Combining parallel detection of proton echo planar spectroscopic imaging (PEPSI) measurements with a data‐consistency constraint improves SNR. (20th October 2015)

• Record Type: Journal Article
• Authors: Tsai, Shang‐Yueh; Hsu, Yi‐Cheng; Chu, Ying‐Hua; Kuo, Wen‐Jui; Lin, Fa‐Hsuan
• Is Part Of: NMR in biomedicine. Volume 28:Number 12(2015:Dec.)
• Page Start: 1678
• Page End: 1687
• Publication Date: 2015-10-20
• Subjects: proton echo planar spectroscopy imaging -- data consistency -- SNR -- regularization; Nuclear magnetic resonance -- Periodicals; Magnetic Resonance Spectroscopy -- Periodicals; 574
• DOI: 10.1002/nbm.3425
• Languages: English
• ISSNs: 0952-3480

Abstract

Abstract : One major challenge of MRSI is the poor signal‐to‐noise ratio (SNR), which can be improved by using a surface coil array. Here we propose to exploit the spatial sensitivity of different channels of a coil array to enforce the k ‐space data consistency (DC) in order to suppress noise and consequently to improve MRSI SNR. MRSI data were collected using a proton echo planar spectroscopic imaging (PEPSI) sequence at 3 T using a 32‐channel coil array and were averaged with one, two and eight measurements (avg‐1, avg‐2 and avg‐8). The DC constraint was applied using a regularization parameter λ of 1, 2, 3, 5 or 10. Metabolite concentrations were quantified using LCModel. Our results show that the suppression of noise by applying the DC constraint to PEPSI reconstruction yields up to 32% and 27% SNR gain for avg‐1 and avg‐2 data with λ = 5, respectively. According to the reported Cramer–Rao lower bounds, the improvement in metabolic fitting was significant ( p < 0.01) when the DC constraint was applied with λ ≥ 2. Using the DC constraint with λ = 3 or 5 can minimize both root‐mean‐square errors and spatial variation for all subjects using the avg‐8 data set as reference values. Our results suggest that MRSI reconstructed with a DC constraint can save around 70% of scanning time to obtain images and spectra with similar SNRs using λ = 5. Copyright © 2015 John Wiley & Sons, Ltd. Abstract : An MRSI reconstruction algorithm enforcing the k ‐space data consistency (DC) among channels of a coil array is proposed to suppress noise and consequently to improve SNR. Our results show that the suppression of noise by applying a DC constraint ( λ ) to proton echo planar spectroscopic imaging yields up to 32% and 27% SNR gain for avg‐1 and avg‐2 data with λ = 5, and this results in the lowest spatial variation and root‐mean‐square errors using an avg‐8 data set as reference. DC reconstruction with λ = 5 is recommended to save around 70% of scanning time.