Fluid-attenuated Inversion Recovery Now with Another Acronym: “KRISP FLAIR”

Over the years, numerous MR pulse sequences have been described and discussed in the literature, but few have had the immediate and lasting impact of fluid-attenuated inversion recovery (FLAIR). As is well known by all involved in clinical neuroradiology, FLAIR provides T2-weighted images while suppressing the signal from CSF by use of an inversion pulse. The ability to discern high-signal abnormalities, particularly those adjacent to the ventricles, is improved, and by virtue of a widened, dynamic range, FLAIR frequently allows detection of what otherwise would be subtle findings on conventional spin-echo T2-weighted images. This pioneering work was first described and published by Graeme Bydder and his associates when the technique was originally coined “PIETIR” (prolonged inversion and echo time inversion recovery).

Despite the many advantages of FLAIR, a frequently annoying finding is the presence of unsuppressed CSF, particularly in the basal cisterns. Over the years, schemes have been published to diminish or eliminate this unwanted high signal. In this issue of the AJNR, Herlihy and colleagues (page 896) describe the implementation of a technique with the acronym, KRISP (k-space reordered by inversion time at each slice position) to improve FLAIR imaging even further.

In general, using selective inversion pulses that are thicker than the slices being imaged provides adequate CSF suppression. With this method, the CSF outside the slice is also inverted; hence, if the volume of the inverted CSF stays within the slices being imaged during the inversion time period, which is typically 2000 ms, the CSF signal is nulled. One of these methods employs an inversion pulse that is twice the thickness of the slice being imaged. The disadvantage of this method is that one has to use a 100% interslice gap, because the inversion pulse is twice the slice thickness of the images. To overcome this problem, interleaved acquisition has been proposed whereby the odd-numbered slices are acquired first, followed by acquisition of the even-numbered slices. With this technique, the inversion pulse used is twice the thickness of the slices being imaged to adequately suppress the CSF, and the imaged slices are contiguous. The disadvantage with the interleave acquisition method is that the total scanning time is doubled; however, by using fast spin echo as a data acquisition method, the total scanning time can be reduced. Both of the above-mentioned methods rely on the fact that the bolus of the inverted CSF stays within the slice being imaged during the inversion time period. There may be instances when the CSF flow is so fast that uninverted CSF enters the slice during the inversion time period. This could result in spurious signal from the inflowing CSF.

Another approach is to use a non-selective inversion pulse. With this method, the entire CSF signal is inverted; thus, the problem of uninverted CSF flowing into the slice being imaged is overcome. This is followed by data acquisition of multiple selective slices. The disadvantage with this method is that each of the slices will have a different inversion time; hence, the slices that are farthest away from the inversion time will have partial signal from the CSF. Also with this method, the number of slices that can be acquired is limited because the long echo times used for acquisition of T2-weighted images would result in inversion times being different than the optimum value.

The use of a non-selective inversion pulse provides the optimum method for suppression of CSF signal and it does not suffer from the inflow effect of the unsuppressed CSF that leads to spurious signals when using selective inversion pulses. Herlihy et al use non-selective inversion pulses to suppress the signal from the CSF; however, rather than using conventional data acquisition methods, the authors have proposed a data acquisition scheme that is similar to the fast spin-echo method whereby the effective echo time is at the center of k-space data from which contrast is determined. The authors have proposed to acquire the center of k-space data when the effective inversion time is optimal to suppress the CSF signal and the outer k-space data are acquired at much shorter or longer inversion times. The resultant images do not suffer from unsuppressed CSF signals. In theory, a “krisper” image should be obtained.

All of the above methods do not take into account the RF inhomogeneity that leads to partial inversion of the CSF signal. It is very severe at the edge of the RF transmitter, where CSF may experience a tip angle that is much less than 180°, and the resulting image will have signal from partially inverted CSF. One approach to overcome the problem of RF inhomogeneity is to use adiabatic inversion pulses. The adiabatic RF pulse alters the phase of the spins being excited such that the spins at the edge of the RF transmitter, which experience a smaller excitation tip angle, will be totally inverted at the end of the pulse. The adiabatic RF pulses invert all the CSF through the transmitter regardless of the RF inhomogeneity. These RF pulses can be designed to be selective or non-selective. When using selective excitation, one still has to take into account the inflow effects of the uninverted CSF, as described above.

It is certain many new schemes will be developed to effectively suppress the CSF. Recent developments have improved the suppression of the CSF signal on FLAIR images considerably. Maybe in the near future we will see a combination of non-selective adiabatic inversion pulses combined with k-space reordered acquisition, as described by Herlihy and colleagues. With this combination, one would overcome the problem of unsuppressed CSF flowing into the slices being imaged, as well as the RF inhomogeneity problem of the transmitter that leads to spurious signal from CSF in FLAIR imaging.