Fluid-Attenuated Inversion Recovery Preparation: Not an Improvement over Conventional Diffusion-Weighted Imaging at 3T in Acute Ischemic Stroke

Methods

Image Analysis

DW images and FLAIR DW images were transferred to a commercial workstation (Advantage Windows, version 3.1; GE Medical Systems). The mean and SD of the signal intensity was measured in a region of interest (ROI). A standard circular ROI was used (approximate area of 140 mm²) and reduced to a smaller ellipse when necessary to fit entirely within a given tissue type. These ROIs were placed in eight readily identifiable areas: 1) ischemic tissue (on the section with the largest hyperintense ischemic region visible on the DW image with a corresponding hypointense ADC region), 2) CSF (in the body of the lateral ventricle), 3) gray matter (occipital interhemispheric fissure at the same section level as for CSF), 4) white matter (periventricular white matter at the same level), 5) putamen (on the section where the putamen is widest), 6) globus pallidus (on the same section as for putaminal measurements), 7) insular cortex (on the same section as for putaminal measurements), and 8) air (ROI placed on the top right corner of the section with putaminal measures).

ROIs for nonischemic tissues were all placed contralateral to the ischemic tissue. Simultaneous measures from each ROI were made from the ADC maps (Functool; GE Medical Systems) derived as follows: ADC = −ln(S_b/S_b=0)/b, where S is the signal intensity, b is the diffusion-weighting b value (in this study, b = 1500 s/mm²). The user-defined image-intensity threshold for this calculation was set to 0 to allow an accurate assessment of the signal intensity in air that was used to define background noise. Typically, a threshold of 20% is used in clinical practice to remove this noise (Fig).

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

DW image, FLAIR DW image and corresponding ADC maps from one patient illustrate differences in ischemic tissue conspicuity (solid arrow) and in insular cortex appearance (dotted arrow). On the ADC maps (middle row), elevated signal intensity in air (noise) is observed with the use of a magnitude threshold of 0 for calculating the ADC. Typically, a magnitude threshold of 20% is used in clinical practice to remove such noise. Third row shows such thresholded ADC maps; greater noise is evident on ADC_FLAIRDWI than on ADC_DWI.

DW images and FLAIR DW images from a given patient, paired with their respective ADC maps (ADC_{DW IMAGE} and ADC_{FLAIR DW IMAGE}), were consecutively analyzed to ensure similar ROI placement and ROI shape and size with each sequence. To avoid observer bias, we randomized the order in which the images were presented. Interrater reliability was assessed in an independent analysis of the images by two readers (J.E.S., D.K.C.) blinded to clinical and other imaging information. One reader (J.E.S.) made repeat measurements, obtained not less than 24 hours apart, for the evaluation of intrarater reliability.

Results

The median interval from symptom onset to MR imaging was 2.6 hours (range, 1.7–4.6 hours). Fourteen patients had anterior-circulation strokes (nine left and five right, middle cerebral artery [MCA] territory), and six patients had posterior-circulation infarcts. The patients’ median age was 67.5 years (range, 20–83 years). Thirteen were men, and seven were women, and 50% of the patients had National Institutes of Health scores greater than five (range, 1–24). In seven, thrombolytic infusion (intravenous tissue plasminogen activator) commenced before imaging.

Figure 1 illustrates the differences in mean signal intensity, SNR, and CNR on images obtained with DW imaging and FLAIR DW imaging and their respective ADC maps. Tables 1 and 2 summarizes these differences. The mean signal intensity of brain tissues on FLAIR DW imaging was significantly lower than that of DW imaging (P < .05). The SNR was significantly higher for DW imaging than for FLAIR DW imaging in all ROIs (P < .05). Similarly, CNR was larger for DW imaging than for FLAIR DW imaging in all ROIs (P < .05).

Similar results were observed when we reviewed the ADC maps (Table 2). As expected, the SNR_ADC and the CNR_ADC were also greater for ADC_{DW IMAGING} than for ADC_{FLAIR DW IMAGING} because of fluid nulling, although this change did not reach statistical significance for all ROIs.

