Brain (2000), 123, 2150–2159
Opposite ictal perfusion patterns of subtracted
SPECT
Hyperperfusion and hypoperfusion
Hyang Woon Lee,1 Seung Bong Hong2 and Woo Suk Tae2
1Department
of Neurology, Ewha Womans University
Hospital and 2Epilepsy Program, Department of Neurology,
Neuroimaging Laboratory, Samsung Medical Center,
Sungkyunkwan University School of Medicine, Seoul,
South Korea
Correspondence to: Seung Bong Hong, MD, Department of
Neurology, Samsung Medical Center, Sungkyunkwan
University School of Medicine, 50, Ilwon-dong,
Kangnam-ku, Seoul, 135-710, South Korea
E-mail: [email protected]
Summary
To investigate the patterns of ictal perfusion and related
clinical factors, single photon emission computed
tomography (SPECT) subtraction was performed in 61
patients who had undergone epilepsy surgery. In addition
to the ictal hyperperfusion region, the ictal hypoperfusion
area was obtained by SPECT subtraction. The ictal
perfusion patterns of subtracted SPECT were classified
into focal hyperperfusion, hyperperfusion-plus, combined
hyperperfusion–hypoperfusion and focal hypoperfusion
only. The concordance rate of seizure localization was
91.8% in the combined analysis of ictal hyperperfusion–
hypoperfusion by SPECT subtraction, 85.2% in hyperperfusion images of SPECT subtraction and 68.9% in the
visual inspection of ictal SPECT. Ictal hypoperfusion
occurred less frequently in temporal lobe epilepsy (TLE)
than in extra-TLE. Mesial temporal hyperperfusion alone
was seen only in mesial TLE while lateral temporal
hyperperfusion alone was observed only in neocortical
TLE. Hippocampal sclerosis had a much lower incidence
of ictal hypoperfusion than other pathologies. Some
patients showed ictal hypoperfusion at the epileptic focus
with ictal hyperperfusion in the neighbouring brain
regions where ictal discharges propagated. Hypoperfusion
as well as hyperperfusion in ictal SPECT should be
considered for localizing epileptic focus. The mechanism
of ictal hypoperfusion could be an intra-ictal early
exhaustion of seizure focus or a steal phenomenon
associated with the propagation of ictal discharges to
adjacent brain areas.
Keywords: subtracted SPECT; partial epilepsy; ictal perfusion pattern; ictal hypoperfusion; combined hyperperfusion–
hypoperfusion
Abbreviations: ETLE ⫽ extratemporal lobe epilepsy; FDG ⫽ [18F]fluorodeoxyglucose; SPECT ⫽ single photon emission
computed tomography; TLE ⫽ temporal lobe epilepsy
Introduction
Both interictal and ictal SPECTs have been used successfully
for localizing temporal lobe epilepsy (TLE) (Jack et al.,
1994; Ho et al., 1995), and have been reported to be less
sensitive but more specific in extratemporal lobe epilepsy
(ETLE) (Marks et al., 1992; Harvey et al., 1993; Ho et al.,
1994; Spencer, 1994; Newton et al., 1995). The low sensitivity
of ictal single photon emission computed tomography
(SPECT) in ETLE may be related to various clinical factors
and technical limitations: (i) inhomogeneous and various
perfusion patterns in brain regions, (ii) difficulty in detection
of subtle perfusion changes, (iii) variability in the amount of
radioisotope delivered to the brain and time taken to inject
radioisotope, (iv) different slice levels and orientations arising
© Oxford University Press 2000
from different patient positioning during the interictal and
ictal SPECT scans, and (v) lack of quantitative assessment
of the difference between interictal and ictal SPECT images,
and other factors.
Recently, computer-aided methods to subtract interictal
from ictal SPECT and to co-register the subtracted SPECT
image to MRI have been developed (Zubal et al., 1995).
SPECT subtraction is considered a useful tool for localizing
seizure focus. However, interpretation of ictal and subtracted
SPECT studies has many pitfalls. First, true ictal SPECT
results are often difficult to obtain because the completion
of radioisotope injection during clinical seizures is not easy,
especially in short duration seizures of extratemporal origin.
Subtracted SPECT in partial epilepsy
Secondly, the perfusion changes of the epileptogenic zone
during the peri-ictal period are not, as yet, known exactly.
The seizure semiology at the time of injection represents
only the seizure discharges occurring at the symptomatogenic
zone, which can be either the epileptogenic area or the
neighbouring brain region. The time of the injection, duration
of seizures, propagation speed and pathway of ictal EEG
discharges, and changing types of clinical seizures can affect
the perfusion patterns of the epileptogenic zones during the
peri-ictal period. Therefore, assessment of the real perfusion
changes in the epileptogenic zone is not simple.
This study aimed (i) to classify the various ictal perfusion
patterns obtained by subtracted SPECT, (ii) to determine
whether the detection of ictal focal hypoperfusion as well as
hyperperfusion in subtracted SPECT improved the seizure
localization, (iii) to clarify the contributing clinical factors
to the determination of the ictal perfusion patterns, and
(iv) to compare the sensitivities of seizure localization in
subtracted SPECT-MRI co-registration, of conventional
visual inspection of interictal and ictal SPECT-MRI and of
interictal [18F]fluorodeoxyglucose (FDG)-PET.
Methods
Patients
Sixty-one patients with intractable epilepsy were included.
All patients underwent the precise presurgical evaluation
according to the Epilepsy Surgery Protocols of Samsung
Medical Center including clinical examination, long-term
video-EEG monitoring, brain MRI, FDG-PET, interictal and
ictal SPECT studies, and SPECT subtraction. All subjects
had epilepsy surgery between 1995 and 1998.
Interictal and ictal SPECT studies
Ethyl cysteinate dimer labelled with 99mTc was injected
intravenously for SPECT studies. Brain SPECT scan was
performed within 90 min after the radiotracer injection
(25 mCi 99mTc–ethyl cysteinate dimer) using a three-headed
Triad XLT system (Trionix Research Laboratory, Twinsburg,
OH, USA) equipped with low-energy and high-resolution
collimators. The transaxial system resolution of this
camera is 6.9 mm full width at half maximum. Images were
reconstructed by filtered back-projection using a Butterworth
filter. Attenuation correction was performed using Chang’s
method (attenuation coefficient ⫽ 0.12 cm–1) (Chang, 1987).
