Neuron loss localizes human temporal lobe epilepsy by in vivo proton magnetic resonance spectroscopic imaging

Neuron Loss Locahzes Human Temporal Lobe Epilepsy by In Vivo Proton Magnetic Resonance Spectroscopic Imaging James W. Hugg, PhD,*tll Kenneth D. Laxer, MD,S Gerald B. Matson, PhD,*$ Andrew A. Maudsley, P h D , W and Michael W. Weiner, MD"t1I

Temporal lobe epileptogenic foci were blindly localized in 8 patients with medically refractory unilateral complex partial seizures using noninvasive in vivo proton magnetic resonance spectroscopic imaging ('H-MRSI) with 4-ml effective voxel size. The brain proton metabolite signals in 8 matched normal controls were bilaterally symmetrical within -+ 10%. The hippocampal seizure foci had 21 -t 5% less N-acetyl aspartate signal than the contralateral hippocampal formations ( p < 0.01). The focal N-acetyl aspartate reductions were consistent with pathology findings of mesial temporal sclerosis with selective neuron loss and gliosis in the surgically resected epileptogenic foci. Proton MRSI correctly localized the seizure focus in all 8 cases. By comparison, MR imaging correctly localized 7 of 8 cases and single photon emission computed tomography correctly localized 2 of 5 cases. No lactate was detected in these interictal studies. No significant changes in choline or creatine were observed. In conclusion, 'H-MRSI is a useful tool for the noninvasive clinical assessment of intractable focal epilepsy. These preliminary results suggest that 'H-MRSI can accurately localize temporal lobe epileptogenic foci. Hugg JW, Laxer KD, Matson GB, Maudsley AA, Weiner MW. Neuron loss localizes human temporal lobe epilepsy by in vivo proton magnetic resonance spectroscopic imaging. Ann Neurol 1993;14.788-794

Presurgical localization of intractable epileptogenic foci is often difficult. Anatomical anomalies are often found by magnetic resonance imaging (MRI) studies, and functional anomalies are often found in positron emission tomographic (PET) and single photon emission computed tomographic (SPECT) studies. Nevertheless, electroencephalographic (EEG) electrode implantation is required for up to 50% of patients with partial epilepsies because neurological imaging and scalp EEG often disagree or are inconclusive [l-31. The need for invasive EEG recordings could be reduced by more sensitive imaging. One approach to presurgical localization is to observe the abnormal interictal metabolism of the seizure focus. We reported three studies of intractable temporal lobe epilepsy using noninvasive phosphorus magnetic resonance spectroscopy methods to localize epileptogenic foci [4-61. All 18 patients were blindly localized by observing interictally increased intracellular p H and inorganic phosphate and decreased phosphomonoesters in the region of the seizure foci. Other investigators have also reported interictally decreased phosphocreatine (PCr)/inorganic phosphate (Pi) in the temporal lobe ipsilateral to the seizure foci {7}.

Another approach to localization, reported here, is to observe a marker of the neuron loss characteristic of intractable focal epilepsy [8}. Mesial temporal sclerosis (MTS, neuron loss and gliosis) is found in the resected hippocampal formation of up to 70% of patients receiving temporal lobectomies for intractable complex partial seizures t9, lo}. Neoplastic or vascular lesions comprise most other pathological findings. AT-acetyl aspartate (NAA) is a putative specific neuron marker not found in mature glial cells {11-15}. Decreased NAA has been found in seizure foci corresponding with neuron loss [ 16-18}. To detect decreased NAA and, thus, neuron loss, we used noninvasive in vivo proton magnetic resonance spectroscopic imaging ('H-MRSI) { 19, ZO}. MRSI is analogous to conventional magnetic resonance imaging (MRI) except that the signal intensity in each voxel is proportional to the concentration of protons in metabolites, rather than the protons in water and lipid. Because metabolite concentrations are about 10,000 times less than water, the spatial resolution (effective voxel sizes of 1 to 4 ml) and signal-to-noise ratio of metabolite images are considerably less than water to lipid MR images. MRSI simultaneously ob-

From the 'MR Unit, Department of Veterans Affam Medical Center, and Departments of ?Radiology, $Neurology, §Pharmaceutical Chemistrv, and IIMedicine, University of California, San Francisco, CA

