> I’m trying to interpret the output parameters from eddy. The output for each volume consists of 16 values with the first 6 pertaining to subject movement. The last 10 represent the EC-induced field as a quadratic model. For one volume, the 10 EC parameters were:
> -0.1300904739 -0.007051537702 -0.09431927767 0.0005882496583 -0.0002554671209 -0.00105972332 -0.002009765693 0.001173003314 -0.002431384829 -2.57769128
> How do I interpret these
The three first encodes linear gradients in the x-, y- and z-directions respectively. The scale is Hz/mm, so the first term there (-0.13) indicates an x-gradient with stronger field for lower x-coordinate (the minus sign) with the strength 0.13Hz/mm. If you want to know the linear part you just add up the x-, y- and z-gradients.
The quadratic terms are a little more tricky to interpret since there is an internal re-scaling of them to ensure that the Hessian for each Gauss-Newton step does not get too poorly conditioned.
The final number is (tentatively) the constant term of the EC field. I say tentatively because it is very hard to disentangle the constant term from subject movement (translation) in the PE-direction. The sum of the two is well determined (which is what matters for the correction), but the individual values are not.
> and how do I get it in voxel space?
I am assuming that by that you mean “how do I generate a field from them”. As I indicated above it is a little tricky with the quadratic terms, but it is easy to get eddy to write them out for you. Just add "--fields” to your command line and you will get an eddy current-induced field per input volume. Provided you have given the correct “total readout time” (the fourth column in the acqp file) those fields will be scaled in Hz.