Interictal injection of 25 mCi 99mTc-ECD was performed
in patients who had no documented seizure activities in the
previous 24 h period or more. For ictal studies, patients
received an injection of radiotracer during a seizure activity
or immediately after cessation of clinical seizures. The
SPECT was considered as ictal when the radiotracer was
injected during the clinical or EEG seizure activity and
postictal when injected after the seizure ended. The patients
were continuously monitored by a long-term video-EEG
monitoring system during this phase. As soon as seizure
2151
activity was witnessed or alarmed by seizure button, the
radiotracer was injected rapidly by a trained EEG technician
or nurse.
MRI imaging
The MRI scanning was performed with a GE Signa 1.5 Tesla
scanner (GE Medical Systems, Milwaukee, Wisc., USA).
The spoiled gradient recalled volumetric MRI was scanned
with the parameters of no gap, 1.6 mm thickness, 124 slices,
TR (repetition time)/TE (echo time) ⫽ 30/7, flip angle ⫽
45, 1 NEX (number of excitation), coronal. The voxel
dimension was 0.875 ⫻ 0.875 ⫻ 1.6 mm.
The fluid attenuated inversion recovery images were
scanned with oblique coronal, 1.0 mm gap, 4.0 mm thickness, 32 slices, TR/TE ⫽ 10002/127.5, 1 NEX, while the
axial images of the fluid attenuated inversion recovery were
also obtained, with 2.0 mm gap, 5.0 mm thickness. The T2weighted images were scanned with 0.3 mm gap, 3.0 mm
thickness, 56 slices, TR/TE ⫽ 5300/99, flip angle ⫽ 90°,
3 NEX, oblique coronal. T2 axial images were also scanned
with 2.0 mm gap, 5.0 mm thickness.
Image processing for SPECT subtraction
The SPECT subtraction technique was processed on an offline SUN Ultra 1 Creator workstation (Sun Microsystems,
Calif., USA) with a commercial software package,
ANALYZE 7.5 (Biomedical Imaging Resource, Mayo
Foundation, Minn., USA).
All biomedical images were transferred from each scanner
console to the Unix workstation by a 4 mm DAT (digital
audio tape) device.
The SPECT subtraction procedures consisted of the
following processes.
Step 1: ictal–interictal SPECT registration. Before
subtracting each pixel value of interictal SPECT from ictal
SPECT images, the three-dimensional position of interictal
SPECT was transformed to ictal SPECT. In all cases, root
mean square distance was within 1 voxel. Correct registration
is important to optimise the sensitivity of the subtraction
technique, and inaccurate registration may produce false
perfusion differences.
Step 2: normalization of radioisotope uptake level. Levels
of radioisotope were normalized as each patient had different
uptake levels, and the normalization factor was calculated
over the whole brain (O’Brien et al., 1998).
Step 3: ictal-transformed interictal SPECT subtraction. To
obtain the cerebral perfusion difference, the ictal SPECT
was subtracted from transformed and normalized interictal
SPECT. Differences in radioisotope uptake level were
calculated pixel-by-pixel by subtraction and the subtracted
SPECT was divided separately into positive (hyperperfusion)
and negative (hypoperfusion) images.
Step 4: noise erasing. To erase subtraction noise, the
standard deviation of each subtracted SPECT was calculated
2152
H. W. Lee et al.
and the two standard deviations were adjusted to erase the
noise. This was done to minimize the influence of random
noise. Some patients may have had multiple regions of hyperor hypoperfusion, and in these cases, the cerebral cortical
pixels were considered significant only when they showed
the largest interictal–ictal difference and were congregated
in a localized area. The other increased pixels, which were
randomly distributed throughout the image and not
congregated in a localized area, were regarded as noise.
Step 5: MRI-subtracted SPECT registration. The accurate
anatomical localization of epileptogenic focus is important
for successful epilepsy surgery (Zubal et al., 1995; O’Brien
et al., 1998). For localization of difference images, we coregistered subtracted SPECT with SPGR MRI of the patient’s
whole brain (Fig. 1). In most cases, the error ranges of
SPECT-MRI registration were within 3 voxels.
Interpretation of interictal, ictal and subtracted
SPECTs
When interictal SPECT was subtracted from ictal SPECT,
the brain region with the positive values indicated a hyperperfusion area, and those with negative values a hypoperfusion area. Four different kinds of SPECT images were
interpreted separately by two reviewers (H.W.L. and W.S.T.)
who were blinded to clinical data, results of other tests and
surgical outcome. Four different sets of SPECT images were
as follows: (i) conventional visual comparison of interictal–
ictal SPECT, (ii) hyperperfusion images of subtracted SPECT
(‘positive results of subtraction’), (iii) hypoperfusion images
of subtracted SPECT (‘negative results of subtraction’), and
(iv) both hyper- and hypoperfusion images of subtracted
SPECT (hyperperfusion–hypoperfusion combined analysis).
These four sets of SPECT images were presented to the first
two reviewers one set at a time, in random order of sets and
patient sequences, so that all reviewers were blinded to their
previous interpretations to the other sets. If the two reviewers
disagreed, a third blinded reviewer was used (S.B.H). Agreement of the third reviewer with one of the primary reviewers
served as the final determination.
Reviewers localized the abnormality in each SPECT
image to one of 16 sites (i.e. either right or left: frontal,
frontotemporal, temporal, frontoparietal, temporoparietal,
parietal, occipitoparietal or occipital) or classified the image
as non-localizing or non-lateralizing. The final decision of
localization was reached by the agreement of all reviewers.
If the three reviewers agreed to no significant perfusion
change or failed to reach an agreement on localization after
discussion, that image was considered as non-localizing.