Address correspondence to Dr Weiner, MR Unit, Department of Veterans Affairs Medical Center, 4150 Clement St (1lM), San Francisco, CA 94121

Received Feb 19, 1993, and in revised form Jun 15 Accepted for publication Jun 15, 1993

TPresent address Neurology Department, Henry Ford Health SCIence Center, 2795)W Grand Blvd, Detroit, MI 48202.2689

788 Copyright 0 1093 by the American Neurological Association

Focal EpilepJy Patient Demographics and Localization _____

Patient No./ SexIAge (yr)

Duration (yr)

llFi37

2lM120

9 19

31Ml22 4iM119

19 18

5iM141 61Mi34

22

3

?IF136

34

8IFi39

20

Etiology None Birth injury? Birth injury? Birth injury? febrile Sz! None None Meningitis Mumps, encephalitis

EEG

NAA(g) IpsilContra

'HMRSI

+

LT

82

RT

79

RT LT

81 70

RT RT LT RT

85 78 81 78

+ + + +

+ +

MRI

SPECT

+

-

Ganglioglioma

sz 1

Sz free

+

Ganglioglioma MTS

Sz free Sz free

ND ND ND

MTS ND MTS MTS

Sz free ND Sz free Sz free

+

+ +

Surgical Outcome

MTS

+ + +

+

+

Pathology

X

>90%

100%

True Positive ( +) False Positive ( X ) False Negative ( - )

OV 0%

EEG = electroencephalogram; MRSI = magnetic resonance spectroscopic imaging; MRI = magnetic resonance imaging; SPECT = single photon emission computed tomography; L = left; K = right; T = temporal lobe; NAA (IpsiiConrra) = ipsilateralicontraiateral N-acetyl aspartate signal ratio; MTS = mesial temporal sclerosis; ND = not done; Sz = seizure; J, = decreased.

tains metabolite spectra from multiple voxels throughout the field of view. From these spectra metabolite images are reconstructed. 'H-MRSI using PRESS volume preselection has been implemented for clinical studies [2 l}. We developed spectroscopic imaging software for acquiring, reconstructing, and displaying metabolite images registered t o MR images and for selecting spectra f r o m multiple volumes of interest E22). Recent reports demonstrating t h e clinical feasibility of *H-MRSI have detected decreased NAA in chronic human brain infarcts {23-26}, intracranial tumors 12 1, 271, and chronic multiple sclerosis plaques [28), consistent with decreased neuronal density in t h e pathological regions. Therefore, t h e goal of this study was t o determine if 'H-MRSI could be used t o blindly localize temporal lobe epileptogenic foci b y observing reduced NAA signal.

tient had unilateral complex partial seizures refractory to medical treatment. In all patients the seizures arose from the mesial temporal region, documented by interictal and ictal recordings on scalp and, as necessary, subdural EEGivideo telemetry. Seizure localization required the recording of a minimum of four seizures with localized onsets consisting of either voltage attenuation or rhythmic spikes that began or preceded the clinical onsets. All patients demonstrated exclusively unilateral seizure onsets and unilateral discharge predominance interictally. The patients had been seizure free at least 24 hours prior to study, and no ictal event occurred during study. All patients were being considered for surgical resection of the epileptogenic tissue to treat their refractory seizure disorder. All studies were performed in vivo prior to lobectomy, which 7 of the 8 patients have now received. Electrocorticography, including acute depth EEG recordings, was performed at the time of surgery as a final confirmation of epileptogenic localization prior to resection.

Methods Haman Subjects

MRI and SPECT Stadiu

Sixteen subjects were studied between December 1990 and May 1991 including 8 patients (5 men, 3 women) with unilateral medically refractory temporal lobe complex partial seizures and 8 normal sex- and age-matched (within 5 5 yr) controls. The patients were sequential cases referred to the UCSF Comprehensive Epilepsy Center who met the selection criteria below. All studies were approved by UCSF Human Research Committee and written informed consent was obtained prior to each study. Demographic patient data are shown in the Table. The 9 years and the mean mean age of the patients was 31 duration of seizures was 18 2 9 years. Seizure etiology was unknown in all cases, although birth injury was considered a possibility in 3 cases with early onset. Patients were excluded from this study if they demonstrated focal MRI abnormalities that suggested neoplastic or vascular pathologies. Each pa-

