Sam entropy: Difference between revisions

From MEG Core
Jump to navigation Jump to search
Content added Content deleted
(Created page with " Return to Source Localization - SAM ==Description== The sam_entropy program computes the rank-vector entropy time series for a single-state d...")
 
No edit summary
 
Line 25: Line 25:
FilterType: Type of filter (FFT vs. IIR) used for bandpass
FilterType: Type of filter (FFT vs. IIR) used for bandpass
TimeStep: Sampling interval for output image
TimeStep: Sampling interval for output image
ImageMetric: Must be RankVectorEntropy, with the TAU (>0) and DIM (4 or 5) parameters defined
ImageMetric: Must be RankVectorEntropy, with the TAU and DIM parameters defined
RemoveBaseline: Specifies if and how a baseline should be removed
RemoveBaseline: Specifies if and how a baseline should be removed
CovType: must be GLOBAL
CovType: must be GLOBAL
Line 33: Line 33:
ImageFormat: Designates whether the weights are in +orig or +tlrc space
ImageFormat: Designates whether the weights are in +orig or +tlrc space
ImageDirectory: Directory where the output images are written, default is the SAM directory
ImageDirectory: Directory where the output images are written, default is the SAM directory
PrefixLength: Number of characters in the MEG dataset name used for naming MRI and output files.
PrefixLength: Number of characters in the MEG dataset used for naming MRI and output files.


==Output Files==
==Output Files==

Latest revision as of 13:40, 12 March 2019

Return to Source Localization - SAM

Description

The sam_entropy program computes the rank-vector entropy time series for a single-state dataset. The output image reflects the change in entropy relative to the mean entropy. Computing the rank vector entropy requires choosing several parameters. The embedding space dimension W, the integrator time constant tau, and a lag D. For sam_entropy, W must be either 4 or 5, and the lag is calculated from the sampling rate and the upper frequency limit. Entropy is calculated by first transforming the time series to a rank vector. A sub-window of W samples (with a lag of D between samples) is chosen from the time series. The measured signal from the W samples is replaced with the rank of each value. For a window W, there are W! possible rank vectors. A simple lookup table is used to convert the rank vector to a single state symbol. The window is then advanced by a single sample, and the same calculation occurs for the next data vector. As state symbols are generated, they are accumulated in a probability histogram, from which entropy can be calculated. In order to allow the distribution of states to vary over time, a decay parameter (tau) is defined for the histogram counts. Intuitively, this measure of entropy, or complexity, represents the amount of information which can be encoded in the signal at any given time.

This program is experimental. Multi-epoch data must be stored continuously.

Usage

 sam_epi -r <dataset_name> -m <parameter_file_name> [options]

The -r flag designates the dataset name (with or without the .ds suffix), and -m designates the parameter file name.

Other options:

 -v  Verbose mode, without this flag sam_entropy works silently except for error messages
 -n  Normalize by the variance

Required Parameters:

 CovBand: Bandpass for the covariance matrices (and directory for weights files)
 ImageBand: Bandpass for the image
 SmoothBand: designates the lowpass filter for smoothing the Hilbert envelope
 FilterType: Type of filter (FFT vs. IIR) used for bandpass
 TimeStep: Sampling interval for output image
 ImageMetric: Must be RankVectorEntropy, with the TAU and DIM parameters defined
 RemoveBaseline: Specifies if and how a baseline should be removed
 CovType: must be GLOBAL

Optional Parameters:

 ImageFormat: Designates whether the weights are in +orig or +tlrc space
 ImageDirectory: Directory where the output images are written, default is the SAM directory
 PrefixLength: Number of characters in the MEG dataset used for naming MRI and output files.

Output Files

The output consists of a single 3D+time NIFTI image of rank vector entropy.