I am trying to understand the cost function values that are coming out flirt. I am using the following command

flirt -in inimg.nii -ref refimg.nii -schedule $FSLDIR/etc/flirtsch/measurecost1.sch -cost somecostfunction

The cost function values are not what I would expect and in many cases are out side of the possible range of many of the cost functions. For instance, for some reason I am getting negative numbers for mutual information. In one of my tests I set the reference and input image as the same image. I get the following values for different cost functions

mutualinfo = -0.972365

corratio = 0.0138204

normcorr = 0.00646079

normmi = -1.39582

Is this what I would expect?

a cost-function is just a proxy for “how well registered” two images are. In general their actual values is of no consequence. The only thing that matters is where in the parameter space that the minimum occurs. So for example with something like MI it is actually the maximum MI that indicates that the images are registered, so therefore one uses the negation of the MI instead. Also, any terms in the calculation of the MI that are constant (with respect to the parameters) one can chose not to calculate, thereby speeding up the calculations. Therefore the values you see may not be in the range you expect, nor even have the expected sign.

I think the details are in MJ’s original flirt paper.

Jesper

Aaron Tanenbaum

 

---

The materials in this message are private and may contain Protected Healthcare Information or other information of a sensitive nature. If you are not the intended recipient, be advised that any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited. If you have received this email in error, please immediately notify the sender via telephone or return mail.

---

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1