Detection of delayed focal MR changes in the lateral hippocampus in transient global amnesia

Methods.

Within 27 months, 31 consecutive right-handed patients presented to our emergency facilities (15 men and 16 women; age 44 to 83 years, mean 64 ± 7 years) with clinically identified TGA.10 All underwent serial MRI examinations for 3 days, starting 2 to 8 hours after onset of TGA with two follow-up MRI studies performed after 24 and 48 hours.

The following diagnostic criteria11 for TGA had to be fulfilled: 1) The attack had to be witnessed by an observer to provide objective information; 2) anterograde amnesia was present; 3) no clouding of consciousness (no psychomotor slowing) or loss of personal identity; 4) cognitive impairment was limited to amnesia (no apraxia or aphasia); 5) no recent history of head trauma or seizures in the preceding 10 years. Only one of the patients reported previous episodes of TGA. There were no focal neurologic signs, symptoms, or features suggesting epilepsy in all patients studied (table). In 12 of 31 patients, various potentially stressful emotional events (i.e., funerals, etc.) preceded symptom onset. None of the patients had consumed alcohol shortly before symptom onset; one patient regularly used benzodiazepines at bedtime. EEG was performed in all patients but did not reveal focal or generalized abnormalities indicating epileptic activity.

In addition, all patients underwent standard procedures of stroke work-up: EKG, echocardiography, and extra- and intracranial Doppler sonography (see the table).

MRI was performed with a 1.5 T scanner (either a Magnetom Vision or a Magnetom Sonata; Siemens Medical Systems, Erlangen, Germany) with echo planar hardware (gradient power 25 mT/m and rise time 83 mT/m/ms; and gradient power 40 mT/m and rise time 200 mT/m/ms). After using orthogonal localizers, standard transverse continuous 5-mm images, and field of view (FOV) of 240 mm, proton density, T2-weighted (turbo spin echo [TSE] 2,620 milliseconds/14 milliseconds/85 milliseconds/5 mm/FOV 240 mm), T1-weighted (SE 530 milliseconds/12 milliseconds/5 mm), DW (echo planar, SE, repetition time 4,000 milliseconds/echo time [TE] 144 milliseconds, 240-mm FOV, 5-mm slice thickness, 128 × 128 matrix, three b values = 0 to 1,000, some up to 2,000 s/mm², TE = 100 milliseconds, matrix 96 × 128, diffusion gradients in three orthogonal planes) images in the transverse oblique plane were acquired. Further DW sequences aligned with the hippocampus and coronal sequences positioned perpendicular to the hippocampus with differing phase-encoding directions to evaluate the medial temporal lobes with reduced artifacts were performed. Maps of the apparent diffusion coefficient (ADC) were obtained by a linear least-squares fit on a pixel-by-pixel basis after averaging of the direction-dependent DW images. ADC maps and DW (isotropic, b = 1,000 s/mm² and b = 2,000 s/mm²), T2-weighted, and T1-weighted images were analyzed for acute and chronic abnormalities.

We currently follow up all patients for persisting memory deficits; at present, these long-term outcome data are not available.

The study was approved by the local ethics committee. For the initial scan, at least the accompanying relatives gave consent if the patient was not able to do so. All patients were studied after the cessation of symptoms, and all gave written informed consent before proceeding with MRI. In two patients, follow-up MRI could be obtained only after day 2: one on day 4 and one on day 5 (see the table: day 4, Patient 30; day 5, Patient 31).

Results.

