Magnetic resonance spectroscopic imaging (MRSI) provides unique characterization of metabolites within cancerous or benign tissue. Unfortunately, the technique is often plagued by signal contamination (e.g. from fat) from outside the tissue of interest, particularly in voxels near organ boundaries. Post-processing filters are typically applied, which have the undesirable side effect of further lowering spatial resolution. [1] The purpose of this study is to demonstrate and validate in phantoms and healthy volunteers a spatial localization technique that features spatially localized signal excitation and therefore eliminates the need for post-processing filtering.

Pulse Sequence: The sequence was customized based on 3D-PROSE; the excitation pulse was replaced by a series of four hard pulses weighted with a Point Spread Function (PSF)–Choice [2] encoding scheme along two spatial directions. A 512-pulse echo planar train read out with alternating polarity was added along the third direction during an FID acquisition. Each blip collects 32 points during the ramp and plateau.

Phantom Validation: The point-spread-function obtained with this technique was verified at 3T, on a GE prostate phantom (metabolic solution enclosed in a 10-cm diameter sphere). A PRESS-localized spectroscopy box with the volume set to one eighth of an imaging voxel was prescribed. Sixteen acquisitions were collected with 0, ¼, ½ and ¾-voxel shifts in each of the two PSF-encoding directions. The localization quality of the technique was also evaluated using an in-vitro phantom consisting of a lime immersed in extra-virgin olive oil. The absence of contamination from neighboring tissue was demonstrated under the condition of poor spatial resolution, which is often the case in spectroscopic imaging to balance SNR. The imaging volume was 8×8×8 cm3, and acquisition matrix was 8×8×8 along EPI/PSF/PSF2 encoding directions, resulting in a voxel size of 1 ml.

In vivo Study: Healthy volunteer scans were conducted in this IRB-approved study. Imaging was performed at 3T using an 8-element cardiac phase array coil with the following parameters: TE/TR = 85/1200 ms, EPSI readout along the R/L direction, imaging volume = 8×8×8 cm3, acquisition matrix = 8×8×8 along EPI/PSF/PSF2 encoding directions, voxel size = 1ml. The customized PROSE sequence was prescribed on T2 weighted oblique axial slices, and a spectroscopy press box was selected tightly within the prostate, and no saturation bands were prescribed.

PSF Mapping: As shown in Fig. 1, the mapped PSF from the customized PSF-choice encoding scheme is relatively free of ripples surrounding the main peak, compared to the standard Fourier-based encoding scheme. The PSF is demonstrated for choline (Cho), creatine (Cr) and the citrate (Cit) metabolic peaks.

Lipid Contamination: By selecting the spectroscopy press box across compartment of both lime and oil, the spectra in Fig 2a demonstrates predominantly citrate signal within the lime voxels, and predominantly lipid signal within the oil voxels. Note the lipid signal was attenuated due to spectral filtering inherent to the PROSE sequence. The signal is constrained within the press box in the sagittal plane where PSF-Choice encodings were applied in both directions (Fig. 2b left), while there is evident ringing artifact along the EPI read out direction due to the Fourier-based encoding (Fig. 2b right).

Volunteer Study: Major metabolic peaks within prostate have been identified, including Cho/Cr/Cit located at around 3.2/3/2.6 ppm respectively, and a representative spectrum within a single voxel is displayed in Fig. 3b.

While the spatial localization was achieved along both PSF-encoding directions, further improvement is needed in constraining the localization along EPI read out direction to avoid fat signal contamination into selected press box.We have demonstrated the features of PSF maps at metabolic peaks in a 2D Gaussian-based PSF encoding scheme and also the comparisons of signal contamination from surrounding tissue in in-vitro phantom. We further investigated its feasibility with in-vivo studies for prostate MRSI