Content-type: text/html Manpage of orientbiasmap


Section: User Commands (1)
Index Return to Main Contents



orientbiasmap - Computes statistics of populations of diffusion tensors oriented over a hemisphere.



orientbiasmap -schemefile <scheme> -snr <value> [options]



Used to evaluate the robustness of an imaging scheme to imaging noise, as a function of orientation.

The orientation bias is evaluated by sampling an identical Gaussian diffusion test function at different orientations. For each point on a regular grid of spherical coordinates, we synthesize noisy data and evaluate the results. The regular grid of spherical coordinates (theta, phi) is sub-optimal in that it oversamples the pole of the hemisphere (where theta is small) compared to the equator. However, it is convenient for 2D display.

For each point on the grid, we synthesize a set of DWI images using the imaging scheme, fit the diffusion tensor, and produce the following statistics. All statistics are written as 2D matrices in CSV files, with each row at a constant theta and each column at a constant phi. The output file extensions are:

  meanfa.csv - Mean FA recovered from the noisy data. 
  stdfa.csv  - Standard deviation of FA
  meantr.csv - Mean trace
  stdtr.csv  - Standard deviation of trace
  dyadl1.csv - Principal eigenvalue of the mean dyadic tensor.

  dyadbias.csv   - The angle (in degrees) between the primary eigenvector of the mean dyadic 
                   tensor and that of the test function.

  coneUncertainty.csv - Aperture f the 100 * (1 - alpha) % cone of uncertainty (in degrees),
                        such a cone centered on the mean vector would contain 
                        100 * (1 - alpha) % of samples orientations. 

The diadbias.csv is a measure of the angle between the mean of the sample axes, and the actual test function. It ranges betwen 0 (perfect alignment) and 90 (orthogonal), but in practice it should not be more than a few degrees.

The interpretation of the mean and standard deviation of FA and Tr will depend on the input test function. The default test function FA is 0.8, with a trace of 2.1E-9. The typical trace of diffusion tensors in the human brain is about 2E-9 m^2 / s. If your scheme file does not have b-values in matching units (s / m^2), then adjust trace accordingly. For example, if your scheme file has b=1000 s / mm^2, you would want to do something like "-trace 2.1E-3".



pointset2scheme -inputfile camino/PointSets/Elec012.txt -bvalue 700 -addzeromeas 1 -numscans 3 > M1_N12_b700_R3.scheme

orientbiasmap -schemefile M1_N12_b700_R3.scheme -trace 2.1E-3 -fa 0.75 -snr 12 -outputroot M1_N12_b700_R3_ldt_



-schemefile <code>
The imaging scheme to be evaluated.

-snr <value>
The estimated signal to noise ratio in the b=0 image.

-outputroot <root>
String prepended to output file names.

-fa <value>
The FA of the test function. The default is 0.8.

-trace <value>
The trace of the test function. The default is 2.1E-9.

-samples <number>
The number of synthetic measurements to generate at each point. The default is 500.

-model <code>
The model used to fit the tensor data, may be linear (ldt), weighted linear (ldt_wtd) or another single-DT fitting method. The default is ldt.

-step <thetaStep> <phiStep>
The resolution of the grid, in degrees. The default is equivalent to "-step 2 4". The sphere is not sampled evenly in this way; the poles (theta close to 0) will be oversampled compared to the equator. However it is convenient to display on a regular grid.

-alpha <alpha>
The alpha value to use in computing the cone of uncertainty. Default 0.05.



Philip Cook <>



invstats(1), mfrstats(1)






This document was created by man2html, using the manual pages.
Time: 02:07:11 GMT, December 04, 2017