*

A separate independent MRI study was performed (1.5 T Signa, General Electric, Milwaukee, WI) and interpreted by a neuroradiologist. An MRI finding of increased T2-weighted

signal or atrophy in one hippocampal formation was considered evidence for unilateral MTS [2, 31.Five of the patients were studied by '""Tc-HMPAO SPECT, in which a finding of interictal hypoperfusion was considered evidence for localization. The constantly observed patients had no seizures during SPECT studies.

Proton MRSI Study In all cases acquisition and analysis of proton MRSI studies were performed without knowledge of the location of the seizure focus. Proton MRSI studies were conducted with a Gyroscan S 15 whole body imaging and spectroscopy system (Philips Medical Systems, Shelton, CT) operated at 2 T. The MRSI studies were completed between December 1990 and Hugg et al: Neuron Loss in Epilepsy by MRSI

789

May 1971. For reduced motion of the subject’s head, a vacuum-molded head holder (Vac-Pac, Olympic Medical, Seattle, WA) and padded straps were employed. A saddle-type imaging headcoil was employed. Proton MR images of the head were first obtained: 16 sagittal locator slices (7.1 mm thick, 0.7 mm gap, TR = 500 msec, TE = 30 msec) and 16 transaxial slices (7.1 mm thick, 1.4 mm gap, T R = 2,500 msec, TE = 30 and 80 msec). These slices were angulated 20 degrees supraorbitally beyond the orbital-auditory meatal (OAM) plane observed on the sagittal locator slices to give a consistent anatomical perspective, facilitating comparisons between patients and controls. This slice orientation also reduced susceptibility artifact and lipid signal from the sphenoid sinus and orbital muscle regions in MRSI studies. The magnetic field was then shimmed (nonlocalized) using Philips computer-controlled optimization for a water resonance line width of less than 15 Hz (0.17 ppm at 86 MHz). Custom acquisition software was written to perform threedimensional (3D) gradient phase encoding for the ‘H-MRSI studies 1271. Using the MR images to guide the MRSI study, a volume of interest (VOI) was selected to encompass as much as possible the temporal lobe region and to avoid lipid in the skull, orbital muscles, and sphenoid bone. Typical VOI dimensions were 10 cm x 10 cm (transaxial plane) x 5 cm. A double spin-echo (PRESS) sequence [30, 3 l} was used for VOI localization in combination with water suppression and 3D gradient phase-encoding MRSI {29]. The VO1 localization sequence employed amplitude-modulated RF pulses in combination with slice selection gradients. Gradient spoiler pulses flanking the two 180-degree refocusing pulses were applied to reduce spurious transverse magnetization excited by imperfections in these pulses. The echo was sampled with a 1 kHz bandwidth beginning shortly after the last gradient pulse with 2 1% ( 108 msec) of the 5 12-msec sampling interval occurring before the echo center at TE = 272 msec. The optimal RF pulse length for ‘H-MRSI was determined by minimizing the water signal excited by the second refocusing pulse in the double spin-echo sequence in the absence of gradient spoiler pulses. The sequence was executed without water suppression for adjustment of the gradient spoiler pulses (“gradient tuning”) to maximize the water echo signal. Localized shimming was next performed, typically resulting in a water resonance line width of 6 Hz (0.07 ppm) or less. Two WEFC water suppression pulses { 3 2 ) preceded the start of the VOI localization sequence with delay times tl and t2that empirically minimized longitudinal water magnetization. The pulses were frequency-selective adiabatic inversion pulses with 50-Hz bandwidth centered on the water resonance. Gradient spoiler pulses dispersed spurious transverse magnetization excited by imperfect longitudinal inversion. The double WEFT sequence was more effective for suppressing the water resonance than a single WEFT sequence, presumably because the brain tissue and cerebrospinal fluid (CSF) water T1 relaxation approximated a double exponential. Typically, t , = 800 msec (ranging from 750 to 830 msec in various subjects) and t2 = 50 msec were used with a repetition time T R = 1.7 seconds. Between the two water suppression WEET pulses, three cosine-sinc outervolume-suppression pulses were applied to further reduce the skull lipid signal {29). Finally, a ’H-MRSI study was performed with 3D phase

