Address correspondence to Horst Urbach, M.D. at Dept. of Radiology/Neuroradiology, University of Bonn Medical Center, Sigmund Freud Str. 25, D-53105 Bonn, Germany. Email: email@example.com
Summary:Background and Purpose: Focal cortical dysplasia of Taylor's balloon-cell type (FCD-BC) are a frequent cause of pharmacoresistant epilepsy in young patients. In order to characterize FCD-BC, we coupled MRI and histopathology, and analyzed the clinical outcome following epilepsy surgery.
Methods: From an epilepsy data bank with 547 histological specimens, 17 FCD-BC were re-evaluated of which high resolution MRI was available. Five additional FCD-BC were prospectively identified by MRI. Histopathological and immunohistochemical features were related to MRI. Outcome following lesionectomy was analyzed as determined on routine examinations 3, 6 and 12 months following surgery.
Results: All but one lesion were located outside the temporal lobe. A markedly hyperintense funnel-shaped subcortical zone tapering towards the lateral ventricle was the characteristic finding on FLAIR MRI. Histopathologically, the subcortical zone of the FCD-BC displayed hypomyelinated white matter with radially oriented balloon cells and gliosis. Dysplastic neurons were found in the adjacent, disorganized cortex. All patients with complete lesionectomy were seizure free one year following surgery.
Conclusion: Focal cortical dysplasias of Taylor's balloon-cell type (FCD-BC) have characteristic MRI and histopathological findings. MRI recognition is important, since outcome following resective surgery is favorable.
Malformations of the cerebral cortex are a frequent cause of pharmacoresistant epilepsies and developmental disorders (1). Various terms have been introduced to classify the underlying pathology such as cortical dysgenesis (2,3), focal cortical dysplasias (4) or glioneuronal hamartias/hamartomas (5,6). Unfortunately, the nomenclature is not uniform, and different diagnostic terms are even used for malformative lesions that appear identical on histological specimens. One of these peculiar lesions is characterized by focal cortical disorganization with enlarged, dysplastic neurons and the presence of bizarre balloon cells within the cortex and subcortical white matter. Taylor et al. first described it in 1971 (7) detailing the history of ten patients suffering from long-standing epilepsy: Six of ten lesions contained “grotesque cells, probably of glial origin,” within the cortex and subcortical white matter, and four lesions did not. Both lesion types were initially merged and denominated as focal cortical dysplasia of Taylor. Other neuropathological terms for the lesions containing balloon cells include focal cortical dysplasia (2), focal transmantle cortical dysplasia (8), focal cortical dysplasia of Taylor, balloon cell subtype (9), severe cortical dysplasia, type II (10), type II focal cortical dysplasia (11), forme fruste of tuberous sclerosis or type III focal cortical dysplasia (11), balloon cell changes (4) and glioneuronal hamartoma with TS cells (5).
With the recent progress in magnetic resonance imaging (MRI), malformations of the cerebral cortex are increasingly recognized during the presurgical evaluation of patients with pharmacoresistant epilepsies. A major advantage of high resolution MRI is the topographical characterization of the lesion with respect to its size, location and extension compared to the anatomically less preserved histological specimen. Furthermore, a reliable distinction of certain groups of pharmacoresistant epilepsy patients before surgery would allow to develop appropriate treatment strategies. In this study, we re-evaluated the specific topography of 17 lesions classified as focal cortical dysplasia of Taylor's balloon cell type (FCD-BC ) using high resolution MRI and correlated the findings with histopathological features and postoperative outcome. A diagnostic paradigm of the presurgical work-up of epilepsy patients was established which allowed us to prospectively identify five additional FCD-BC patients in a 5-month period.
MATERIAL AND METHODS
The study group was derived from a consecutive series of 547 patients who underwent surgery for pharmacoresistant epilepsy between 1996 and 2000. Upon histopathological analysis, 17 patients (3.1%) with morphological features of focal cortical dysplasia of Taylor's balloon cell type (FCD-BC, Table 1) were identified.
Table 1. Clinical and morphologic characteristics of patients with FCD-BC
Duration of epilepsy (yr)
Location of lesion
Size of lesion (mm)
Size of resection (mm)
Outcome/ Engel class
Patients 1–17 were retrospectively evaluated, patients 18–22 prospectively identified.
TSC, extracerebral features of the tuberous sclerosis complex have been (yes)/have not been (no) identified.
a Clinical follow-up <3 months. Twenty patients underwent total lesionectomy, one patient incomplete lesionectomyb‡, and another patient multiple subpial transections (MST) onlyc‡.
