Introduction

Transient global amnesia (TGA) is still an enigmatic disease and patients present with a remarkable and characteristic clinical syndrome. A TGA episode is often described as a threatening experience by patients and their relatives for the profound and disturbing clinical symptoms. The disease TGA was first defined by Fisher and Adams as the temporary loss of anterograde and recent retrograde memory lasting less than 24 h with preservation of alertness, attention and self-identity and without other neurological deficits [1]. Various etiologies for TGA episodes have been suggested including ischemic insult [2], transient mesiotemporal ischemia induced by venous congestion [3–5], migraine [6], spreading depression [7] and epileptic seizures [8]. An involvement of temporomesial structures, such as the hippocampus has been suggested by several studies [9,10]. Episodes of TGA have a relatively unique clinical presentation and diagnosis can often be made based solely on the clinical symptoms. In cases of equivocal clinical findings diffusion-weighted imaging (DWI) represents a valuable diagnostic adjunct in establishing the diagnosis of TGA. Initial reports on magnetic resonance imaging (MRI) changes in TGA were contradictory in some studies [11–25] demonstrating diffusion restriction (hyperintense) in lesions of temporomesial structures by DWI, while other studies did not confirm these findings [26–29]. It is now generally accepted that a substantial proportion of TGA patients have characteristic small punctuate hyperintense lesions in DWI studies (Fig. 1) although the published detection rates vary broadly. A recent study showed that these lesions are located exclusively in the lateral portion of the hippocampus, i.e. the CA1 region [11]. Comparable punctuate hippocampal DWI hyperintensities may be very rarely seen in posterior cerebral artery (PCA) stroke (see pattern 4 in Szabo et al. [30]) but these patients do not present with TGA symptoms. Hence in the context of TGA DWI can be regarded as a diagnostic tool with a very high specificity. With respect to sensitivity the published reports vary widely (12–86%) regarding the proportion of patients in whom the characteristic DWI lesions were detected.

Fig. 1
figure 1

Example of a characteristic hippocampal lesion in diffusion-weighted imaging (b = 1000 s/mm2)

Full size image

To improve sensitivity and detection rates previous reports investigated the impact of slice thickness, diffusion-weighting factor and timing of the MRI examination relative to TGA symptom onset [11–14,24,25]. Almost all previously published results were acquired in relatively small patient populations. A recently published study in a large sample of 200 TGA patients who were examined with a DWI protocol (1.5 T, 5 mm slice thickness with 90% of all patients being examined in less than 24 h) showed a low diagnostic yield when performed early in the course of TGA [24].

These and previous results have now been extended by a retrospective MRI data analysis in a large population of 198 TGA patients, comparing the detection of DWI lesions for different slice thicknesses, field strengths and detection rates are related to the time interval between TGA symptom onset and MRI examination. The aim of this study was to identify and affirm factors which increase detection rates and if possible to derive imaging protocol recommendations to fully exploit the diagnostic capabilities of DWI in the context of patients with suspected TGA but equivocal clinical findings.

Methods

Following approval from the institutional review board for this retrospective study a comprehensive data base search was performed to identify all MRI studies performed in this institution (major university hospital in Berlin, Germany) throughout the last decade (January 2000 to July 2011). The aim was to retrieve all studies performed including patients with clinically suspected TGA who were referred for MRI assessment by the neurology department and 239 patients were identified. Emergency reports and discharge letters of these patients were scrutinized to corroborate the diagnosis of TGA. Patients where the discharge letter revealed that TGA was not the final diagnosis (n = 12) and for whom no DWI study was available (MRI study not retrievable n = 10, DWI not performed n = 19) were excluded. A total of 198 patients were included (123 women, 75 men, mean age ± SD = 64.7 ± 8.5 years) in the further analysis. For these TGA patients DWI studies with a slice thickness of 3 mm, 5 mm, or 6 mm in the institutes picture archiving and communication system (PACS) were retrieved. After grouping the studies according to DWI slice thickness a total of 47 DWI studies were identified with 3 mm slice thickness, 59 with 5 mm and 92 with 6 mm. The time span in days between onset of TGA symptoms and MRI examination was recorded with day 0 being the day of symptom onset.

The DWI studies were reviewed by two raters (MS and CM) independently for the characteristic small punctuate hyperintensities in the hippocampus on a standard PACS workstation (GE Centricity RA 1000, GE Healthcare Munich, Germany). Both raters noted the occurrence of these signal abnormalities as well as the lesion side (left, right or bilateral), lesion number (single, multiple) and lesion diameter. In six cases the two raters differed regarding the occurrence of a lesion. In all six cases the lesion suspected by one of the raters was small and not well defined. As all six cases were inconclusive the consensus rating was “no hyperintense lesion present”. The raters also noted the type and field strength of the MR scanner used. Most DWI studies (21 out of 27) on the 3 T scanners were performed with a slice thickness (ST) of 3 mm and most of the studies on the 1 T scanner (58 out of 60) were performed with 6 mm ST. Consequently, there was a high correlation between magnetic field strength and slice thickness. Therefore, three different subset analyses were performed. The first subset analysis included the data from 1.5 T scanners only (n = 26/54/31 with respect to 3/5/6 mm ST) and another two subset analyses compared the detection rates between 1.5 and 3 T at 3 mm ST and between 1 T and 1.5 T at 6 mm ST.

Statistics

Tests were carried out for the presence of DWI hyperintense lesions, the occurrence of single versus multiple lesions, lesion side as well as for differences in lesion diameter between the three groups. The c2-test was used for testing the rate of DWI hyperintense lesions. Due to the small number of observations (n < 5) in some cells of the contingency table the Freeman-Halton extension of the Fisher exact test was used for testing group differences with respect to single versus multiple lesions and lesion side. A one-way ANOVA was chosen to test differences in lesion diameter. A p-value < 0.05 was regarded as being significant for all tests.

