For its analysis, SMAL uses the raw MRS data, acquired with a phased array coil. To achieve proper coil combination, we assume that the raw data from each coil is provided as as a separate channel in the data. There are defaults set for the number of non-water-suppressed transients, and sampling rate, but these can be set as optional inputs to the user interface.
The main MRS.api user interface assumes that the data is organized as an array with shape: (transients, echos, channels, timepoints). For example, typical data might have the shape: (68, 2, 32, 4096), where 68 transients are acquired, each with two echos (GABA-edited and not GABA-edited), with a 32-channel phased array coil, acquiring 4096 samples in each FID curve at 5 kHz.
Alternatively, we can accept a full path to the data saved as a nifti file. Following the nifti convention, the data array in the file needs to be saved with the shape: (transients, 1, 1, echos, channels, timepoints). That is, for this data, the data array in the nifti file will have the shape: (4096, 1, 1, 68, 2, 32). Data acquired at the Stanford CNI, will be saved as a nifti file in this convention (see also the example data provided at the Stanford Data Repository).