Right fourth occipital gyrus (O4)
20 × 20 × 10
30 × 20 + MST
Class I A
Right superior frontal gyrus (F1)
8 × 8 × 8
25 × 30
Class I A
Left medial frontal gyrus (F2)
30 × 20 × 20
45 × 30
Class I A
Right superior frontal gyrus (F1)
8 × 6 × 8
80 × 60 + MST
Class I A
Right angular gyrus
20 × 20 × 15
20 × 36 + MST
Class I A
Right superior parietal gyrus (P1)
20 × 20 × 20
33 × 30 × 38
Class I A
Right inferior parietal gyrus (P2)
30 × 30 × 30
50 × 50 × 25
Class I A
Right superior frontal gyrus (F1)
20 × 15 × 20
50 × 30 + MST
Class I A
Left pre- and postcentral gyrus
30 × 30 × 40
Left rectus and medial orbial gyri
15 × 20 × 15
20 × 20
Class I A
30 × 30 × 30
40 × 50
Class I A
Right inferior frontal gyrus (F3), right angular gyrus
25 × 30 × 25; 30 × 30 × 30
30 × 30 × 15 + MST
Class I A
Left cingulate gyrus
10 × 70 × 10
15 × 30
Class I A
Right middle occipital gyrus (O2)
30 × 50 × 30
35 × 50 + MST
Class I C
Left sub- and precentral gyrus
20 × 20 × 25
25 × 70
Class I A
Left superior frontal gyrus (F1)
10 × 15 × 20
70 × 35 + MST
Class I A
Right postcentral gyrus
15 × 8 × 8
15 × 8 + MST
Class I A
Right superior frontal gyrus (F1), right superior temporal gyrus (T1)
High resolution magnetic resonance (MRI) and computed tomography imaging was routinely performed during the presurgical evaluation of all patients.
MRI was performed using a 1.5 Tesla system (Gyroscan ACS-NT, Philips Medical Systems, Best, The Netherlands). Sagittal T1-weighted spin echo (slice thickness 5 mm, interslice gap 0.5 mm), axial FLAIR and T2-weighted fast spin echo (slice thickness 5 mm, interslice gap 0.5 mm), coronal FLAIR (slice thickness 2 to 5 mm, interslice gap 0.2 to 0.5 mm) and T2-weighted fast spin echo (slice thickness 2 to 5 mm, interslice gap 0.2 to 0.5 mm), coronal T1-weighted inversion recovery (slice thickness 5 mm, interslice gap 0.5 mm) and axial T1-weighted spin echo sequences (slice thickness 5 mm, interslice gap 0.5 mm) before and following Gd-DTPA injection were acquired.
Computed tomography scans were not enhanced and obtained with a slice thickness of 4 mm infratentorially and 8 mm supratentorially (Somatom Plus, Siemens AG, Erlangen, Germany).
The following aspects of the lesions were documented: location, size, shape, calvarial remodeling, focal widening of the subarachnoid space, gray matter involvement, white matter involvement, signal intensity on T1-weighted, T2-weighted and FLAIR- images, calcifications, contrast enhancement.
Surgical specimens submitted for neuropathological evaluation were microscopically analyzed using haematoxylin-eosin (HE), haematoxylin-eosin-luxol-fast-blue (HE-LFB) and Nissl staining. Immunohistochemical studies included reactions with a monoclonal antibody directed against Vimentin (V9, Dako), polyclonal antibodies directed against glial fibrillary acid protein (GFAP), a monoclonal antibody directed against human neurofilament protein (2F11, Dako), a monoclonal antibody directed against neuronal specific enolase, a monoclonal antibody to synaptophysin (SY 38, Dako), a monoclonal antibody directed against Ki67 (MIB1, Dako), a monoclonal antibody directed against CD 34 (QBend 10, Immunotech), and polyclonal antibodies directed against nestin (nestin 129, R. D. G. McKay, Bethesda, NIH, USA) using standard protocols and the avidin-biotin-peroxidase complex with diaminobenzidine as chromogen. Vimentin and nestin immunoreactivity were simultaneously detected using dual channel laser scanning microscopy (Leica TSC-NT, Bensheim, Germany) and fluorescence labeled secondary antibodies with different wavelengths (FITC and Texas Red).
All patients were re-examined at routine intervals three months, six months and one year following surgery. Seizure outcome was determined according to Engel's classification (12).
