MR spectroscopic imaging (MRSI) is an important tool in clinical applications. As the spectrum is measured in either the frequency or the time domain, MRS is divided into the frequency-resolved and the time-resolved techniques (1, 2, 3). The early frequency-resolved techniques have been supplanted by the time-resolved technique with faster acquisition and higher sensitivity. We introduce a fast frequency-resolved MRSI technique, termed phase-cycled spectroscopic imaging (PCSI). PCSI uses an ultra-low flip-angle steady state to achieve high acquisition efficiency and faster frequency sweeping by changing cycled RF phase and using flexible non-uniform sampling, and greatly reduces RF energy deposition. With its intrinsic water and fat suppression, performing PCSI more closely resembles routine clinical scans by eliminating outer volume suppression steps. We demonstrate its feasibility using simulations, phantom and human studies.

PCSI is the first pulsed frequency-resolved MRSI method. Using a series of ultra-small RF pulses to excite a target frequency, PCSI acquires an image at that frequency. By sweeping through discrete frequency points using RF phase-cycling, PCSI acquires a series of images at a range of frequencies and generates a spectrum in each voxel. PCSI makes frequency sweeping more flexible, allowing non-uniform frequency sampling of selected frequencies to speed up acquisition.

We performed simulations with multiple small flip angles to validate its feasibility and investigate the importance of the choice of flip angle. In simulations, a spectrum was created with four peaks: water, Cho, Cr, and NAA, with magnitudes of 10000, 0.6, 0.8, and 1.2, respectively. The profiles of the transverse magnetization were computed for different flip angles (0.24, 1, 3 degree) using TR=2.4 ms, T1=1300 ms, T2=250 ms.

All imaging studies were performed on a Siemens 3T scanner. PCSI acquisition uses a modified bSSFP sequence with the cycled RF phase changing for each measurement. A single axial slice was selected, and the advanced shimming mode was used. The system frequency was decreased by 200 Hz for phantom, 190 Hz for human. The protocol was as follows: ms; ms; flip angle for phantom, for human; acquisition matrix 32×32; resolution 6.25 × 6.25×15 mm³; 23 averages; 143 measurements with non-uniform sampling; total time 4:28 minutes.

For comparison, single-voxel spectra were acquired using a single voxel spectroscopy (SVS) sequence with TE =135 ms, voxel size = 20×20×20 mm³, and the total time 4:24 minutes. Chemical shift imaging (CSI) data were acquired from a patient with TE=135 ms; voxel size 10×10×15 mm³; acquisition matrix 10×10; and total time 8:07 minutes.

Figure 1a shows the original generated spectrum. Fig. 1b shows the simulated spectrum with an optimal a=0.24°. All metabolic peaks are observed (Fig.2b), with heights proportional to the original values. Fig. 1c shows metabolic peaks are barely detectable in the simulated spectrum with α=1°. Those peaks are not appreciable in Fig. 1d, with α=3°.

Figure 2a shows PCSI signals of the phantom at the center (ROI size: 18.75×18.75×15 mm³). “Zooming in” inset demonstrates three metabolite peaks on the real signal, which was converted to the spectrum (Fig. 2b). For comparison, Fig. 2c shows a single-voxel spectrum using SVS. PCSI and SVS spectra (Fig. 2b, 2c) were similar and had consistent peak positions.

Figure 3a shows a PCSI spectrum from a volunteer (ROI size: 18.75×18.75×15 mm³), in which three metabolite peaks can be identified and fitted. The corresponding SVS spectrum is shown in Fig. 3b, with inset showing ROI’s location similar to that in Fig. 3a. Both spectra show consistent peak positions and relative peak heights after alignment. Figure 4 shows metabolic parametric maps were generated for two repeated measurements to demonstrate a good robustness of PCSI. Two sets of maps were interpolated and overlaid on a T2-weighted image for comparison.

Figure 5 shows CSI parametric maps and PCSI maps from a patient with a multifocal anaplastic astrocytoma. There were four lesion foci (1-4) shown in one T2w image, in which lesion 1 corresponds to a surgical cavity and lesions 2-4 represent tumor foci. Fig. 5 shows the importance of full coverage of the whole slice due to the heterogeneity of tumors.

We demonstrated PCSI, a new flexible way, to perform MRSI in the frequency domain. PCSI uses a new frequency-sweep technique and a non-uniform sampling scheme that speed up acquisition. PCSI also offers intrinsic water and fat suppression, reducing operator dependence and greatly simplifying scanning procedures. Further, PCSI substantially reduce SAR using ultra-small RF pulses. PCSI paves the way to faster, simpler, and broader clinical applications of MRSI.