-ku, Tokyo 160-8582, Japan. E-mail:


Summary:  Purpose: Rasmussen encephalitis is a progressive inflammatory process with difficult-to-control focal or lateralized seizure activity, leading to hemispheric dysfunction and atrophy in advanced stages. Anatomic changes of atrophy may be subtle in earlier phases of the disease, and progressive changes on serial scans may be difficult to detect. We report a case of early-stage Rasmussen encephalitis with a relatively stable clinical course in whom we performed magnetic resonance imaging (MRI)-based volumetric analysis over an interval of 1 year, to assess for volumetric changes.

Methods: Volumetric analysis was performed on two successive MRI scans obtained at age 5 and 6 years, by using the CARDVIEWS program (J Cogn Neurosci, 1996). The images were segmented into gray- and white-matter structures according to signal intensity of their borders semiautomatically, with manual corrections. The cerebral cortex was further subdivided into smaller parcellation units according to anatomic landmarks identifiable on MRI.

Results: Stable left cerebral hemispheric atrophy and progressive atrophy in the left precentral gyrus, left inferior frontal gyrus, and left cerebellar atrophy were detected over the 1-year interval.

Conclusions: Volumetric analysis enables early detection and quantification of anatomic changes, identification of focal involvement, and assists in determining the severity of disease and timing for surgical interventions such as hemispherectomy.

Rasmussen encephalitis is a progressive inflammatory disease, associated with focal or lateralized seizure activity and encephalopathic changes (1). The inflammatory process is initiated unilaterally and usually remains confined to one hemisphere, although rarely the process may spread to the contralateral hemisphere. Atrophy in the involved hemisphere is frequently seen in advanced stages. Hemispherectomy has been indicated for cases with a clear progressive course. Better outcome is obtained for seizure control and residual function when performed at a younger age, and before contralateral spread (1). New technology using MRI-based volumetric analysis (2) has the potential of allowing clinicians to detect subtle changes in cortical volume in earlier stages of the disease, and progressive changes on serial scans that may be difficult to detect even by expert visual inspection. We report an MRI-based volumetric analysis of the brain in a girl with Rasmussen encephalitis whose disorder had been active for 5 years. The volumetric analysis in this instance was undertaken on two sequential scans obtained over a year during which the clinical course was relatively stable.


Case history

This previously healthy 6-year-old girl had onset of seizures at age 2 years, described as focal clonic activity in the right upper extremity, with occasional spread to the right leg, body, and face. Her birth history and early development was reported as normal. After the onset of the seizures, she developed right arm and leg weakness, and left-hand preference. She also experienced delay in acquiring developmental milestones since the onset of seizures, but no regression had been noticed. She had been treated with multiple antiepileptic drugs (AEDs), but she was able to obtain only partial seizure control with one seizure every 1 to 2 months. No progression was seen in the right arm and leg weakness or frequency of seizures.

The electroencephalogram showed left hemisphere slowing, maximal in the parasagittal areas, with frequent independent spike-and-wave complexes in the left frontal and temporal areas, activated by sleep. Neuroimaging studies obtained at ages 3, 5, and 6 years had shown atrophy in the left cerebral hemisphere and left cerebellum. A biopsy in the frontal pole obtained at age 5 years showed findings consistent with Rasmussen encephalitis, with diffuse gliosis, rare perivascular lymphoid cells, and rare abnormal neurons with cell bodies positive for phosphorylated neurofilaments.


Volumetric analysis was performed on two successive MRI scans obtained at ages 5 and 6 years. The study was approved by the institutional board for clinical investigations (Children's Hospital, Boston). The brain biopsy in the left frontal pole was obtained between the two studies. For both studies, MRI was performed on General Electric 1.5-Tesla Signa (Milwaukee, WI, U.S.A.) by using coronal IR-prepped FAST-SPGR T1-weighted spoiled gradient-echo pulse sequence: TR = 13.8 ms, TE = 2.8 ms, TI = 300 ms, flip angle = 25°, FOV = 24 cm, 124 contiguous 1.5-mm coronal slices, matrix = 192 × 192, averages = 1.

Volumetric analysis

Volumetric analysis was performed with a custom-designed software CARDVIEWS (Center of Morphometric Analysis, Department of Neurology, Massachusetts General Hospital, Boston, MA, U.S.A.), detailed by Caviness et al. (2). Coronal images were adjusted for XYZ axis, with the anterior–posterior axis (Y-axis) corresponding to the anterior commissure–posterior commissure line, and the superior–inferior axis (Z-axis) set orthogonal to the Y-axis, passing through the interhemispheric fissure. Reconstructed coronal images were analyzed by using the CARDVIEWS program. The images were segmented into gray- and white-matter structures according to signal intensity of their borders semiautomatically, with manual corrections (3). The cerebral cortex was further subdivided into smaller parcellation units (PUs) according to anatomic landmarks identifiable on MRI, such as sulci and coronal planes defined by intersection of sulci.