790 Annals of Neurology

Vol 34

encoding. An ellipsoid in k-space with 16 x 16 x 12 gradient phase-encoding steps along principal axes was sampled over a 16 x 16 x 14.1-cm’ field of view { 3 3 } . The phaseencoding gradients were applied between the 70-degree excitation pulse and the first 180-degree refocusing pulse of the double spin-echo sequence. A single acquisition was acquired for each phase-encoding step. Total acquisition time was 45 minutes.

MRSI Processing and Display MRS images were reconstructed [22) using Fourier transforms with mild spatial gaussian apodization and zero-filling to 32 x 32 interpolated voxels in 12 transaxial planes (nominally 0.85 cm thick). The nominal in-plane resolution before spatial filtering was 1.0 cm and the nominal voxel size was 0.85 cm3. Computer simulation and MRS imaging of a 0.5 cm3phantom were used to determine the spatial point spread function, including the effects of elliptical k-space sampling, apodization, zero-filling, and Fourier reconstruction {333. At the full width at half maximum of the point spread function, the effective in-plane resolution was about 1.25 cm, the effective slice thickness was 1.1 cm, and the effective oblate elliptical voxel size was about 1.7 cmj. Total proton metabolite images were generated by total spectral integration during reconstruction. Additional images of individual proton metabolites were produced during interactive display and analysis by spectral integration over individual resonances. The MRS images were interpolated to 64 x 64 voxels for display. The display software provided simultaneous display and spatial registration of the MR and metabolite images {22). To demonstrate the anatomical location of metabolite distributions in MRS images, these images were often overlaid by the corresponding high-pass filtered T2-weighted proton MR image, which delineated highcontrast tissue edges, such as ventricles, sulci, and calvarium ( e g , Figure Ib).

Spectral Fitting Individual proton spectra selected from the full MRSI study data sets were fit by a least-squares method (NMR1 program, New Methods Research, Inc., Syracuse, NY). A 1-Hz exponential spectral line broadening was applied to optimize signal-to-noise ratio. Spectra were interpolated by zero-filling and Fourier transformation to a digital resolution of 0.5 Hz/ point. Choline (Cho), creatine + phosphocreatine (Cr), and NAA peaks were identified by their chemical shifts [GI. Proton magnitude spectra were fit with gaussian line shapes to determine metabolite chemical shifts and peak integrals.

Localization of Seizure Foci The higher spatial resolution proton MR image was used to select single effective voxel spectra centered on the hippocampal formation. The MR image recorded during the MRSI study was used only for anatomically consistent selection of spectra for bilateral symmetry comparisons. No attempt was made to use that MR image to localize the seizure focus and thus to potentially bias the MRSI results. The extracted spectra were labeled by coded file names so that the investigator fitting the spectra could not inadvertently use any knowledge of the MR images to bias the spectral fitting re-

No 6 December 1993

Fig 1 . (a) T2-weighted transaxial magnetic resonance imaging (MRI) scan of temporal lobe epilepsy of Patient 3 . The right hippocampalformation had increased signal intensity. A single photon emission computed tomographic study of this patient showed ltft temporal lobe hypoperfiusion ii.e., incorrect localizametabolite tion). (bj Corresponding N-acetyl aspartate map with a high-pass Jiltered MRI overlay. Note the low NAA signal in globw pallidus luhich contains iron) and oentrides. The PRESS z?olume (white box) was inadoertently placed a q w metric-ally. sults. Based on previous studies 116, 171 we expected a power 1-p >80% to detect a 20% unilateral decrease in NAA signal comparing 8 patients and matched controls using an unpaired, one-tailed t test with significance LY = 0.05.

