The proposed approach aims at reducing the acquisition time of Magnetic Resonance Spectroscopic Imaging (MRSI) through application of Magnetic Resonance Fingerprinting (MRF) ² based on dictionary generation of metabolites.

A retrospective analysis was performed on the Ethanol (CH₃CH₂OH) phantom data sets. These datasets were acquired on Siemens 1.5T scanner with dimensions of 16 X 16 X 1024 with 1024 being the time points. A spectra was obtained with TE/TR combination of 30ms/1590ms respectively and Averages=1. 4 bottles having pure distilled water (for reference) in 1 of the bottles and 3 other bottles with different concentrations of ethanol (25%, 50% and 100%) were scanned with total acquisition time of 4 minutes and 30 seconds. The phantom set up is as shown in figure 1. Spectra of ethanol obtained from scanner was noisy due to low TR/TE and less averages.

MRF-Dictionary generation: Free Induction Decay (FID) closely approximating the response of metabolites for the molecular groups of ethanol i.e., CH₃, CH₂ and OH was generated based on equation as in³ given below,

$$S(k)=\sum_{q=1}^NA_{q}exp(-t_{k}/T_{2}^{q})exp(j(2\pi f_{q}t_{k} + \phi_{q}))$$

Where, S is the simulated FID having ‘q’ number of metabolite peaks each with amplitude A_q having an apparent relaxation time T₂^q, a frequency f_q and a phase φ_q and t_k is the k^th time point in the FID. FIDs for above mentioned metabolites were generated for given TE/TR and for different T₂ values between the range 50-720ms, 50-750ms and 20-90ms for CH₃, CH₂ and OH respectively. Spectra obtained by taking Inverse Fast Fourier Transform (IFFT) were added up and stored in dictionary.

Pattern matching: Pattern matching between the original spectra and dictionary was performed using vector dot product. To find the ppm shift of the spectra, cross correlation and circular shift operations were performed on the matched spectra from dictionary. The original and matched spectra were subjected to the following processing steps in jMRUI⁴: (a) Apodization (b) Phase Correction (c) Frequency Shift.

The spectra obtained from scanner was matched with one of the spectra in dictionary. Signal evolution curves for original spectra obtained from scanner and dictionary matched spectra are shown in figure 2. From figure 2, we can observe that the patterns of scanner spectra and spectra from dictionary were matching. Ratio of the 3 peaks in the spectrum obtained from dictionary was observed to be 0.57, 0.327 and 0.097 for CH₃, CH₂ and OH respectively as expected.Application of MRF on MRSI has been performed for the first time. This method aids in accelerated MRSI. Future work includes generation of metabolite maps and incorporation of CS for denoising of MRSI data.