The concordance rates to the final epileptic focus proven
by epilepsy surgery as well as scalp and invasive ictal EEG
were calculated in subtracted SPECT, conventional visual
inspection of interictal and ictal SPECT, MRI and interictal
FDG-PET studies. The various ictal perfusion patterns in
subtracted SPECT were described, and we tested the
relationship between the ictal perfusion patterns of subtracted
SPECT and clinical factors such as the locations of seizure
foci (temporal versus extratemporal origins), types of tissue
pathology (hippocampal sclerosis, cortical dysplasia and
others), and the time taken to inject the radioisotope.
Injection time of radiotracer for SPECT study
For determination of seizure length and injection timing,
the seizure onset was taken as the time of earliest indication
of warning (verbalized or pushing the seizure button) or of
abnormal movements, behaviour or impaired awareness. The
end of a seizure was the time when ictal movements or
behaviour ceased. To confirm the ictal period, ictal EEG was
interpreted simultaneously. The time of the injection was
taken as the time when the plunger on the syringe containing
the radiotracer was fully depressed. The exact time of
injection was recorded by the epilepsy fellow or the staff.
The time of the injection in relation to the seizure occurrence
was determined accurately by replaying the videotape that
had recorded the clinical seizure and radioisotope injection
and reviewing ictal EEG during that seizure.
Injection time of the radioisotope in each patient was
normalized to seizure duration in order to accommodate
seizures of different lengths. Therefore, the normalized
percentage time (t) was calculated by [(tracer injection time –
seizure end time)/total seizure duration] ⫻ 100, where the
time and duration were measured in seconds. Thus, positive
percentage time values (t ⬎ 0) represent time after seizure
cessation (postictal period) while negative percentage time
(–100 ⬍ t ⬍ 0) represent time from seizure onset to seizure
end (ictal period). For example, t ⫽ –50% injection time
means the injection was performed at the midpoint of the
seizure, t ⫽ –100 at seizure onset, and t ⫽ 0 at seizure end.
Statistical analysis
McNemar’s χ2-test for paired proportions was used to compare the proportion of the patients localized by subtracted
SPECT-MRI co-registration with the proportion localized by
traditional visual inspection of interictal and ictal SPECT,
MRI and PET studies. Inter-observer agreement between the
first two reviewers was determined by Cohen’s kappa (κ)
scores using generalized κ-type statistics. Agreement was
considered poor when κ was ⬍0.4, good when κ was 0.4–
0.75 and excellent when κ was ⬎0.75 (Landis and Koch,
1977). The χ2-test was used to compare the categorical data,
and Fisher’s exact test was used for dichotomous variables
to compare the numbers of various perfusion patterns in
different seizure foci and different pathological groups. The
significance level was set at P ⬍ 0.05 for all tests.
Results
Clinical features
Demographics
Of the total 61 patients, 35 were males and 26 were
females. The patients’ ages ranged from 1 to 48 years
Subtracted SPECT in partial epilepsy
Table 1 Epilepsy syndrome classification of patients
(n ⫽ 61)
TLE
Mesial
Neocortical
ETLE
Total
Non-lesional (%)
Lesional (%)
Total (%)
17
7
10
12
29
26
20
6
6
32
43
27
16
18
61
(27.9)
(11.5)
(16.4)
(19.7)
(47.5)
(42.6)
(32.8)
(9.8)
(9.8)
(52.5)
(70.5)
(44.3)
(26.4)
(29.5)
(100.0)
(mean ⫽ 24 ⫾ 9.9 years). The mean age of seizure onset
was 12 ⫾ 8.1 years, and the mean duration of seizure history
was 12 ⫾ 7.4 years.
The final epilepsy classifications were 27 (44.3%) mesial
TLE, 16 (26.4%) neocortical TLE, and 18 (29.5%) ETLE.
Thirty two (52.5%) had lesions on MRI and the other 29
(47.5%) had no lesions. MRI of mesial TLE patients showed
hippocampal sclerosis in 20 and no lesion in seven. Six of
the 16 neocortical TLE and six of the 18 ETLE patients had
lesion shown by MRI (Table 1).
2153
(21/61) of side-by-side conventional visual inspections, in
3.3% (2/61) of hyperperfusion subtracted SPECT and in
14.8% (9/61) of hypoperfusion subtracted SPECT and in 8.2%
(5/61) of combined hyperperfusion–hypoperfusion analyses.
The inter-observer agreement was significantly poorer for the
conventional visual inspection (κ ⫽ 0.23, P ⬍ 0.05) than it
was for either hyperperfusion subtracted SPECT (κ ⫽ 0.82,
P ⬍ 0.05), hypoperfusion subtracted SPECT (κ ⫽ 0.63,
P ⬍ 0.05) or the hyperperfusion–hypoperfusion combined
analysis (κ ⫽ 0.71, P ⬍ 0.05).
MRI was performed in all 61 patients and FDG-PET
in 52 patients. The comparison of concordance rates was
performed in these 52 patients who underwent all MRI, PET
and SPECT studies. The concordance rates to epileptic focus
were 57.7% (30/52) by MRI, 63.5% (33/52) by PET, 84.6%
(44/52) by hyperperfusion image of subtracted SPECT and
90.4% (47/52) by hyperperfusion–hypoperfusion combined
analysis of subtracted SPECT (Table 3). The seizure
localization rates of subtracted SPECT were significantly
higher than those of MRI and PET studies (P ⬍ 0.01 by
McNemar’s χ2-test).
Postsurgical outcome and pathology
All 61 patients had surgery after presurgical evaluation
including long-term video-EEG monitoring, MRI, interictal–
ictal SPECT, FDG-PET and intracranial EEG recording if
necessary. Of these, an invasive study was performed in 42
patients (68.9%); these included eight of the 27 mesial TLE
patients and all patients with neocortical TLE (16/16) or
ETLE (18/18).
The mean postoperative follow-up period was 30 ⫾ 10.3
months and the classification of postsurgical outcome was
evaluated according to the Engel’s classification (Engel et al.,
1993). Forty-seven patients (77.0%) were seizure free (class
I). Ten patients (16.4%) were almost seizure-free except for
rare disabling seizures since surgery (class II). Four patients
(6.6%) had a worthwhile reduction of seizures (class III). The
tissue pathology was abnormal in all patients. Hippocampal
sclerosis was found in 24 patients (39.4%), gliosis and/or
cortical dyslamination in 20 patients (32.8%), cortical
dysplasia in 15 patients (24.6%) and oligodendroglioma in
one patient (1.6%).