In 26 of 31 patients, small, focal parenchymal hyperintense lesions were noted on MR DWI after 48 hours (figures 1 to 3⇓⇓). In 23 cases, these lesions corresponded with a small area of reduction in the ADC maps. All lesions were of a very small size of approximately 1 to 2 mm in diameter. Fifteen patients exhibited a left hippocampal lesion, five patients bilateral, and six a right hippocampal lesion. Lesions were usually localized in the upper part of the hippocampus next to the lateral ventricle. All 26 lesions were detectable on day 2 (48 hours), 23 after 24 hours, but only 2 lesions in the hyperacute phase (Patients 1 and 13). In 11 patients, retrospective analysis suggested that a minor hyperintensity was already present on the preceding images that was not noted as abnormal for the subtle hyperintensity. DW measurements with higher b values (b = 2,000 s/mm²) and in isotropic acquisition tended to be more specific with regard to lesion identification while reducing artifacts (see figure 3). All lesions identified on DWI were confirmed in the coronal plane or on conventional T2-weighted MRI or fluid-attenuated inversion recovery images. Hippocampal lesions (pre-existing or new) other than the ones described above could not be detected in any patient.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. (A) Schematic drawing of focal diffusion-weighted imaging lesions located in 26 patients with transient global amnesia and in a segmented hippocampus (projection in neurologic convention: left side = left hemisphere). (B) T2 lesion and the related neurovascular anatomy of the hippocampal structures. Schematic drawing of the borderzone between the upper and lower hippocampal artery: the hypoxia-susceptible sector of Sommer.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Original MRI with selected slices of a left hippocampal lesion of Patient 14. Lesion development: no significant diffusion-weighted imaging lesion (DWI and ADC-Maps) hyperacutely (left slide), but detection of a significant DWI and ADC lesion in the rostral part of the left hippocampus after 48 hours.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Classic transient global amnesia, 24 hours after onset (Patient 21): contrast of the techniques employed. (A) Three directions, isotropic acquisition, b value = 1,000; (B) single direction (phase), b value = 2,000; (C) apparent diffusion coefficient map, three directions, isotropic acquisition, b values = 0, 1,000, 2,000.

In five patients, no lesion was detectable on follow-up scans; in two of the negative scans, patient imaging could not be performed after 24 and 48 hours but only after 96 (Patient 31) and 120 hours, respectively. No predictors of negative follow-up imaging (age, sex, duration of amnesia or risk factors) could be found (Wilcoxon matched pairs test).

Discussion.

In contrast to earlier studies, including a previous study from our group,6,9⇓ 26 of 31 patients with TGA developed punctate small lesions on MR DW images when investigated 24 to 48 hours after the onset of symptoms. The high number of patients included reflects a regional stroke awareness program12 with very early admittance of large numbers of patients with suspected stroke (>800 patients/year within 6 hours of symptom onset). While applying identical clinical criteria in patients with the typical age range,13 only imaging time points were changed. All lesions were centered in a particular anatomic part of either or both hippocampi: pes and fimbria hippocampi. Lateralization of the lesion locations showed a preference for left hippocampal structures (i.e., 15 left sided + 5 bilateral), similar to findings in unilateral posterior cerebral artery stroke causing amnesia, where about 85% of the lesions were located in the left hemisphere.14 The consistent locations of lesions in anatomic regions likely to be involved in the development of memory disturbances suggest that DWI detected biologic abnormalities. By using the patients as their own controls in a follow-up design and by using transverse and coronal planes, results were ascertained.

The results of this study, although positive, may well be in line with earlier negative studies. Previous studies using DWI were rather designed to detect early stroke lesions that may be detected before T2 signal change becomes apparent after focal ischemia. Indeed, ischemic lesions being seen without delay have been reported in patients with TGA as an embolic complication of vertebral artery angiography.15–17⇓⇓ DWI lesion was reported in the left parahippocampal gyrus2 and in the right hippocampus15 by different authors already 8 hours after symptom onset. The hippocampal DWI lesions associated with TGA in these studies15–17⇓⇓ are larger than the ones demonstrated in the current study and may therefore have been identified earlier. Lesions in our patients were clearly identified with a delay, in the second or even third examination at 24 and 48 hours, respectively, but in only 2 of 31 cases in the acute phase. The timing of the development of DWI-visible lesions was therefore different from what one would expect in acute stroke. However, lesions were very small, and there is a possibility that the pathology was already present at the first MRI but was only identified once signal change was prominent enough in the second scan. This is also consistent with the observation of several scarcely visible lesions, definitely identifiable after 24 or 48 hours. Furthermore, a delayed appearance of ischemic lesions has been reported in small white matter stroke lesions18 and brainstem and internal capsule infarction.19 This possibility may eventually be explored once higher-resolution DWI sequences and higher-field strength MR systems can be used in TGA patients.

