Basis simulation is necessary step for modern MRS analysis and the this vignette will explain how to achieve this with spant. It is advisable to follow the examples given in the metabolite simulation vignette before following this guide.
Load the spant package:
A basis set is a collection of signals to be fit to the MRS data. In spant we start with a list of molecular definitions containing the relevant information for each signal - such as chemical shifts and j-coupling values:
mol_list <- list(get_mol_paras("lac"),
get_mol_paras("naa"),
get_mol_paras("cr"),
get_mol_paras("gpc"))
In the next step we convert these chemical properties into a
collection of signals (a spant basis_set
object) with the
sim_basis
function. When fitting, the signal parameters
(e.g. sampling frequency) and pulse sequence (e.g. echo-time) must match
the MRS data acquisition protocol.
basis <- sim_basis(mol_list, pul_seq = seq_slaser_ideal,
acq_paras = def_acq_paras(N = 2048, fs = 2000, ft = 127.8e6),
TE1 = 0.008, TE2 = 0.011, TE3 = 0.009)
stackplot(basis, xlim = c(4, 0.5), y_offset = 50, labels = basis$names)
In 1H MRS broad resonances from lipids and macromolecules are often included in addition to metabolites:
mol_list_mm <- append(mol_list, list(get_mol_paras("MM09", ft = 127.8e6)))
basis_mm <- sim_basis(mol_list_mm, pul_seq = seq_slaser_ideal,
acq_paras = def_acq_paras(N = 2048, fs = 2000, ft = 127.8e6),
TE1 = 0.008, TE2 = 0.011, TE3 = 0.009)
stackplot(basis_mm, xlim = c(4, 0.5), y_offset = 50, labels = basis_mm$names)
Note the field strength is often required to simulate these broad
resonances as their linewidth is usually specified in ppm. spant also
includes the functions sim_basis_1h_brain
and
sim_basis_1h_brain_press
to produce commonly used sets of
basis signals:
basis <- sim_basis_1h_brain()
stackplot(basis, xlim = c(4, 0.5), y_offset = 20, labels = basis$names)
Basis sets can be exported for use with LCModel with the
write_basis
function, and sim_basis_1h_brain has the option
lcm_compat
to remove signals that are usually generated
within the LCModel package: