Meeting Banner
Abstract #3968

Metabolite Map Estimation from Undersampled Spectroscopic Imaging Data Using N-Compartment Model

Itthi Chatnuntawech1, Berkin Bilgic1, Borjan Gagoski2, Trina Kok1, Audrey Peiwen Fan1, Elfar Adalsteinsson1, 3

1EECS, Massachusetts Institute of Technology, Cambridge, MA, United States; 2Fetal-Neonatal Neuroimaging & Developmental Science Center, Boston Children's Hospital, Harvard Medical School, Boston, MA, United States; 3Harvard-MIT Division of Health Sciences and Technology, Cambridge, MA, United States

Specific physiological abnormalities could be detected by irregular change of metabolite concentration in specific brain regions. The combination between fully sampled spectroscopic imaging data and segmented structural image has been used to estimate metabolite value at each voxel. This abstract presents an N-compartment-model method with polynomial masks to obtain metabolite maps from undersampled spectroscopic imaging data. With the assumption that metabolite value within the same tissue type is slowly varying, the information of tissue boundaries is obtained from segmented structural image. Then, a regularized reconstruction with priors is formulated to reconstruct the metabolite maps. By acquiring only a subset of k-space samples, the acquisition process is sped up, while high reconstruction quality is retained via prior knowledge of tissue boundary and structure of data.

Abstract PDF Presentation Video

Keywords

abnormalities acceleration accommodate accurate acquiring acquisition addition allowed amplitude applied appropriate assumes audience boundaries brain chemical children chosen coefficients combination compartment compartments composition computed concentration conducted consideration constant constraint corner depict detected determined developmental deviation division domain dominating enforced equal errors estimation evaluated excite expected experiment expressed expressing extended fetal formulated fully generated grid gridded hand health hospital impose in vivo includes instance institute investigators irregular isotropic knowledge laser leads least length limitation lipid maps mask masks matrix medical metabolite minute minutes model modeled neonatal neuron norm numerical observation orientation original peak phantom physiological polynomial post prior priors process proposed quality real reconstruct reconstructed reconstruction reconstructions reduced regularization regularized resemble resolution respected respectively retained sampled samples scaling scanner school science sciences scientists segmented sets setup slowly smaller solution solving space sparsity spectra spectroscopic sped square structural structure subset superposition suppressed target technology term tissue transform type unknowns upper various varying volume water wavelet