Introduction

3D TSE and GRASE acquisitions are recommended for brain ASL because of their high SNR and compatibility with optimal background suppression ^[1]. While 3D TSE acquisitions are more robust to B₀ inhomogeneities, they suffer from image blurring due to T₂ decay along the long echo trains. Several methods have been proposed to reduce or compensate T₂ blurring, including k-space filtering ^{[2, 3]} and variable flip angle schemes ^[4]. While variable flip angle methods can reduce T₂ blurring, it needs optimization and may be more sensitive to B₁ inhomogeneities compared to constant refocusing flip angle method. In this study, we compared the variable flip angle schemes ^[4] implemented in 3D Cartesian TSE acquisition, with k-space filtering method ^[2] using a constant refocusing flip angle scheme for robust 3D ASL images of brain perfusion.

Methods

Variable flip angle schemes: Variable flip angle schemes were designed using the extended phase graph (EPG) algorithm ^[5] using gray matter T₁ and T₂ ^[6], inversely calculated from the predefined signal shapes along the echo train. The signal shapes were defined either as a Fermi or Hann function (‘direct Fermi’ or ‘direct Hann’ ^[4]), or the adapted forms (‘optimal Fermi’ or ‘optimal Hann’ ^[4]) which used increased amplitudes in the first few echoes to improve SNR. Maximum amplitudes and linewidths allowed by the EPG algorithm were used for these signal shapes for the benefit of SNR, by means of iterating among a range of different values. A Fermi filter was then used with optimized radius (r = 56) and width (w = 8) to increase the overall SNR of the images. The resultant signals of different variable flip angle schemes were simulated as well using the EPG algorithm to validate the calculated flip angles.
k-space filtering method: A k-space filtering method was designed as the inverse of the EPG simulated signal along the echo train to compensate for the T₂ decay modulation, combined with a Fermi function to minimize the noise contribution in the later echoes. The k-space filter intensity increases before reaching the 50% of the ETL and progressively decays to 0 at the end of the echo train. This k-space filtering method was applied to ASL images acquired with a constant refocusing flip angles of 120°.
MR Acquisition: This study used pseudo-continuous ASL (pCASL) with optimized background suppression and a 3D TSE Cartesian acquisition with Spiral Profile Reordering (CASPR) on a 3T scanner (Ingenia, Philips Healthcare) ^[7]. 3D ASL was performed in 2 healthy volunteers with IRB approval. The imaging parameters were: TR/TE = 6000/14 ms, FOV = 220x220x110 mm³, matrix = 64x64 with 36 slices, acquired resolution = 3.5x3.5x6 mm³, reconstructed resolution = 3x3x3 mm³, label duration = 1.8 s, post-label delay = 1.8 s, 1 repetition, 4 background suppression pulses and acquisition time = 3:00 minutes. A M₀ image was acquired using same acquisition parameters in 1:30 minutes. Other parameters were echo spacing = 2.8 ms and ETL = 80. Each volunteer was scanned with 5 ASL acquisitions with different refocusing flip angle schemes, including constant, direct Fermi, direct Hann, optimal Fermi, and optimal Hann.
Image analysis: The k-space filtering combined with the Fermi windowing was implemented on the scanner reconstruction platform (Philips Recon 2.0). A blur metric was used to quantitatively determine the blurring of the images ^[8].

Results

Fig. 1 shows the flip angle schemes of different methods along the echo train (ETL = 80), where the ‘optimal’ methods have larger values than ‘direct’ methods, in particular for the first few refocusing flip angles. Fig. 2c-f show the EPG simulated signals of different flip angle designs, where the signals at the peripheral k-space are better maintained than the constant flip angle design (Fig. 2a), to which a truncated k-space filter (Fig. 2b) was applied to demodulate the signal decay. Fig. 3c-f show the 3D brain ASL images in a volunteer using different variable flip angle schemes, compared against the constant flip angle image without (Fig. 3a) and with (Fig. 3b) k-space filtering, showing reduced blurring with both variable flip angle and k-space filtering methods. Fig. 4 shows the blur metric values calculated at the central slices in the 3D ASL images acquired in 2 healthy volunteers with variable flip angle schemes, compared against the values of constant flip angle images without and with k-space filtering. ASL images acquired with constant refocusing flip angles combined with k-space filtering showed similar blur metric values compared with images acquired using variable flip angle schemes (Fig. 4).

Discussion and Conclusion

Variable flip angle schemes were designed and applied to 3D brain ASL images acquired with 3D Cartesian TSE, and its performance in reducing T₂ blurring was compared against k-space filtering with constant flip angle methods. Optimized k-space filtering with constant refocusing flip angle could achieve similar deblurring provided by variable flip angle schemes, and can be more robust to B₁ inhomogeneities. Future study will consider comparing these two methods systematically in more volunteers and evaluate its sensitivity to B₁ inhomogeneities.

Acknowledgements

This work was partly supported by the NIH/NCI grant U01CA207091.