The concordance rates of seizure localization to
the final epileptogenic focus
The concordance rate of seizure localization was 68.9% (42/
61) by conventional visual inspection of interictal and ictal
SPECT. The concordance rates of subtracted SPECT were
85.2% (52/61) in hyperperfusion images, 23.0% (14/61) in
hypoperfusion images and 91.8% (56/61) in hyperperfusion–
hypoperfusion combined analysis (Table 2). The concordance
rates of subtracted SPECT for both neocortical TLE and
ETLE were significantly higher than those of conventional
visual inspection of interictal and ictal SPECT (P ⬍ 0.05 by
McNemar’s χ2-test). The third reviewer was used in 34.4%
Patterns of ictal perfusion changes in
subtracted SPECT
The ictal perfusion patterns of subtracted SPECT images
were classified as follows: (i) focal hyperperfusion, (ii)
hyperperfusion-plus, (iii) combined hyperperfusion–
hypoperfusion, (iv) focal hypoperfusion only, and (v) nonlocalizing or non-lateralizing (Table 4). Focal hyperperfusion
was defined as increased blood flow observed only in
the epileptic focus. Hyperperfusion-plus was defined as
hyperperfusion seen in the epileptic focus and adjacent
or contralateral cortical regions. Combined hyperperfusion–
hypoperfusion was the hypoperfusion area as well as
hyperperfusion area observed in epileptic focus at the same
subtracted SPECT image. Focal hypoperfusion implies only
decreased perfusion without hyperperfusion, which is seen
in the epileptic focus.
In TLE patients, focal hyperperfusion pattern was found
in 24/43 (55.8%) cases. Perfusion patterns in TLE could be
subdivided as follows (Table 5): (i) ipsilateral mesial temporal
only (Fig. 1A), (ii) ipsilateral lateral temporal hyperfusion
only (Fig. 1B), (iii) ipsilateral mesial and lateral temporal
hyperperfusion (Fig. 1C), and (iv) bilateral temporal
hyperperfusion (Fig. 1D). Hyperperfusion-plus in TLE (9/43,
21.0%) sometimes showed ipsilateral temporal and adjacent
frontal or even parietal hyperperfusion, but more commonly
appeared as bilateral temporal hyperperfusion with ipsilateral
temporal predominance that was found in 8/43 (18.6%)
patients [4/27 (14.8%) in mesial TLE and 4/16 (25.0%) in
neocortical TLE]. The most common perfusion pattern of
TLE was focal hyperperfusion of both ipsilateral mesial and
lateral temporal regions in 17/43 (39.6%) cases [11/27
(40.8%) in mesial TLE and 6/16 (37.5%) in neocortical
2154
H. W. Lee et al.
Subtracted SPECT in partial epilepsy
2155
combined hyperperfusion–hypoperfusion in 5/18 (27.8%).
Focal hypoperfusion with radiotracer injection during the
ictal period was seen in 3/18 (16.6%) cases, one of whom is
illustrated (Fig. 1G).
Overall, four of 61 patients (6.6%) showed only focal
hypoperfusion in epileptic foci (Table 7). One of them was
neocortical TLE and the remaining three were ETLE. No
mesial TLE patient showed focal hypoperfusion in
subtracted SPECT.
TLE]. Focal hyperperfusion in the mesial temporal region
alone was found in 9/27 (33.3%) of mesial TLE and focal
lateral temporal hyperperfusion in 4/16 (25.0%) of neocortical
TLE. Bilateral temporal hyperperfusion was found in both
mesial and neocortical TLE patients (Fig. 1C), in 4/27
(14.8%) and 4/16 (25.0%) patients, respectively. A combined
hyperperfusion–hypoperfusion pattern was observed in 5/43
(11.6%) of TLE patients; 2/27 (7.4%) in mesial TLE and 3/
16 (18.7%) in neocortical TLE (Table 6). One of these
patients with the combined hyperperfusion–hypoperfusion
pattern has been illustrated (Fig. 1E). He was a 24-year-old
man who had suffered frequent complex partial seizures or
secondarily generalized seizures for ⬎17 years. The brain
MRI revealed cortical dysplasia in the left posterior lateral
and basal temporal cortex. The EEG showed initial ictal
rhythm on the left posterior lateral temporal region and
propagation of ictal EEG discharges to the anterior temporal
area. The radioisotope injection was performed 50 s before
the end of clinical seizure, and the EEG discharges showed
build-up in the bilateral anterior temporal regions with left
side predominance at that time. On subtracted SPECT, ictal
hypoperfusion was found at the left posterior temporal cortex
while ictal hyperperfusion was seen at the left anterior
temporal region. In TLE, only one patient (2.3%) who
had neocortical TLE showed focal hypoperfusion without
hyperperfusion at the epileptic focus despite being injected
with radioisotope during the ictal period.
In ETLE, focal hyperperfusion (Fig. 1F) was observed
in 8/18 (44.4%), hyperperfusion-plus in 1/18 (5.6%), and
Clinical factors related to ictal perfusion
patterns of subtracted SPECT
Locations of seizure foci (temporal and extratemporal
origins), types of tissue pathology (hippocampal sclerosis,
cortical dysplasia and others), and the time of the radioisotope
injection were tested in their relationship with ictal
perfusion patterns.
The numbers of patients showing a hypoperfusion area
in epileptic foci (either focal hypoperfusion or combined
hyperperfusion–hypoperfusion patterns) were significantly
higher in ETLE than TLE patients (P ⬍ 0.05 by Fisher’s
exact test) (Table 4).