We retrospectively evaluated 11 male and six female patients (age six to 48 years, mean age 24 years), who had suffered from pharmacoresistant focal epilepsies between two and 42 years (table). Fifteen patients had no extracerebral features of tuberous sclerosis by routine clinical examination. Two females (#7, #12) had skin lesions (facial angiofibromas, hypomelanotic macules) and retinal hamartomas as major features of tuberous sclerosis (13). All patients underwent surgery following extensive presurgical evaluation as described previously (14–16). Four patients were operated guided by intraoperative electrocorticography and in 12 patients subdural grid and/or strip electrodes were implanted. Subdural electrodes were implanted, when 1) surface EEG recordings failed to show ictal activity or revealed generalized or multifocal pattern (n = 8), 2) electrical stimulation mapping was needed due to the location of the lesion close to eloquent cortex areas (n = 6), or 3) MRI showed more than one lesion (n = 1). EEG recordings from subdural electrodes demonstrated ictal EEG activity in all patients with one lesion. In patient #12, two lesions in the right frontal and parietal lobes were covered by two grid electrodes. Only the parietal lesion showed ictal EEG activity and was resected subsequently.
For a period of five-months, we prospectively identified additional five patients (one male, four female, age five to 35 years, mean age 23 years) with MRI findings suggestive of FCD-BC. All patients underwent surgery following subdural grid implantation and showed histopathological signs of FCD-BC (Figure 1; table). Of these patients, four had no extracerebral features of tuberous sclerosis. In one female (#19), cardiac rhabdomyoma and hypomelanotic macules as major features of tuberous sclerosis were found (14). This patient revealed lesions in the right frontal and temporal lobes. EEG recordings from subdural grid electrodes demonstrated ictal EEG activity of the subsequently resected frontal lesion only.
Taken together, 22 patients had 24 lesions: Ten lesions were located in the frontal lobe, one in the cingulate gyrus, six in the parietal lobe, one in the temporal lobe, and two in the occipital lobe, respectively. Another four lesions involved the pre- and/or postcentral gyrus. Lesion diameters ranged from 6 to 70 mm.
Nineteen lesions were rather circumscript, from which 11 lesions involved two adjacent gyri, four lesions one gyrus, and four lesions the cortex in the depth of a sulcus only (see Fig. 2). One lesion in the cingulate gyrus had a band-like shape. Focal thinning of the skull was observed in one occipital lesion, and the subarachnoid space was focally widened in additional 12 lesions.
All lesions involved the cortex. Cortical thickness was normal in four, and the cortex was focally thickened in twelve lesions. In eight lesions, cortex and underlying white matter could not be separated from each other.
Signal intensity of the disturbed cortex and subcortical white matter was isointense to the surrounding gray matter on T1-weighted images. Lesions were therefore only visible, if the cortex was markedly thickened (n = 4) or if there were adjacent changes in the subcortical white matter (n = 6). On T2-weighted fast spin echo images, 16 lesions were slightly hyperintense and seven lesions isointense to the surrounding gray matter. On FLAIR fast spin echo images, all lesions were slightly hyperintense. None of the lesions was calcified, one lesion showed band-like contrast enhancement directed towards the lateral ventricle.
On FLAIR fast spin echo images, all but two lesions showed markedly hyperintense zones in the subcortical white matter. A comparison of coronal and axial MRI revealed a funnel-shaped three-dimensional structure of these hyperintense zones (Fig. 1 A and B). In general, the funnel's tip pointed towards the lateral ventricle, reaching the ventricular wall in 18 of 22 lesions. The base of the funnel extended towards the ascending part of the gyrus when the entire gyrus seemed affected from cortical disorganization. When the depth of a sulcus was affected, the hyperintense zones seemed to encapsulate the disarrayed cortex. The distinct geometry of this lesion was a neuroradiological hallmark. In fact, some of the smaller lesions were only identified due to the funnel-shaped subcortical hyperintensity accompanied by a focal widening of the subarachnoid space (Fig. 2A).
On T2-weighted fast spin echo images, strikingly hyperintense zones were visible in 18 lesions. Compared to FLAIR MRI however, T2-weighted fast spin echo images were noted of lower diagnostic accuracy since small subcortical hyperintensities could be easily missed due to high signal CSF (Fig. 2).
Similar cerebral imaging abnormalities were observed in patients (#'s 7, 12, 18) with definite tuberous sclerosis, and the neuroradiological hallmarks described above uniquely characterized brain lesions of FCD-BC and TSC patients.