A consensus was obtained between two investigators (M.T. and N.M.) for the assignment of fissures and anatomic landmarks. Volume measurements for segmented structures and PUs were calculated from imaging parameters and number of voxels included in each structure (2). Percentage of volume reduction in each structure was obtained by comparing the volume reduction between the two studies with the volume measurement in the first study. A map of the PUs is shown in Fig. 1.

Figure 1.

Map of parcellation units. The map of the parcellation units as defined by Caviness et al. (2) is shown. A: Lateral surface. B: Medial surface. C: Superior surface. D: Inferior surface. E: Superior opercular surface. F: Inferior opercular surface. The following cortical parcellation units are assigned to the lobes: Lateral frontal lobe: FP, F1, F2, F3t, F3o, PRG, FOC, FO, CO; Medial frontal lobe: PAC, CGa, SMA, FMC, FOC, SC; Parietal lobe: POG, SGa, SGp, SPL, AG, CGp, PCN, PO; Occipital lobe: OLs, OLi, OP, CN, SCLC, CALC, LG, OF; Temporal lobe: TP, T1a, T1p, T2a, T2p, T3a, T3p, TO2, TO3, TFa, TFp, TOF, PP, H1, PT, PHa, PHp, INS).

Data analysis

Volume measurements of segmented structures were compared between right and left. PUs were grouped into lobes in each cerebral hemisphere to compare for symmetry and longitudinal changes. Less variation was seen in volume of lobes compared with individual PUs, among normal controls (4). The volume measurements were first compared in the lobes, and then in individual PUs. As the two studies were performed in a single case, intersubject variability was not of concern.


Segmented and parcellated images are shown in Fig. 2. Volumetric measurements of segmented structures, lobes, and selected PUs in the lateral frontal lobe from the two studies are shown in Fig. 3. A decrease in left hemisphere volume was detected in both MRI studies (first study: left hemisphere, 381 cc; right hemisphere, 441 cc; second study: left hemisphere, 385 cc; right hemisphere, 444 cc). Comparing the two studies, decreased volume was seen in the left lateral frontal lobe [71.4 to 67.2 cc: 4.2 cc (5.9%) reduction]. Specifically, volume reduction was seen in the left precentral gyrus [PRG; 13.7 to 12.3 cc: 1.4 cc (10.2%) reduction], and inferior frontal gyrus pars opercularis [F3o; 2.7 to 1.8 cc: 0.9 cc (33.3%) reduction]. These focal reductions occurred despite the lack of volume loss of the entire hemisphere. The frontal pole (FP), where the biopsy was performed, had a minimal decrease in volume [19.9 to 19.0 cc: 0.9 cc (4.5%) reduction], which we did not consider as significant.

Figure 2.

Segmentation and parcellation. Magnetic resonance images from the two longitudinal studies at ages 5 and 6 years are shown for comparison (top row; initial study at 5 years, bottom row; follow-up study 1 year later). No obvious changes are seen when comparing the images from the two studies. Slices of T1-weighted FSPGR images, segmented structures, and parcellation units are shown at the level of (A) basal ganglia, amygdala, hippocampus, and thalamus (decreased volume in the left cerebral hemisphere is noted); (B) brainstem, fourth ventricle, and cerebellum (decreased volume in the left cerebral hemisphere and cerebellum is noted).

Figure 3.

Volumes of segmented brain structures and lobes of the neocortex. Volumes of the segmented brain structures, lobes, and parcellation units of the neocortex are compared from the two longitudinal studies (No.1; initial study at 5 years, No. 2; follow-up study 1 year later). A: Cerebral hemispheres, cortex, and white matter (decreased volume is seen on the left). B: Cerebellar hemispheres, cerebellar cortex, and cerebellar white matter (decreased volume is seen on the left). C: Basal ganglia, amygdala, hippocampus, and thalamus (a minimal volume decrease is seen in the caudate). D: Lobes of the neocortex (lateral frontal, medial frontal, parietal, occipital, and temporal lobes). E: Parcellation units F3o (inferior frontal gyrus, pars opercularis) and PRG (precentral gyrus).