With respect to tissue pathology, ictal hypoperfusion of
subtracted SPECT was found significantly less frequently in
patients with hippocampal sclerosis than in patients with
other pathologies such as cortical dysplasia, gliosis and
cortical dyslamination (P ⬍ 0.05 by χ2-test) (Table 8). In
Table 2 The concordance rates of seizure localization in visual inspection of interictal–ictal
SPECT, hyperperfusion images, hypoperfusion images and hyperperfusion–hypoperfusion
combined analysis by subtracted SPECT-MRI co-registration (n ⫽ 61)
Subgroup (n)
Visual inspection (%)
Subtracted SPECT (%)
Hyperperfusion
MTLE (27)
NTLE (16)*
ETLE (18)*
Total (61)*
22
8
12
42
(81.5)
(50.0)
(66.7)
(68.9)
24
14
14
52
(88.9)
(87.5)
(77.8)
(85.2)
Hypoperfusion
Combined
2
4
8
14
24
15
17
56
(7.4)
(25.0)
(44.4)
(23.0)
(88.9)
(93.8)
(94.4)
(91.8)
MTLE ⫽ mesial TLE; NTLE ⫽ neocortical TLE; ETLE ⫽ extratemporal lobe epilepsy. *P ⬍ 0.05 by
McNemar’s χ2-test. The concordance rates of seizure localization in interictal–ictal combined analysis
by subtracted SPECT were significantly higher than the conventional interpretation of interictal and ictal
SPECT in NTLE and ETLE, but not in MTLE.
Fig. 1 Various perfusion patterns of subtracted SPECT in TLE and ETLE patients are illustrated. The perfusion patterns of subtracted
SPECT in TLE are (A) hyperperfusion in only mesial temporal region of right mesial TLE, (B) hyperperfusion in lateral temporal region
of a left neocortical TLE, (C) hyperperfusion in both mesial and lateral temporal areas in a right mesial TLE, and (D) hyperperfusionplus pattern of subtracted SPECT: bilateral temporal hyperperfusion with left predominance in a left neocortical TLE. (E-1 and E-2)
Combined hyperperfusion–hypoperfusion pattern of subtracted SPECT; E-1 is a positive image showing hyperperfusion in left anterior
temporal region and E-2 is a negative image showing hypoperfusion in left posterior and basal lateral temporal cortex from the same
interictal and ictal SPECT images. (F and G) Perfusion patterns of subtracted SPECT in ETLE patients; (F) focal hyperperfusion in a
left supplementary motor area seizure and (G) focal hypoperfusion in a right frontal lobe epilepsy. B-2, F-2 and G-2 are conventional
ictal SPECT of patients in B, F and G while B-1, F-1 and G-1 are conventional interictal SPECT of patients in B, F and G.
2156
H. W. Lee et al.
the group with hippocampal sclerosis, only one patient
showed ictal hypoperfusion with hyperperfusion and none
revealed only ictal hypoperfusion in subtracted SPECT.
The relationship between ictal perfusion patterns and the
time of the radiotracer injection is shown in Fig. 2. Sometimes
unexpected coupling could be found between the perfusion
pattern and the injection time. Four patients (6.6%) who had
the tracer injection postictally showed focal hyperperfusion
at the seizure focus (Table 9). Another four patients (6.6%)
who had the injection during the early ictal period showed
Table 3 The concordance rates of seizure localization in
MRI, interictal FDG-PET and subtraction SPECT (n ⫽ 52)
MRI
FDG-PET
Subtracted SPECT
Hyperperfusion image
Hypoperfusion image
Combined analysis*
Concordant
(%)
Non-concordant/
non-localizing (%)
30 (57.7)
33 (63.5)
22 (42.3)
19 (36.5)
44 (84.6)
7 (13.5)
47 (90.4)
8 (15.4)
45 (86.5)
5 (9.6)
*P ⬍ 0.01 by McNemar’s χ2-test. The seizure localization
concordance rate of combined analysis by subtracted SPECT was
significantly higher than that of MRI or interictal FDG-PET.
Table 4 Ictal perfusion patterns of subtracted SPECT and
locations of epileptic focus (temporal and extratemporal
origins) (n ⫽ 61)
Focal hyperperfusion
Hyperperfusion-plus
Combined hyperperfusion–hypoperfusion*
Focal hypoperfusion*
Non-localizing/non-lateralizing
TLE (%)
ETLE (%)
24
9
5
1
4
8
1
5
3
1
(55.8)
(21.0)
(11.6)
(2.3)
(9.3)
(44.4)
(5.6)
(27.8)
(16.6)
(5.6)
TLE ⫽ temporal lobe epilepsy (n ⫽ 43); ETLE ⫽ extratemporal
lobe epilepsy (n ⫽ 18). *P ⬍ 0.05 by Fisher’s exact test. The
number of combined hyperperfusion–hypoperfusions and focal
hypoperfusion was significantly higher in the ETLE than the TLE
group.
only focal hypoperfusion of subtracted SPECT in epileptic
focus. Seven patients (11.5%) who had the injection during
the ictal period revealed the combined hyperperfusion–
hypoperfusion pattern.
Discussion
Recent studies have demonstrated that the SPECT subtraction
enhanced the seizure localization rate to between 66.7 and
100% (Zubal et al., 1995; O’Brien et al., 1998). Our study
showed that SPECT subtraction improved the localization
rate (91.8%) of epileptic focus and provided us with an
accurate anatomical orientation of ictal perfusion changes
through SPECT-MRI co-registration. In addition to ictal
hyperperfusion, a new finding of ‘ictal hypoperfusion’ in
subtracted SPECT further improved the seizure localization
of ictal SPECT. In mesial TLE, subtracted SPECT had a
higher sensitivity of seizure localization than conventional
ictal SPECT, but no statistical significance was observed. In
neocortical TLE and ETLE, however, the seizure localization
rates of subtracted SPECT were significantly higher than the
conventional visual inspection of interictal and ictal SPECT.