Results The brains appeared grossly normal at surgery, but the mesial temporal structures were firm to suction during subpial resection. Increased astrocytes and decreased neurons were found in the resected dentate gyrus and Amrnion’s horn in all 7 who underwent surgery. MTS was found in 5 patients. In 2 patients unsuspected gangliogliomas were resected. Postoperatively (12 to 18 months) 6 of the 7 patients who underwent surgery have been seizure free (see Table). The proton metabolite maps were bilaterally symmetric (within 10%) in all controls and patients, except that significant asymmetries were observed in the patient hippocampal regions on NAA metabolite maps. N o lactate was detected in these interictal studies. Figure l a shows a T2-weighted transaxial MR image of temporal lobe epilepsy of Patient 3 in whom the r g h t hippocampus had increased signal intensity (i,e., correct localization). His SPECT study showed hypoperfusion in the left temporal lobe (i.e., false posi-

*

tive localization). Figure 1b shows the corresponding NAA metabolite map with a high-pass filtered MRI overlay. The NAA signal was 19% smaller in the right hippocampal region (i.e., correct localization). Spectra extracted from the two hippocampal 1.7-ml voxels indicated on Figure l a and b are shown in Figure 2a and b. In all 8 patients the interictal ipsilateral seizure foci had less NAA than the homologous contralateral region. The Table shows the ipsilateral/contralateral NAA signal ratios ranging from 70 to 85% with a mean of 79 k 5%. The other proton metabolite ratios in patients demonstrated bilateral symmetry (choline, 93 k 8%; creatine, 102 t 7%). All leftiright metabolite ratios in controls demonstrated bilateral symmetry (NAA, 98 t 5%; choline, 102 7%; creatine, 97 i 8%). Ten homologous bilaterally paired single voxel spectra were compared from each subject (hippocampal, posterior lateral temporal, anterior lateral temporal, posterior lateral frontal, posterior mesial frontal). The only significant asymmetry was observed in the patient hippocampal formation region NAA ratios. The patient ipsilateral/contralateral NAA ratios were significantly less than control lefdright NAA ratios in the hippocampal formation region ( p < 0.01). The NAA peak line widths in ipsilateral hippocampi (166 ? 56 ppb), contralateral hippocampi (138 k 29 ppb), and control hippocampi (127 k 25 ppb) were significantly greater ( p < 0.01) than in patient and control centrum semiovale (88 7 ppb). The accuracy of blinded localization of the epileptogenic focus by ‘H-MRSI, MRI, and SPECT is compared in the Table. Proton MRSI findings of less NAA signal in the ipsilateral interictal foci resulted in correct

*

*

Hugg ec al: Neuron Loss in Epilepsy by MRSI

791

'

A\

NAA

I\

b

Contralateral

a

3.0

2.0

2.5

1.5

PPM Fig 2. (a, 6) Spectra extracted fmm two 1.7-rnl voxels indicated on Figure 1 a and b. The N A A signal u'as 19% srnallev in the right epileptogenic hippacampal region.

localization of all 8 cases. MRI findings of increased T2 signal or atrophy of the hippocampal formation regions resulted in correct localization in 7 of 8 cases. "'"TcHMPAO SPECT findings of reduced perfusion interictally resulted in correct localization in 2 of 5 cases.

Discussion Selective N A A (Newon) Loss The major finding was that use of 'H-MRSI enabled

blinded localization of all 8 epileptogenic foci. Decreased NAA signal is consistent with MTS in the epileptogenic foci. Decreased NAA has also been observed in chronic infarcts 123-261, intracranial tumors 121, 273, and chronic multiple sclerosis plaques [28] where glial cells have proliferated to replace lost neurons. NAA is evidently localized in neuronal cytosol [ll-151, whereas choline and creatine are found in both neurons and glia [14, 151. NAA was reduced about 2 1% on average whereas total phosphorus metabolites were not reduced C4-6, 181, consistent with pathological findings of a greater loss of neurons than glial cells within epileptogenic foci [S-lO]. Stroke also causes a greater focal loss of neurons than glial cells, in part because surviving glia can proliferate, forming