Results

The group comparison showed a significant increase of detection rates (i.e. approximately 8.4% per millimeter reduction in slice thickness) when smaller slice thicknesses were used. The DWI hyperintense lesions were detected in 44.7% when using an ST of 3 mm, in 27.1% when using 5 mm and in 19.6% when using 6 mm. The detection of bilateral and multiple lesions on a descriptive level increased with smaller slice thickness but did not reach statistical significance. The comparison of lesion diameters also did not reveal a significant difference between groups (Table 1, Fig. 2).

Fig. 2
figure 2

Comparison of detection rate (a), rate of singular versus multiple lesions (b), lesion side (c) and lesion diameter with mean ± SD (d) for the different slice thicknesses

Full size image
Table 1 Occurrence rates of diffusion-weighted imaging lesions (number of patients) and lesion diameter (mean ± SD) for different slice thicknesses, p-values (uncorrected for multiple comparison)s for c2-tests detection rate), Fisher exact test (multiple lesions and lesion side) and one-way ANOVA (lesion diameter)
Full size table

Subset Analysis

A subset analysis of the data from all 1.5 T scanners was performed to rule out magnetic field strength as the decisive factor for the findings. The results of the subset analysis (Table 2) were very similar to the results for the whole data set (Table 1).

Table 2 Results for the subset of studies performed at 1.5 T. Occurrence rates of diffusion-weighted imaging lesions (number of patients) and lesion diameter (mean ± SD) for different slice thicknesses, p-values (uncorrected for multiple comparisons) for c2-test(detection rate), Fisher exact test (multiple lesions and lesion side) and one-way ANOVA (lesion diameter)
Full size table

The comparison of 3 mm ST at 1.5 T versus 3 T showed no significant difference in detection rates: 1.5 T (DWI pos/DWI neg = 13/13) and 3 T (DWI pos/DWI neg = 8/13, c2 p = 0.60). The same was true for the comparison of 1 T and 1.5 T at 6 mm ST: 1 T (DWI pos/DWI neg = 11/47) and 1.5 T (DWI pos/DWI neg = 6/25, c2 p = 0.81). The subset analysis confirmed that reduced slice thickness but not increase in magnetic field strength was responsible for the increased detection rate.

Optimal Time Interval

Regarding the time interval between symptom onset and imaging, the lesion detection rate was highest on day 2 after TGA onset in this data set (Fig. 3).

Fig. 3
figure 3

Detection rate of diffusion-weighted imaging hyperintense lesions for the different slice thicknesses with respect to time, day 0 being the day of symptom onset

Full size image

Discussion

This study demonstrated the significant impact on detection rates for TGA lesions when adapting slice thickness (approximately 8.4% increase per millimeter reduction in slice thickness) and timing of the MRI study. Transient global amnesia does not have an impact on mortality or morbidity in affected patients and does not represent a risk factor for stroke or ischemic disease [6]. However, a timely diagnosis will help in reassuring the patients and their relatives of the benign character of this subjectively disturbing disorder. The diagnosis of TGA can be established with much greater confidence if the patient shows the characteristic DWI lesions, as comparable punctuate hippocampal DWI hyperintensities may only be rarely seen in PCA stroke [30] but these patients do not present with TGA symptoms and often show further lesions outside the hippocampus.

The overall detection rate in this population with MRI studies using different slice thicknesses as well as varying time spans between symptom onset and MRI studies was only 28%. However, considering the subgroup with 3 mm slice thickness and a 2-day time interval a detection rate as high as 80% was encountered (Fig. 3) which compares well with previous studies using the 3 mm approach [11,12,25]. In contrast the use of slices >3 mm and the use of different study time points have resulted in varying detection rates ranging from 0% to 86%. It has been recently shown in a large dataset of 200 TGA patients that DWI has a low diagnostic yield when performed early in the course of TGA [24]. These varying detection rates were more questioning than confirming the evidence of a clear TGA-DWI lesion association and questioning the value of DWI as a diagnostic procedure. This study demonstrated the different influencing factors when using DWI as a diagnostic tool in TGA.

The main limitation of this study is the retrospective analysis of MRI studies obtained outside a controlled prospective study setting. By evaluating all available clinical data (emergency reports and discharge letters) the aim was to include “pure” TGA patients only. Another limitation is that the DWI studies were collected using different scanners from different manufacturers and these factors might have influenced the results. Regarding field strength in the subset analysis a significantly increased detection rate could not be found in the DWI studies performed at higher field strengths. However, an advantage of higher magnetic field strengths to detect small lesions would still be expected due to a higher signal-to-noise ratio. This has been shown in recently published study where the detection rate was significantly higher at 3.0 T compared to 1.5 T [25].

The authors are not aware of any studies investigating TGA with slice thicknesses below 3 mm; however, with an increasing number of imaging centers performing DTI allowing higher resolution studies with slice thicknesses of 2 mm are to be expected. Future work will show if detection rates and sensitivity can be further improved when using slice thicknesses below 3 mm and if hippocampal DWI lesions can be found in almost every TGA patient.

Conclusions

In the context of TGA DWI can be a diagnostic tool with high specificity and high sensitivity. This study corroborates the call for an optimized imaging protocol in order to improve the sensitivity of DWI. It was also shown that MRI should be performed on day 2 after symptom onset and the DWI sequence should have a slice thickness of no more than 3 mm.

Acknowledgment

We thank Anne Lesemann for help with data collection.

We also thank Friedrich C. Luft Clinical Scientist Pilot Program funded by Volkswagen Foundation and Charité Foundation for financial support of M. Scheel