Previous studies also tried to link TGA to ischemic events and identified risk factors for TGA that resembled those for TIA20 (arterial hypertension and preceding lacunar infarctions), but a recent study could not identify any association with vascular risk factors in 108 patients.13 In our patients, an embolic origin could be assumed only in less than one-fourth of the patients, whereas almost two-thirds of the patients had vascular risk factors; 50% of those had more than one vascular risk factor.

Hyperintensity on DWI and a reduced ADC have been observed in a number of nonischemic pathologies including prolonged ictal activity, venous thrombosis, multiple sclerosis, and phenylketonuria.21,22⇓ DWI-hyperintense lesions in these pathologies tend to be bigger and may be accompanied in the vicinity by signs of vasogenic edema (venous thrombosis, multiple sclerosis) somewhat different from the punctate lesions observed in this study. The variety of potential pathophysiologic processes and the lack of unequivocal circumstantial evidence make it extremely difficult to be certain in regard to the underlying pathophysiology of hippocampal lesions in TGA. Hyperintense DWI lesions, even when accompanied by a reduced ADC, are highly suspicious for cerebral ischemia but are not definite evidence.

One of the recently discussed mechanisms of TGA is hippocampal oligemia induced by venous congestion due to a Valsalva maneuver.23 The results of our study would not suggest or favor venous congestion as a mechanism of TGA, particularly as there were no signs of vascular or tissue change such as may be observed in cerebral venous stasis or thrombosis.23,24⇓

Spreading depression (SD), a favored mechanism of aura and repeatedly claimed to be associated with TGA, has not been directly observed with MRI in humans, nor can our findings be taken as evidence of SD. In a recent study, cortical recordings in head/brain trauma patients identified clearly the electrophysiologic pattern of SD as it has been observed in numerous experimental studies.25 However, there is no information on DWI or the perfusion status at the time of SD in these patients. If EEG alterations were seen in patients with TGA26 (none in our patients), they do not reflect the findings in SD, to our understanding. Another recent study pointed out that SD is not accompanied by hypoperfusion in primates, indicating that the situation is different from what one would expect looking at studies in mice and rats.27

Local anatomic and functional characteristics of the hippocampus could play a role in the unusual timing of the lesion detection. The hippocampus as part of the limbic system has been noted to show particular anatomic and functional characteristics. TGA often occurs after strong emotional and stressful situations. Emotional arousal has been suggested to lead to metabolic disturbances, namely, enhanced glutamate release,28,29⇓ which would in turn account for increases in energy requirements.30,31⇓ In this context, age-related small-vessel changes might explain an age-dependent susceptibility for TGA in the seventh decade.13 The neurovascular anatomy of the hippocampus has been a focus of previous research: In 1937, when the origin of epileptic seizures was largely unclear, anatomic studies32 identified a “resistant sector of Spielmeyer” (CA 2/3), which was called resistant because of the tolerance to hypoxia, and the hypoxia-susceptible sector of Sommer (CA1) affected by conditions such as CO intoxication or seizure disorders33 (see figure 1B). Hippocampal arterial supply shows an internal anastomosis, providing arcades with an upper and a lower artery.32 The watershed area between the two is the susceptible sector of Sommer, the lateral aspect of the hippocampus, which was also involved by the lesions observed in this study.

Although TGA remains mysterious pathophysiologically, the concept of delayed ischemic mechanisms being involved is appealing. High metabolic rates leading to relative hypoperfusion in the subcortical vascular borderzone in patients with mild vascular changes are known to be associated with delayed ischemic mechanisms.18 Similar processes occurring at a hippocampal level could be the natural history of sporadic TGA.