Regarding other segmented structures, volume reduction of >10% was seen in the left cerebellum [cerebellar hemispheres: 48.8 to 43.0 cc: 5.8 cc (11.9%) reduction; cerebellar cortex: 39.8 to 36.9 cc: 2.9 cc (7.3%) reduction; cerebellar white matter: 8.7 to 7.1 cc: 1.6 cc (18.4%) reduction]. No change in cerebellar volume was seen on the right. Left caudate volume decreased from 3.8 cc to 3.4 cc [0.4 cc (10.5%) reduction] between the two studies. No significant change in volume was seen in other segmented structures between the two studies.


Recent advances in neuroimaging have dramatically increased the sensitivity of detecting subtle anatomic changes. Volumetric analysis by the CARDVIEWS system enables quantitative assessment of such subtle changes in longitudinal studies, which may otherwise be difficult to detect. In our case of Rasmussen encephalitis, volume reduction in the left cerebral hemisphere and the left cerebellum was obvious on visual inspection, even without volumetric analysis. However, atrophy of cortical and subcortical structures was not apparent without formal volumetric determinations. Moreover, the gyrus-by-gyrus analysis undertaken here clearly allows a closer specification of the regional impact of the disease process. Such findings may help in assessing specific neuroanatomic regions and associated functions that are at risk in the patient. In addition, the determination of the rate of progression of atrophy, through quantitative measurements of longitudinal studies of volumetric analysis, could become a criterion for when and whether to undertake hemispherectomy.

In our patient with Rasmussen encephalitis, we detected subtle progressive volume changes from the two successive studies, especially in the precentral gyrus and caudate. Abnormalities have been reported in the precentral gyrus on positron emission tomography (PET) (5) and single-photon emission computed tomography (SPECT) (6) in Rasmussen encephalitis. Because the change in the precentral gyrus was subtle, and with no clinical deterioration, hemispherectomy had been deferred. We will reconsider a surgical option if there is clinical evidence of deterioration of motor or cognitive impairment, or if volumetric studies suggest progression of the disorder.

Normal brain growth must to be considered when analyzing volumetric studies. The patient was age 5 and 6 years at the time of the studies, and the brain volume has not yet reached the full adult size at this age (7). Although the rate of brain growth may be less than that earlier in life, brain growth would still be expected to continue during the year between the two studies.

Both sets of data were analyzed by the same observer (M.T.). Intraobserver variability in the method, considered small compared with interobserver variability, also must be considered in the interpretation of the results (3,4,8,9). Such variability has been reported as <5% on average, ≤8%, by Filipek et al. (9). A difference of >8% seen by the same observer could be considered significant, less likely to be explained by intraobserver variability alone.

No decrease in the cerebral hemisphere volume was detected when comparing the two studies, even in the left hemisphere with the atrophy. The volume increase was seen in the left occipital lobe in the second study. This may correlate with the relatively stable clinical course, with no further clinical progression during the interval of the studies. Regarding the frontal lobe, a 0.9-cc volume reduction in the frontal pole may be attributed to the biopsy in that location; the 1.4-cc reduction in the precentral gyrus and 0.85-cc reduction in the inferior frontal gyrus pars opercularis, which are volume reductions >10%, cannot be attributed to the procedure and are larger than the expected for intraobserver variability (3,9).

The clinical significance of the progressive volume reduction in the left cerebellum is unclear, as no clinical signs were seen suggesting cerebellar dysfunction. No consistent MRI finding of cerebellum has been reported in Rasmussen encephalitis. Considering the fiber connections between the cerebral hemisphere and the cerebellum, cerebellar involvement contralateral to the involved cerebral hemisphere would be expected, rather than ipsilateral (10,11). Geller et al. (12) reported cerebellar involvement contralateral to the involved cerebral hemisphere in Rasmussen encephalitis. The reason for the strong ipsilateral cerebellar involvement in our patient is unknown.


Volumetric measurements of the whole brain by using CARDVIEWS enables quantitative analysis of various brain structures and neocortical regions and allows quantitative comparison of volume measurements in longitudinal studies. In patients with Rasmussen encephalitis, such analysis allows early detection and quantification of anatomic changes, identifies focal involvement, and assists in determining the severity of disease and timing for surgical interventions such as hemispherectomy.

Acknowledgment: The study was supported in part by the Epilepsy Foundation (Research Clinical Fellowship, with support from Pfizer), and the National Epifellows Foundation (Research Grant). This work was presented at the 125th Annual Meeting of the American Neurological Assocation, Boston, Massachusetts, October 15, 2000. We thank Drs. Maria Younes, Lincoln, RI, and Ignacio Valencia, Philadelphia, PA, for helping the preparation of the clinical data.