Hi Steve,
First sorry for taking so long for my reply and thanks for kindly having a look at my output.
Obviously I was a bit too keen to get started and see some sienna output and I had not done enough checking :-p. I will check the pairing of scans one by one for the proper analysis.
For the magnet, the scanner was changed between the waves of data collection but with a similar Philips scanner however, the quality was not as good during the first wave due to the protocol used. Given the article you have published validating the method and showing very little error across different scanners, and if my recollection is correct, even when different slice thicknesses were used, I am a bit surprised that it is an issue. Could you please elaborate, since I am sure that many people involved in longitudinal studies are confronted with the same problem.
Thanks again for your generous advice. I am looking forward to the course in Cardiff in September where no doubt I will learn a lot more but in the mean time I have to keep going...
Cheers
Nic
-----Original Message-----
From: FSL - FMRIB's Software Library on behalf of Steve Smith
Sent: Tue 20/02/2007 20:29
To: [log in to unmask]
Cc:
Subject: Re: [FSL] siena and assessing quality of output
Hi Nicolas, thanks for sending your files.
I wouldn't worry about the slight brain-extraction variations, that
probably won't make much difference to the final SIENA measure.
However, i WOULD be VERY worried about the fact that one timepoint is
left-right flipped wrt the other timepoint, in both datasets you
sent! This completely invalidates the SIENA output. It's very
important to check for this kind of mistake in the input data.
I would also be quite concerned that the images look so different
from each other - were different coils used for the acquisitions? We
would not generally recommend running structural analysis on studies
where the hardware has changed a lot.
Cheers, Steve.
On 18 Feb 2007, at 23:52, Nicolas Cherbuin wrote:
> Hi Steve,
>
> I just wanted to confirm that you have received my previous email (see
> below) saying that I have uploaded these files.
>
> Thanks
>
> Nic
>
>
> Thank you for responding to my email. I have uploaded the file as
> requested. The ID is 886269. My concern is that some non-brain is
> being
> included in one of the paired scans and I am not sure how to deal with
> it. I have had a look at the course notes. Thanks, every bit helps. Is
> there something similar for the surface-based analysis in siena which
> will be my next challenge after resolving the scan quality/
> segmentation
> issue.
>
>
> Is there an intensive fsl/freesurfer course coming up? Being based in
> Australia I cannot attend short courses on the other side of the world
> but I would consider week-long courses particularly if they had a
> strong
> structural component.
>
>
> I am very grateful for fsl software and support.
>
>
> Nic
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
> Behalf Of Steve Smith
> Sent: Tuesday, February 13, 2007 8:39 PM
> To: [log in to unmask]
> Subject: Re: [FSL] siena and assessing quality of output
>
> Hi Nic,
>
> You can upload the files in a single compressed tarfile to
> http://www.fmrib.ox.ac.uk/cgi-bin/upload.cgi
> and then email me the upload ID. I would send the full results of
> running SIENA, including the inputs, and use the "-d" option to keep
> the intermediate stages.
>
> The number output by SIENA is described in the second para of the
> manual - it's the percentage brain volume change (PBVC) between two
> input images, taken of the same subject, at different points in time.
>
> There isn't a huge amount more detail, but you might find it
> informative to look through the course lecture slides:
> http://www.fmrib.ox.ac.uk/fslcourse/
>
> Cheers, Steve.
>
>
>
> On 13 Feb 2007, at 03:33, Nicolas Cherbuin wrote:
>
>> Dear FSLers,
>>
>> I want to run siena to compare a clinical group and a control group
>> with
>> scans four years apart. The scans were of lesser quality in the first
>> than in the second wave. As a first test, as I am no expert, I have
>> run
>> siena on two subjects. I had attached the gifs of the results but
>> it did
>> not make it to the list since the message was too large.
>>
>>
>> My questions are:
>> 1. I would like to know if my results look reasonable, is there a
>> drop
>> folder for somebody to have a look at them? and is the red outline
>> the
>> second scan overlayed on the first one (or the reverse)?
>> 2. What should I do to improve the results and apart from looking
>> at the
>> GIF what is the best way to assess the quality of the registration?
>> 3. At the end of processing siena spits out -6.332, is it the volume
>> difference in mm3? (where can I find all the volumes?)
>> 4. I have looked at the siena website, searched for
>> tutorials,looked at
>> the mailing list, read the attached pdfs, and browsed the wiki but it
>> seems that there is not enough detailed step-by-step information
>> for a
>> new user like me. Is there a detailed tutorial somewhere dealing not
>> just with process but with quality of output and what to do to
>> improve
>> it?
>>
>> Thanks very much for your continuing support
>>
>> Nic
>
>
> ----------------------------------------------------------------------
> --
>
> ---
> Stephen M. Smith, Professor of Biomedical Engineering
> Associate Director, Oxford University FMRIB Centre
>
> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
> ----------------------------------------------------------------------
> --
>
> ---
------------------------------------------------------------------------
---
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
------------------------------------------------------------------------
---
|