792

Annals of Neurology Vol 34

No 6

scar tissue. In chronic stroke, NAA decreased to a greater extent than other metabolites E231, similar to the finding reported here for the epileptogenic focus in temporal lobe epilepsy. Therefore, T2-weighted MRI signal hyperintensity and hippocampal atrophy are less sensitive indicators of neuronal loss than 'H-MRSI because of selective neuronal vulnerability and subsequent glial hyperplasia. Decreased NAA signal could be due to longer T1 relaxation or shorter T2 relaxation times. The subtle T2-weighted MRI signal increase in the hippocampal formation of some cases of MTS is evidence for a -10% increased T2 of water. No significant changes of water T1 are associated with MTS. Therefore, it is unlikely that extraordinarily large changes in NAA relaxation times cause much of the decreased NAA signal. Nevertheless, localized metabolite relaxation measurements should be performed in both controls and patients. None of the patients had obvious dilated CSF spaces or atrophied hippocampi on MRI. N o significant asymmetry in total phosphorus metabolites has been observed C4-6, IS}. Therefore, it is unlikely that increased partial volume of CSF caused much of the decreased NAA signal. The finding of gangliogliomas in 2 of our patients was unexpected. They are the only such cases in our surgical series of over 150 temporal

December 1991

lobe patients. These 2 patients also showed neuronal loss in the resected hippocampi. All proton metabolite images of controls were relatively uniform throughout brain parenchyma with bilateral symmetry within & 10%. Marked deviations from bilateral symmetry were present only in the NAA maps of epilepsy patients. However, the contralateral region in patients used as an internal reference is unlikely to be strictly normal because bilateral neuron loss is possible in patients with predominantly unilateral epileptogenic foci. Techniques under development for molar quantitation of metabolite concentrations measured by proton MRSI should ultimately allow direct comparison of NAA concentrations with controls

line may vary. Residual lipid signal may mask the presence of lactate or interfere with the quantitation of NAA. By carefully shimming and tuning gradients, we minimized these problems.

(2) NAA PEAK SPECIFICITY. The NAA peak may also contain contributions from other molecules. The 2.03 ppm NAA peak appears to be specific for NAA in cortex, whereas NAA-glutamate contributes to the NAA peak in white matter [14). Contamination of hippocampal region spectra by inclusion of a parrial volume of surrounding white matter will dilute the magnitude of NAA reduction and lead to an underestimate of neuronal loss.

1341. N o lactate was detected in these interictal studies. Animal models have shown greatly altered energy metabolites associated with high metabolic expenditures during the ictus. Postictally, intracellular p H increased slowly from acidosis to normal and subsequently to alkalosis well before lactate had normalized {35J. Brain lactate may remain elevated for a considerable time postictally, presumably because glycolysis is continuously enhanced by the alkalotic stimulation of phosphofructokinase r36J. Interictal alkalosis C4-6, 181 is associated with neuronal hyperexcitability and may play a role in seizure pathogenesis. Alkalosis has also been associated with selective neuronal injury, in part by potentiating glutamate toxicity { 37). Glutamate concentrations are elevated in seizure foci (381. However, despite considerable research, the causes of focal epileptic seizures remain unknown.

MRSI Limitutians There are several limitations and problems with the present study.

(1) WATEK~LIPID CONTAMINATION. A major difficulty with in vivo 'H-MRS studies is the presence of intense resonances of water and lipid, which interfere with the proton metabolite signals of interest. Although our acquisition method suppressed most of these unwanted signals, often some water and lipid signal conramination remained, especially in spectra from regions close to the sphenoid sinus or skull. Because there is a large susceptibility contrast between brain tissue and the sinus air spaces, the hippocampal formation is a technically challenging region to study by 'H-MRSI. Susceptibility gradients cause shimming difficulties and, consequently, reduced water and lipid suppression. Large hippocampal susceptibility gradients were reflected in the observation that NAA peaks were 40 to 70% broader in hippocampi than in centrum semiovale with four to eight times the variability in line width. Residual water signal can interfere with the quantitation of choline and creatine because the spectral base-

(3)

EDDY CURRENTS. During switching of magnetic field gradients, eddy currents are induced in the conductive parts of the magnet, creating serious effects for 'H-MRSI. Because eddy currents produce spatially variable phase distortion, 'H magnitude spectra were used. Gradient tuning (see Methods) was used to minimize dynamic interference of eddy currents with shimming.

( 4 ) B~

INHOMOGENEITY. The 'H saddle-type imaging coil had some B, field inhomogeneities that degraded the water suppression, V@I slice profiles, and the sensitivity of signal reception. Improvements have been made by constructing a circularly polarized 'H birdcage resonacor, now in use for ongoing studies.