With the improvement of the SPECT subtraction technique,
the value of hypoperfusion at the epileptic focus has been
suggested by other workers to be only in postictal SPECT
scans (Spanaki et al., 1998; O’Brien et al., 1999). They
reported that the use of SISCOM (subtraction SPECT coregistered to MRI) to detect focal cerebral hypoperfusion in
addition to focal hyperperfusion improved the sensitivity and
specificity of postictal SPECT in patients with intractable
partial epilepsy. In their postictal SPECT study, significantly
higher proportions of hyperperfusion SISCOM images
(65.7%), hypoperfusion SISCOM images (74.3%) and combined hyperperfusion–hypoperfusion SISCOM evaluations
(82.9%) were localizing than by using the conventional
method of side-by-side visual comparison of unsubtracted
images (31.4%, P ⬍ 0.0001) (O’Brien et al., 1999).
Our study was the first attempt to sort the ictal perfusion
patterns by subtracted SPECT–MRI co-registration and to
relate them to clinical factors in both TLE and ETLE. For
postictal hyperperfusion shown in four of the patients of our
Table 5 Patterns of ictal perfusion changes in mesial and neocortical TLE patients
Mesial temporal hyperperfusion*
Lateral temporal hyperperfusion*
Mesial and lateral temporal hyperperfusion
Bilateral temporal hyperperfusion
Hypoperfusion only
Non-localizing/non-lateralizing
Mesial TLE
n ⫽ 27 (%)
Neocortical TLE
n ⫽ 16 (%)
9 (33.3)
0 (0.0)
11 (40.8)
4 (14.8)
0 (0.0)
3 (11.1)
0
4
6
4
1
1
(0.0)
(25.0)
(37.5)
(25.0)
(6.3)
(6.3)
*P ⬍ 0.05 by Fisher’s exact test. Mesial temporal hyperperfusion alone was found only in mesial TLE
and lateral temporal hyperperfusion only in neocortical TLE, while the incidences of other ictal
perfusion patterns were not significantly different between mesial TLE and neocortical TLE.
Subtracted SPECT in partial epilepsy
study, one had a mesial TLE whereas the other three had
ETLE. Another four patients showed focal hypoperfusion at
the epileptic focus even though the radiotracer was injected
during the early ictal period. Subtracted SPECT of 10 other
patients revealed a hypo- as well as a hyperperfusion area
during the ictal period. Postictal hyperperfusion was reported
2157
in previous studies (Rowe et al., 1989, 1991; Newton et al.,
1997). However, the ictal hypoperfusion phenomenon has
been neglected because only ictal hyperperfusion was focused
and the subtraction technique was unavailable.
The ictal perfusion patterns of TLE were subdivided in
the present study. The most common perfusion pattern of
Table 6 The clinical data of the patients with a combined hyperperfusion–hypoperfusion pattern on subtracted SPECT
No.
1
2
3
4
5
6
7
8
9
Injection time (s)
SPECT subtraction
Surgery
From onset
To end
Hyper
Hypo
Resection
⫹49
⫹40
⫹65
⫹30
⫹30
⫹19
⫹21
⫹44
⫹7
⫹34
⫹26
⫹94
⫹50
⫹45
⫹124
⫹51
⫹141
⫹21
Rt ant. mT
Lt ant. mT
Rt mid-lat.T
Lt ant-lat.T
Lt TP
Lt mFP
Rt OF
Lt mFP
Post-lat.T
Rt mid-lat. T
Lt mid-lat. T
Rt post. TO
Lt post. T
Lt P
Lt lat. FP
Rt F
Lt FP
Lt P
Rt mT
Lt mT
Rt TO
Lt T
Lt CP
Lt CP
Rt OF
Lt FP
Lt PO
Pathology
HS
Gliosis
CD
CD
Cort.dysl, gliosis
CD
Cort.dysl, gliosis
CD
CD
Outcome*
I-A
I-A
I-A
I-A
II-A
II-A
I-A
I-A
II-B
No. ⫽ patient number; From onset ⫽ injection time – seizure onset; To end ⫽ seizure end – injection time (⫹ means the injection
was done during ictal period); Hyper ⫽ area showing increased blood flow; Hypo ⫽ area showing decreased blood flow in
subtracted SPECT; Rt ⫽ right; Lt ⫽ left; T ⫽ temporal; F ⫽ frontal; P ⫽ parietal; CP ⫽ centroparietal; TP ⫽ temporoparietal;
FP ⫽ frontoparietal; TO ⫽ temporo-occipital; PO ⫽ parieto-occipital; OF ⫽ orbitofrontal; m ⫽ mesial; ant. ⫽ anterior; post. ⫽
posterior; lat. ⫽ lateral; mid-lat. ⫽ mid-lateral; ant-lat. ⫽ anterolateral; post-lat. ⫽ posterolateral; HS ⫽ hippocampal sclerosis;
Cort.dysl ⫽ cortical dyslamination; CD ⫽ cortical dysplasia. *Engel’s classification (Engel et al., 1993); outcome at minimum 12
months’ follow-up.
Table 7 The clinical data of the patients with ictal hypoperfusion pattern on subtracted SPECT
No.
1
2
3
4
Injection time (s)
SPECT subtraction
Surgery
From onset
To end
Hyper
Hypo
Resection
Pathology
Outcome*
⫹19
⫹27
⫹26
⫹27
⫹55
⫹42
⫹37
⫹34
NL
NL
NL
NL
Lt post-lat.T
Rt F
Rt F
Rt P
Lt post-lat.T
Rt OF⫹lat.F
Rt lat.F
Rt FP
Gliosis
Gliosis
FCD
Cort.dysl
I-B
I-A
I-A
II-A
No. ⫽ patient number; From onset ⫽ injection time – seizure onset; To end ⫽ seizure end – injection time (⫹ means the injection was
done during ictal period); Hyper ⫽ area showing increased blood flow; Hypo ⫽ area showing decreased blood flow in subtracted
SPECT; Rt ⫽ right; Lt ⫽ left; T ⫽ temporal; F ⫽ frontal; P ⫽ parietal; FP ⫽ frontoparietal; OF ⫽ orbitofrontal; lat. ⫽ lateral;
post-lat. ⫽ posterolateral; NL ⫽ non-lateralizing or non-localizing; FCD ⫽ focal cortical dysplasia; Cort.dysl ⫽ cortical dyslamination.