All surgical specimens were reviewed using standard histological and immunohistochemical reactions and showed similar patterns of focal cortical and subcortical alterations. The affected neocortex revealed severe disorganization, whereas adjacent brain tissue demonstrated a regular six-layered cytoarchitecture (Fig. 3 A–C). Reduced numbers of pyramidal neurons were noted in the affected areas. Residual neurons of abnormal, dysplastic phenotype were usually positioned without anatomical orientation or in small aggregates (Fig. 3 D, E). Their enlarged cytoplasm frequently contained Nissl substance clustering at the cell membrane. Dysplastic neurons accumulated cytoplasmic aggregates of neurofilament protein (Fig. 3E, inset), usually restricted to the axonal compartment. Within the dysplastic cortex, bizarre and hypertrophic balloon cells (see below) were observed in all territories, reaching the superficial layers of the cortex as well. Underneath, in close anatomic association, the white matter was hypomyelinated (recognized by reduced staining with luxol fast blue, but see also Fig. 1 D) and showed accumulation of radially oriented balloon cells (Fig. 3 and 4). These balloon cells were characterized by gigantic opaque/eosinophilic cytoplasm and one or more peripherally located nuclei (Figure 4 A).
Immunohistochemically, balloon cells consistently exhibited immunoreactivity for vimentin and nestin (Fig. 3C; 4 D–F). Occasionally, immunoreactivity for CD34 (Figure 4C), GFAP, synaptophysin or neurofilament protein has been recognized in these cells. Ectopic neuropil islands were usually present in the white matter as detected by synaptophysin staining. The affected cortex and white matter showed severe astrogliosis, the latter with additional signs of edema. While hypertrophic astrocytes were often difficult to distinguish from balloon cells, they consistently expressed GFAP rather than vimentin or nestin (Fig. 4 B). Reactive astrocytes could also be observed in adjacent, normal brain tissue. The lesion lacks significant proliferative potential, as assessed by mitotic activity or by antibody detection of the Ki-67 epitope.
Surgery and outcome
Surgery was intended to resect the electrophysiologically defined seizure onset zone while sparing eloquent cortex areas. Seizure onset zones and—if possible—resected cortex areas were always larger than the lesions visible on MRI (see table). Resections including complete removal of (the cortical aspect of) the lesions were achieved in 20 patients (16 retrospectively evaluated and three prospectively identified patients). One patient each in both groups received incomplete lesionectomy or multiple subpial transsections (MST) only due to the unfavorable location of their lesion in the pre- and postcentral gyri. The funnel-shaped part of the lesion through the white matter was not considered relevant for seizure induction because of lack of contact to healthy cortex. Outcome seems to prove that this assumption is correct.
Long term outcome data are only available for the retrospectively evaluated patients. In this group, complete lesionectomy yielded in a seizure free outcome as determined on follow-up examinations three months, six months and one year following surgery (Engel class I). Although the observation period of the prospectively identified patients was shorter than three months, a similar clinical outcome was noted so far. All patients with complete lesionectomy (n = 4) became seizure-free. In both MST-patients, seizure frequency was reduced by more than 75% (Engel class III) (12).
Focal cortical dysplasia of Taylor's balloon cell type (FCD-BC) appears as distinct entity with respect to MRI findings, histopathological features and clinical outcome: On FLAIR MRI, a markedly hyperintense funnel-shaped area in the subcortical white matter is the hallmark of this disease. Focal cortical dysplasia with adjacent hypomyelinated white matter and accumulation of balloon cells within the hypomyelinated white matter characterize this disease histopathologically. If complete surgical resection (of the cortical aspect of the lesion) can be achieved, patients become seizure free.