( 5 ) CHEMICAL SHIFT ARTIFACT. The VOI positions (but not the MRSI voxel positions) were offset spatially for the various 'H metabolites due to the well-known chemical shift artifact. Therefore, we selected no spectra near the VOI edges. Conclusion Despite current limitations in this new and rapidly developing method, 'H-MRSI is a useful tool for the noninvasive clinical assessment of focal epilepsy. These pilot results suggest that focal NAA reductions may be used to localize the epileptogenic foci.

This work was supported by NIH grants HL07102 (J.W.H. through the UCSF Cardiovascular Research Institute), R01-DK33293 (M.W.W.), and R01-CA48815 (A.A.M.), Philips Medical Systems, and the Department of Veterans Affairs Medical Research Service (M.W.W.). Valuable assistance was rendered by R. S. h a , E. Lin, and S. K. Steinman.

References 1. Scheucr ML, Pedley TA. The evaluation and treatment of seizures. N En& J Med 1990;323:1468-1471

Hugg et al: Neuron Loss in Epilepsy by MRSI

793

2. Dowd CF, Dillon WP, Barbaro NM, Laxer KD. Magnetic resonance imaging of intractable complex partial seizures: pathologic and electroencephalographic correlation. Epilepsia 1991;32: 454-459 3. Swartz BE, Tomiyasu U, Delgado-Escuetta AV, et al. Neuroimaging in temporal lobe epilepsy: test sensitivity and relationships to pathology and postoperative outcome. Epilepsia 1992; 33:624-634 4. Laxer KD, Hubesch B, Sappey-Marinier D, Weiner MW. Increased p H and inorganic phosphate in temporal seizure foci, demonstrated by 31P]MRS.Epilepsia 1992;33:618-623 5. Hugg JW, Laxer KD, Matson GB, et al. Lateralization of human focal epilepsy by 3'P MR spectroscopic imaging. Neurology 1992;42:2011-2018 6. Hugg JW, Matson GB, Twieg DB, et al. "Phosphorus MR spectroscopic imaging (MRSI) of normal and pathological human brains. Magn Reson Imaging 1992;10:227-243 7. Kuzniecky R, Elgavish GA, Hetherington HP, et al. In vivo j'P nuclear magnetic resonance spectroscopy of human temporal lobe epilepsy. Neurology 1992;42:1586-1590 8. Siesjo BK, Abdul-Rahman A. A metabolic basis for the selective vulnerability of neurons in status epilepticus. Acta Physiol Scand 1979;106:377-378 9. Sagar HJ, Oxbury JM. Hippocampal neuron loss in temporal lobe epilepsy: correlation with early childhood convulsions. Ann Neurol 1987;22:334-340 10. Jackson GD, Berkovic SF, Tress BM, et al. Hippocampal sclerosis can be reliably detected by magnetic resonance imaging. Neurology 1990;40:1869- 1875 11. Nadler JV, Cooper JR. N-Acetyl-L-aspartic acid content of human neural m o u r s and bovine peripheral nervous tissues. J Neurochem 1972;19:313-3 19 12. Koller KJ, Zaczek R, Coyle JT. N-Acetyl-aspartyl-glut~dte: regional levels in rat brain and the effects of brain lesions as determined by a new HPLC method. J Neurochem 1984; 43:1136-1142 11. Birken DL, Oldendorf WH. N-Acetyl-L-aspartic acid: a literature review of a compound prominent in 'H-NMR specrroscopic studies of brain. Neurosci Biobehav Rev 1989;13:23-3 1 14. Perroff OAC, Spencer DD, Alger JR, Prichard JW. High-field proton magnetic resonance spectroscopy of human cerebrum obtained during surgery for epilepsy. Neurology 1989;39: 1197-1202 15. Gill SS, Small RK, Thomas DGT, et al. Brain metabolites as 'H NMR markers of neuronal and glial disorders. NMR Biomed 1989;2:196-200 16. Matthews PM, Andermann F, Arnold DL. A proton magnetic resonance spectroscopy study of focal epilepsy in humans. Neurology 1990;40:985-989 17. Hugg JW, Matson GB, Husted CH, et al. Chronic focal encephalitis studied by 'H and j'P MR spectroscopic imaging. Soc Magn Reson Med 1991;10:779 (Abstract) 18. Hugg JW, Laxer KD, Matson GB, Weiner MW. Lateralization of human focal epilepsy by P-31 and H-1 MR spectroscopic imaging: Rabi young investigator award paper. Magn Reson Imaging 1992;2(P):33 (Abstract) 19. Brown TR, l n c a i d BM, Ugurbil K. NMR chemical shifr im-