*Engel’s classification (Engel et al., 1993); outcome at minimum 12 months’ follow-up.
Table 8 Ictal perfusion patterns of subtracted SPECT in different pathology (HS; CD and
others)
Focal hyperperfusion
Hyperperfusion-plus
Combined hyperperfusion–hypoperfusion*
Focal hypoperfusion*
Non-localizing
HS
n ⫽ 24 (%)
CD
n ⫽ 15 (%)
Others
n ⫽ 20 (%)
15
6
1
0
2
9
1
4
1
0
7
2
5
3
3
(62.5)
(25.0)
(4.2)
(0.0)
(8.3)
(60.0)
(6.7)
(26.7)
(6.7)
(0.0)
(35.0)
(10.0)
(25.0)
(15.0)
(15.0)
HS ⫽ hippocampal sclerosis; CD ⫽ cortical dysplasia; Others ⫽ gliosis/cortical dyslamination.
*P ⬍ 0.05 by χ2-test; the incidences of combined hyperperfusion–hypoperfusion and focal
hypoperfusion were significantly lower in HS than CD and others.
2158
H. W. Lee et al.
mesial and neocortical TLE was hyperperfusion of both
ipsilateral mesial and lateral temporal regions. Focal hyperperfusion of mesial temporal region was observed only in
mesial TLE and focal lateral temporal hyperperfusion was
seen only in neocortical TLE. This observation could be an
important finding in differentiating neocortical from mesial
TLE. In a previous study, some differences of perfusion
patterns on ictal SPECT in subgroups of TLE were also
reported (Ho et al., 1996). These authors observed that TLE
with hippocampal sclerosis or foreign-tissue lesion in the
mesial temporal lobe showed ictal hyperperfusion in the
ipsilateral mesial and lateral temporal regions, whereas TLE
with foreign-tissue lesion in the lateral temporal lobe had
ictal hyperperfusion in bilateral temporal lobes with ipsilateral
predominance.
The mechanism for postictal hyperperfusion and ictal
hypoperfusion is not, as yet, completely understood. Only in
mesial TLE have two factors been suggested for the ictal
Fig. 2 The ictal perfusion patterns of subtracted SPECT versus
the injection time of radiotracer. The percentage injection time
means the percentage of total seizure duration (‘–100’ means the
seizure onset time and ‘0’ is the seizure end time). A negative
value indicates an ictal period while a positive value indicates a
postictal period. F-Hyper ⫽ focal hyperperfusion; Hyper-plus ⫽
hyperperfusion-plus; Combined ⫽ combined hyperperfusion–
hypoperfusion; F-Hypo ⫽ focal hypoperfusion.
and postictal temporal hyperperfusion (Newton et al., 1992).
The increased peri-ictal regional cerebral blood flow has
been associated with the increased extracellular cations (H⫹
and K⫹) in experimental systems (Leniger-Follert, 1984) and
with increased non-oxidative glucose metabolism in human
and experimental hippocampal seizures (Fox et al., 1988;
Ackermann and Lear, 1989). Our results, however, can
provide some explanation of the new finding of ‘ictal hypoperfusion’. In the patient illustrated in Fig. 1E, the initial seizure
onset zone (posterior temporal area) showed hypoperfusion
in subtracted SPECT whereas the zone with a later active
build-up of ictal discharges (anterior temporal region) had
hyperperfusion. From this observation, ictal hypoperfusion
at the epileptic focus can be due to ‘intra-ictal early exhaustion
of an initial seizure focus’ or a ‘steal phenomenon’ from
neighbouring brain regions where ictal discharges are
propagated. It is well known that the ictal SPECT represents
the perfusion status of the brain when the radiotracer is
injected. However, even during the ictal period, the regional
cerebral blood flow continues to change because ictal EEG
discharges propagate and change in their amplitude and
frequency. Therefore, the whole picture of ictal perfusion
changes may not be sufficiently described by the interpretation
of ictal hyperperfusion alone.
Several factors may be related to the determination of ictal
perfusion patterns; the speed of ictal propagation to adjacent
brain regions, the intensity of ictal discharges, the area of
the brain with active ictal discharges at the time of injection,
the energy status of different brain regions including the
epileptic focus, and the metabolic state of the epileptic focus
when the radioisotope reaches the brain tissue. Previous
studies accepted the injection timing of radiotracer as an
important factor affecting results of an ictal SPECT study
(Rowe et al., 1991; Newton et al., 1992, 1997; Spanaki et al.,
1998; O’Brien et al., 1999). The location of seizure focus
and the tissue pathology have also appeared to be important
in determining the perfusion patterns or in some cases even
more important than the injection timing itself. In our study,
the injection times of the radiotracer were interpreted after
they were normalized, as the duration of individual seizures
was highly variable (range, 6–209 s; mean, 52 ⫾ 39.6 s). Of
course, the absolute duration of an individual seizure may
Table 9 The clinical data of the patients with a postictal hyperperfusion pattern on subtracted SPECT
No.
1
2
3
4
Injection time (s)
SPECT subtraction
Surgery
From onset
To end
Hyper
Hypo
Resection
Pathology
Outcome
⫹77
⫹31
⫹10
⫹29
⫺4
⫺12
⫺11
⫺5
Lt O
Rt T
Rt F
Lt F
NL
NL
NL
NL
Lt TO
Rt mT
Rt F
Lt F
CD
HS
Gliosis
Gliosis
I-A
I-A
II-A
III-A
No. ⫽ patient number; From onset ⫽ injection time – seizure onset; To end ⫽ seizure end – injection time (– means the injection was
done during postictal period); Hyper ⫽ area showing increased blood flow; Hypo ⫽ area showing decreased blood flow in subtracted
SPECT; Rt ⫽ right; Lt ⫽ left; T ⫽ temporal; F ⫽ frontal; O ⫽ occipital; TO ⫽ temporo-occipital; m ⫽ mesial; CD ⫽ cortical
dysplasia; HS ⫽ hippocampal sclerosis. *Engel’s classification (Engel et al., 1993); outcome at minimum 12 months’ follow-up.
