[IGSTK-Developers] Re: Landmark3DRegistration

[Andinet Enquobahrie]

Hello David,

Thank you for pointing that out.  A while ago, I had plan to add an 
error checking mechanism, but for some reasion it slipped my mind...
I will add one soon...

-Andinet

> Hi Andinet,
>
> I looked at the Landmark3DRegistration registration code that
> you just committed, and it looks like it requires the following
> sequence of events:
> 1) ImageLandmark is added
> 2) TrackerLandmark is added
> 3) repeat 1 and 2 at least two more times
> 4) compute transform
>
> This is good, the only thing that is missing is that at
> step 4, there should be a check to make sure that the
> resulting registration is accurate enough.
> A simple check is to make sure that the RMS of the registration
> (the square root of the sum of the squares of distances between
> the image and tracker landmarks after registration) is less than
> some threshold e.g. 3 mm, that the application should be able to
> specify via a Set method.
>
> A succinct description of error calculation for landmark registration
> is given in the following paper:
> "Fiducial Point Placement and the Accuracy of Point-based, Rigid Body
> Registration", West et al., Neurosurgery 48:810-817, 2001.
> The methods described in the paper are a bit more advanced than what
> we want to implement in this iteration, but the paper also provides a
> very good description of all of the issues involved.
> I'll see if I can dig up a copy.
>
> - David
>
>
>
>
>