Search

Nuclear magnetic resonance spectroscopy (MRS) allows for the determination ofatomic structures and concentrations of different chemicals in a biochemicalsample of interest. MRS is used in vivo clinically to aid in the diagnosis ofseveral pathologies that affect metabolic pathways in the body. Typically, thisexperiment produces a one dimensional (1D) 1H spectrum containing several peaksthat are well associated with biochemicals, or metabolites. However, since manyof these peaks overlap, distinguishing chemicals with similar atomic structuresbecomes much more challenging. One technique capable of overcoming this issueis the localized correlated spectroscopy (L-COSY) experiment, which acquires asecond spectral dimension and spreads overlapping signal across this seconddimension. Unfortunately, the acquisition of a two dimensional (2D)spectroscopy experiment is extremely time consuming. Furthermore, quantitationof a 2D spectrum is more complex. Recently, artificial intelligence has emergedin the field of medicine as a powerful force capable of diagnosing disease,aiding in treatment, and even predicting treatment outcome. In this study, weutilize deep learning to: 1) accelerate the L-COSY experiment and 2) quantifyL-COSY spectra. We demonstrate that our deep learning model greatly outperformscompressed sensing based reconstruction of L-COSY spectra at higheracceleration factors. Specifically, at four-fold acceleration, our method hasless than 5% normalized mean squared error, whereas compressed sensing yields20% normalized mean squared error. We also show that at low SNR (25% noisecompared to maximum signal), our deep learning model has less than 8%normalized mean squared error for quantitation of L-COSY spectra. These pilotsimulation results appear promising and may help improve the efficiency andaccuracy of L-COSY experiments in the future.

1H Magnetic Resonance Spectroscopic imaging (SI) is a powerful tool capable of investigating metabolism in vivo from mul- tiple regions. However, SI techniques are time consuming, and are therefore difficult to implement clinically. By applying non-uniform sampling (NUS) and compressed sensing (CS) reconstruction, it is possible to accelerate these scans while re- taining key spectral information. One recently developed method that utilizes this type of acceleration is the five-dimensional echo planar J-resolved spectroscopic imaging (5D EP-JRESI) sequence, which is capable of obtaining two-dimensional (2D) spectra from three spatial dimensions. The prior-knowledge fitting (ProFit) algorithm is typically used to quantify 2D spectra in vivo, however the effects of NUS and CS reconstruction on the quantitation results are unknown. This study utilized a simulated brain phantom to investigate the errors introduced through the acceleration methods. Errors (normalized root mean square error >15%) were found between metabolite concentrations after twelve-fold acceleration for several low concentra- tion (

Attempts have been made to reduce the total scan time in multi-dimensional J-resolved spectroscopic imaging (JRESI) using an echo-planar (EP) readout gradient, but acquisition duration remains a limitation for routine clinical use in the brain. We present here a significant acceleration achieved with a 4D EP-JRESI sequence that collects dual phase encoded lines within a single repetition time (TR) using two bipolar read-out trains. The performance and reliability of this novel 4D sequence, called Multi-Echo based Echo-Planar J-resolved Spectroscopic Imaging (ME-EP-JRESI), was evaluated in 10 healthy controls and a brain phantom using a 3 T MRI/MRS scanner. The prior knowledge fitting (ProFit) algorithm, with a new simulated basis set consisting of macromolecules and lipids apart from metabolites of interest, was used for quantitation. Both phantom and in-vivo data demonstrated that localization and spatial/spectral profiles of metabolites from the ME-EP-JRESI sequence were in good agreement with that of the EP-JRESI sequence. Both in the occipital and temporal lobe, metabolites with higher physiological concentrations including Glx (Glu+Gln), tNAA (NAA+NAAG), mI all had coefficient of variations between 9-25%. In summary, we have implemented, validated and tested the ME-EP-JRESI sequence, demonstrating that multi-echo acquisition can successfully reduce the total scan duration for EP-JRESI sequences.