Subtracted SPECT in partial epilepsy
be important as the radiotracer only enters the brain 30–60 s
after the injection. If the seizure was very brief, it might take
some time, even twice as long as the duration of a seizure,
for it to enter into the brain and the perfusion to reverse
itself. Tables 6, 7 and 9, providing details about absolute
times of the radiotracer injection for individual patients,
illustrate that in several cases the perfusion patterns were not
related to the injection timing. At least in these cases, the
type of the seizure, location of seizure onset with the cortical
connectivity of that area, the rapidity of seizure propagation
as well as the tissue pathology might significantly influence
the perfusion changes.
In summary, subtracted SPECT–MRI co-registration
revealed that ictal perfusion changes during partial seizures
were various and showed hypo- as well as hyperperfusion
during the ictal period. Use of the SPECT subtraction
technique and careful interpretation of SPECT with other
clinical features are recommended for more accurate
localization of epileptic focus.
References
Ackermann RF, Lear JL. Glycolysis-induced discordance between
glucose
metabolic
rates
measured
with
radiolabeled
fluorodeoxyglucose and glucose. J Cereb Blood Flow Metab 1989;
9: 774–85.
Chang L. A method for attenuation correction in computed
tomography. IEEE Trans Nucl Sci 1987; NS-25: 638–43.
Engel J Jr, Van Ness PC, Rasmussen TB, Ojemann LM. Outcome
with respect to epileptic seizures. In: Engel J Jr, editor. Surgical
treatment of the epilepsies. 2nd ed. New York: Raven Press; 1993.
p. 609–21.
Fox PT, Raichle ME, Mintun MA, Dence C. Nonoxidative glucose
consumption during focal physiologic neural activity. Science 1988;
241: 462–4.
Harvey AS, Hopkins IJ, Bowe JM, Cook DJ, Shield LK, Berkovic SF.
Frontal lobe epilepsy: clinical seizure characteristics and localization
with ictal 99mTc-HMPAO SPECT. Neurology 1993; 43: 1966–80.
Ho SS, Berkovic SF, Newton MR, Austin MC, McKay WJ, Bladin
PF. Parietal lobe epilepsy: clinical features and seizure localization
by ictal SPECT. Neurology 1994; 44: 2277–84.
Ho SS, Berkovic SF, Berlangieri SU, Newton MR, Egan GF,
Tochon-Danguy HJ, et al. Comparison of ictal SPECT and interictal
PET in the presurgical evaluation of temporal lobe epilepsy. Ann
Neurol 1995; 37: 738–45.
Ho SS, Berkovic SF, McKay WJ, Kalnins RM, Bladin PF. Temporal
lobe epilepsy subtypes: differential patterns of cerebral perfusion
on ictal SPECT. Epilepsia 1996; 37: 788–95.
Jack CR Jr, Mullan BP, Sharbrough FW, Cascino GD, Hauser MF,
2159
Krecke KN, et al. Intractable nonlesional epilepsy of temporal lobe
origin: lateralization by interictal SPECT versus MRI. Neurology
1994; 44: 829–36.
Landis JR, Koch GG. The measurement of observer agreement for
categorical data. Biometrics 1977; 33: 159–74.
Leniger-Follert E. Mechanisms of regulation of cerebral microflow
during bicuculline-induced seizures in anaesthetized cats. J Cereb
Blood Flow Metab 1984; 4: 150–65.
Marks DA, Katz A, Hoffer P, Spencer SS. Localization of
extratemporal epileptic foci during ictal single photon emission
computed tomography. Ann Neurol 1992; 31: 250–5.
Newton MR, Berkovic SF, Austin MC, Rowe CC, McKay WJ,
Bladin PF. Postictal switch in blood flow distribution and temporal
lobe seizures. J Neurol Neurosurg Psychiatry 1992; 55: 891–4.
Newton MR, Berkovic SF, Austin MC, Rowe CC, McKay WJ,
Bladin PF. SPECT in the localisation of extratemporal and temporal
seizure foci. J Neurol Neurosurg Psychiatry 1995; 59: 26–30.
Newton MR, Berkovic SF. Interictal, ictal, and postictal singlephoton emission computed tomography. In: Cascino GD, Jack CR
Jr, editors. Neuroimaging in epilepsy: principles and practice.
Boston: Butterworth-Heinemann; 1996. p. 177–91.
O’Brien TJ, So EL, Mullan BP, Hauser MF, Brinkmann BH, Bohnen
NI, et al. Subtraction ictal SPECT co-registered to MRI improves
clinical usefulness of SPECT in localizing the surgical seizure focus.
Neurology 1998; 50: 445–54.
O’Brien TJ, So EL, Mullan BP, Hauser MF, Brinkmann BH, Jack
CR Jr, et al. Subtraction SPECT co-registered to MRI improves
postictal SPECT localization of seizure foci. Neurology 1999; 52:
137–46.
Rowe CC, Berkovic SF, Sia ST, Austin M, McKay WJ, Kalnins
RM, et al. Localization of epileptic foci with postictal single photon
emission computed tomography. Ann Neurol 1989; 26: 660–8.
Rowe CC, Berkovic SF, Austin MC, McKay WJ, Bladin PF. Patterns
of postictal cerebral blood flow in temporal lobe epilepsy: qualitative
and quantitative analysis. Neurology 1991; 41: 1096–103.
Spanaki MV, Spencer SS, Wisniewski G, MacMullan J, Seibyl J,
Zubal IG. Evolution and localization of postictal blood flow changes
in partial seizures demonstrated by SPECT: use of quantitative
difference images. J Epilepsy 1998; 11: 25–33.
Spencer SS. The relative contributions of MRI, SPECT, and PET
imaging in epilepsy. [Review] Epilepsia 1994; 35 Suppl 6: S72–89.
Zubal IG, Spencer SS, Imam K, Seibyl J, Smith EO, Wisniewski
G, et al. Difference images calculated from ictal and interictal
technetium-99m-HMPAO SPECT scans of epilepsy. J Nucl Med
1995; 36: 684–9.
Received April 11, 2000. Revised June 10, 2000.
Accepted June 15, 2000.