A MRI pattern describing the lesions as either radial hyperintense bands, wedge-shaped, non specific or tumefactive has repeatedly been noted in patients with tuberous sclerosis (17) and in patients with FCD-BC (8,9,18). These studies encompassed Proton-density and T2-weighted images, on which cortical lesions are difficult to delineate from the intermediate or high signal CSF. We acquired a FLAIR fast spin echo sequence with intermediate spatial resolution (slice thickness 2 to 5 mm) and multiple excitations (number of excitations 2 to 4). This sequence proved to be helpful, since the signal of CSF is nulled and marked subcortical hyperintensities can be clearly distinguished from the only slightly hyperintense cortex. Using FLAIR sequence, we were able to identify even small FCD-BC that could have been easily missed on T2-weighted fast spin echo sequences (see Fig. 2). We did not compare spin echo and fast spin sequences. Albeit the latter have a higher contrast resolution, they are impractical due to a markedly longer acquisition time. From an imaging point of view, the aforementioned MRI pattern has been found useful to differentiate FCD-BC and low-grade gliomas (9). In small lesions, gyral scarring could have been mistaken for FCD-BC, particularly when the depth of a sulcus was affected and the subarachnoid space showed focal widening (ulegyria). It's the patient's history, and the characteristic FLAIR pattern (which shows hyperintense cortex and subcortical white matter in ulegyria) that helps to distinguish both entities. Focal cortical dysplasias of Taylor's balloon cell type also attract attention with regard to the location of the lesions and the outcome following surgery. As described by Kuzniecky et al. (18) and in contrast to many other epileptogenic lesions (14,19), FCD-BC are usually located outside the temporal lobe. Approximately half of the lesions can be identified during the surgical procedure by a focal widening of the subarachnoid space (3), the macroscopic aspect of the brain's surface (Fig. 1 C), or rarely focal thinning of the skull (calvarial remodelling). A small lesion that affects the cortex in the depth of a sulcus, however, is not easily detected. As suggested here, there is an excellent outcome following complete excision of the lesion. Lesion and perilesional tissue must be resected; the extent of resection is determined with ictal EEG recordings, electrical stimulation mapping and anatomical findings (resection includes a gyrus and ends at a sulcus e.g.). It seems noteworthy that not the resection of the subcortical, balloon cell-rich parts of the lesion (which are of highest diagnostic value), but rather the really complete resection of the dysplastic cortex seems to determine the clinical outcome. This observation must be confirmed with a longer follow-up and a larger number of patients, but it is extremely useful, since sparing the subcortical white matter reduces the surgical risk of encroaching on motor or other white matter fiber tracts.
All patients with complete lesionectomy were seizure free at follow-up examinations obtained one year after surgery. This outnumbers the data from large series of patients with mesiotemporal or neocortical temporal lobe epilepsy as well as frontal epilepsy (14–16,19,20), in which a mean of 67% of patients were seizure free following surgical treatment. It also outnumbers the data from a series of 35 patients with different malformations of cortical development, in which 49% of patients were seizure-free and another 23% had rare seizures at latest postoperative follow-up (21). It remains to be evaluated, whether a better identification and topographic characterization or FCD-BC by preoperative MRI or a different intrinsic epileptogeneity is the cause for the better surgical outcome.
A successful neuropathological detection of FCD-BC relies on well-preserved anatomical specimens (Fig. 1). If small or highly fragmented biopsy specimens of the lesion are obtained for histology, characteristic neuropathological features may be missed. Reliable identification and classification of FCD-BC is made easier taking into account the facts from this study and aiming for an subpially developed en-bloc resection of the lesion (if at all possible) with some parts of the funnel-shaped lesion attached to the cortical block. Immunohistochemical studies using antibodies directed against vimentin and nestin particularly highlight the balloon cell component and thus represent a diagnostic hallmark for histopathological diagnosis (Figs. 3, 4). Additionally, CD34-immunoreactivity is helpful for the differential diagnosis since it does not occur in reactive astrocytes, and is on the other hand so abundant in surgical specimens of ganglioglioma (12). Further histological differential diagnoses include other glioneuronal neoplasms, i.e. dysembryoplastic neuroepithelial tumors. However, neuropathological evaluation should provide evidence for proliferative activity of neoplasms and neuronal as well as glial tumor cell populations.
Balloon cells are the neuropathological hallmark in FCD-BC lesions as well as in cortical tubers of tuberous sclerosis. These gigantic, dysplastic, often multinucleated cells share characteristics of both neuronal and glial lineage indicating an origin from pluripotent brain cells (4). It is suggested, that balloon cells fail to commit or to differentiate into a specific cell type within the first trimester. With the presence of balloon cells, it is reasonable to consider cortical tubers and FCD-BC as disorders of neuronal and/or glial proliferation (1), like it is for other disorders containing balloon cells (e.g. hemimegalencephaly) (1,23).
It is still controversial whether FCD-BC and tuberous sclerosis are two different entities or whether FCD-BC represents a “forme fruste” of tuberous sclerosis. The present study supports the latter hypothesis in two ways: Morphological and clinical findings are highly comparable in FCD-BG and tuberous sclerosis patients. And tuberous sclerosis patients can have the same excellent clinical outcome following complete lesionectomy as have FCD-SC patients.