aging in three dimensions. Proc Natl Acad Sci USA 1982; 79:3523-3526 20. Maudsley AA, Hilal SK, Perman WH, Simon HE. Spatially resolved high resolution spectroscopy by "four-dimensional" NMR. J Magn Reson 1983;51:147-152 2 1. Luyten PR, Marien AJH, Heindel W, et al. Metabolic imaging of patients with intracranial tumors: H-1 MR spectroscopic imaging and PET. Radiology 1990;176:791-799 22. Maudsley AA, Lin E, Weiner MW. Spectroscopic imaging display and analysis. Magn Reson Imaging 1992;10:471-485 23. H u g JW, Duijn JH, Macson GB, et al. Elevated lactate and alkalosis in chronic human brain infarction observed by 'H and 3LPMR spectroscopic imaging (MRSIj. J Cereb Blood Flow Metab 1992;12:7 34-744 24. Duijn JH, Matson GB, Maudsley AA, et al. Human brain infarction: proton MR spectroscopy. Radiology 1992;181:7117 18 25. Petroff OAC, Graham GD, Blamire AM, et al. Spectroscopic imaging of stroke in humans. Neurology 1992;42:1349-1354 26. Ford CC, Griffey RH, Marwiyoff NA, Rosenberg GA. Multivoxel 'H-MRS of stroke. Neurology 1992;42:1408-1412 27. Segebarrh CM, Baleriaux DF, Luyten PR, denHollander JA Detection of metabolic heterogeneity o f human intracranial tumors in vivo by 'H NMR spectroscopic imaging. Magn Reson Med 1990;13:62-76 28. Arnold DL, Matthews PM, Francis GS, et al. Proton magnetic resonance spectroscopic imaging for metabolic characterization of demyelinating plaques. Ann Neurol 1992;31:235-241 29. Duijn JH, Matson GB, Maudsley AA, Weiner MW. 3D phase encoding 'H spectroscopic imaging of human brain. Magn Reson Imaging 1992;10:315-319 30. Ordidge RJ, Bendall MR, Gordon RE, Connelly A. Volume selection for in vivo spectroscopy. In: Govil G, Khetrapal CL, Sarans A, eds. Magnetic resonance in biology and medicine. New Delhi: Tata-McGraw-Hill, 1985:387-397 31. Bottomley PA. Spatial localization in NLMR spectroscopy in vivo. Ann NY Acad Sci 1987;508:333-348 32. Part SL, Sykes BD. T1 water eliminated Fourier transform NMR spectroscopy. J Chem Phys 1972;56:3182-3184 33. Hugg JW, Maudsley AA, Weiner MW, Matson GB. Reduced voxel size and contamination in MR spectroscopic imaging. SOC Magn Reson Med 1992;11:3813 (Abstract) 34. Duijn JH, Husted CA, Matson GB, et al. Molar quantitation of in vivo proton merabolites in human brain with l D MR spectroscopic imaging. Soc M a p Reson Med 199211:38O7 (Abstract) 35. Siesjo BK, von Hanwehr R, Nergelius G, et al. Extra- and intracellular p H in the brain during seizures and in the recovery period following arrest of seizure activity. J Cereb Blood Flow Metab 1985;5:47-57 36. Ui M. A role of phosphofructokinase in pH-dependent regulation of glycolysis. Biochim Biophys Acta 1966;124:310-322 37. Giffard RG, Weiss JH, Choi DW. Extracellular alkalinity exacerbates injury of cultured cortical neurons. Stroke 1992;23:1817182 1 38. Perry TL, Hansen S. Amino acid abnormalities in epileptogenic foci. Neurology 1981;31:872-876

794 Annals of Neurology Vol 34 No 6 December 1993