Hi,
Problem solved.
You filled the non-brain part of the image with ones not zeros.
If you just subtract 1 from the whole image then it works OK.
There is some bias field present and you might want to play with
the --iter and --lowpass options to get a good bias field removal,
but apart from that I don't think you'll have any trouble.
All the best,
Mark
On 9 Jul 2009, at 15:37, Ronan Joseph Kelly wrote:
> Hi Mark,
>
> Thanks for that. Yes I've skull-stripped the brains all right, and
> I've expanded the image dimensions etc. I took a look at the
> histogram; it starts with a really tall peak at the beginning,
> followed by a much smaller, smoother one after that. The rest of the
> histogram is flat. How would I go about removing the bias field?
> I'll upload an image for you to take a look.
>
> Cheers,
> Ronan
>
> 2009/7/9 Mark Jenkinson <[log in to unmask]>
> Hi,
>
> Just to comment on this - FAST does not use stored
> values, it always estimates them from the data. The
> issue is that for some reason it can't do an initial
> split between different classes, which may be for the
> reasons I quoted in the previous email.
>
> However, you can manually initialise values too,
> although the failure is often an indication of a bigger
> problem which needs to be solved.
>
> All the best,
> Mark
>
>
> On 9 Jul 2009, at 15:06, Christoph Blau wrote:
>
> Hiya,
>
> I think we have to create our own kmeans file, as the ones stored are
> for human data.
>
> from the FAST pages under ADVANCED OPTIONS:
>
> "Use file of initial tissue-type means" tells FAST to use a text file
> with mean intensity values (separated by newlines) for the starting
> mean values of the different classes to be segmented. This is then
> used instead of the automated K-means starting parameter estimation.
>
> For rats, I'm getting means (from paravision, on manual regions) in
> the order of:
> Grey matter: 50,000
> White matter: 38,000
> ventricle: 90,000
>
> This is for rats, though, so mice might be different again.
> Not sure how you have to write the file though.
>
> 2009/7/9 Ronan Joseph Kelly <[log in to unmask]>:
> Hi,
>
> I've just started using Fast 4.1 to segment some high-res mouse data.
> However when I run the command I get a strange error back:
>
> Exception: Not enough classes detected to init KMeans
>
> Does anyone have any suggestions for what's going on? Or any tips to
> improve
> my results once it's working?
>
> Thanks in advance,
> Ronan
>
> --
> Ronan Kelly
> Neuroinflammatory Research Group
> Trinity College Institute of Neuroscience
> University of Dublin, Trinity College
> Dublin 2
> Ireland
>
>
>
>
> --
> Christoph Blau, MVB
> Trinity College Institute of Neuroscience,
> Lloyd Building,
> Trinity College,
> Dublin 2,
> Ireland
> Phone: +353-(0)1-8968412
>
>
>
>
>
> --
> Ronan Kelly
> Neuroinflammatory Research Group
> Trinity College Institute of Neuroscience
> University of Dublin, Trinity College
> Dublin 2
> Ireland
|