Interrater and intrarater reliabilities for each tissue and technique were good to excellent. Interrater reliability ranged from 0.73 to 0.96 for DW imaging and 0.89 to 0.97 for FLAIR DW imaging. Intrarater reliability ranged from 0.81 to 0.99 for DW imaging and 0.78 to 0.99 for FLAIR DW imaging.

Discussion

FLAIR DW imaging had an SNR significantly lower than that of DW imaging (18–40% lower), and the same trend was observed for ADC maps. A reduction in the signal intensity from tissues on FLAIR DW imaging is expected and agreed with the findings of others (5–7). This occurred because the fraction of the signal intensity due to a partial volume of CSF or water in each ROI is reduced by FLAIR preparation. Differing fractions of fluid in each tissue explains the variability in this reduction in signal intensity in the regions. For example, the gray matter ROI likely contained more CSF contamination than did the putaminal ROI.

Unexpectedly, the signal intensity in air (background noise) was higher on FLAIR DW imaging than on DW imaging. Although statistically significant, the difference was small and probably due to the FLAIR DW imaging having a longer TR. This may cause a change in the degree of artifact on FLAIR DW imaging, as a mild increase in ghosting could bias the noise estimates. This small increase in noise, coupled with the more profound decrease in tissue signal intensity, resulted in lower SNR measures for FLAIR DW imaging compared with DW imaging.

FLAIR DW images also showed lower conspicuity between normal tissues and ischemic tissue, as measured with the CNR (26–60% lower). SNR, and particularly CNR, can be considered an objective measure of clinical lesion conspicuity. These results suggest that the sensitivity of FLAIR DW imaging for detecting acute ischemic stroke detection may be lower than that of DW imaging.

The main advantage proposed for the FLAIR DW imaging sequence is related to the estimation of ADC. Some have suggested that lower ADC_{FLAIR DW IMAGING} values might identify damaged tissue more accurately than the larger values obtained on DW imaging (5, 7). This is important, as ADC values are increasingly used to assess the severity and duration of ischemia. In addition, in diffusion tensor imaging, the use of a FLAIR-preparatory pulse may improve the delineation of white matter tracts close to the ventricles (11). These potential advantages might be negated if the increased noise associated with FLAIR DW imaging obscures the presence of ischemic tissue. Figure 1 illustrates this hypothesis. In this patient, the ischemic lesion in the temporal cortex was less conspicuous on FLAIR DW imaging than on DW imaging, although with both sequences ischemic tissue was clearly visible more superiorly. However, no follow-up images were obtained for this patient; therefore, the tissue outcome of this differential appearance of the lesion with the two sequences remains uncertain.

Figure 1 illustrates another potential advantage of the FLAIR DW imaging sequence, which is an absence of the insular cortex hyperintensity that is often seen on conventional DW imaging. This insular hyperintensity can sometimes be misinterpreted as being due to ischemia. The use of FLAIR DW imaging might increase the specificity in detecting ischemic tissue, but at the cost of reduced sensitivity. Whether there are differences in the clinical detection of ischemic tissue by using the two image acquisition methods is the subject of an ongoing study (12).

We used a 3T MR machine in this study. The current standard field strength for clinical MR imaging units is 1.5T. The primary advantage of imaging at a higher field strength is an increase in SNR, which we have used to improve spatial resolution. However, a disadvantage is that susceptibility-induced image distortions may be greater due to faster T2* relaxation when single-shot echo-planar imaging is used (as in this study). Both DW imaging and FLAIR DW imaging have little T1 weighting and thus are not affected by increased T1 relaxation in parenchyma at 3T. In addition, the FLAIR TI does not change with field strength, so that CSF suppression with FLAIR DW imaging would be the same at 1.5T (13). Theoretically, field strength does not affect diffusion-imaging properties, and the choice of b value does not affect the degree of fluid suppression. Therefore, these results should be reproducible at other b values, including b = 1000 s/mm² at 1.5T (the current clinical standard).

Conclusion

Our SNR and CNR results suggest that FLAIR DW imaging provides ischemic tissue conspicuity lower than that of DW imaging. Deriving ADC values from FLAIR DW imaging may have advantages, but FLAIR DW images may be inferior to DW images for the visual detection of hyperacute strokes.