Hi,
The published papers and documentation on SIENA detail all
of the tests which have been run to test how SIENA performs
under different circumstances. We have not explicitly tested
distorted images or a wide range of pathologies, beyond what
is detailed in the papers. The main thing is to get the best
quality images that you can. If you are doing a longitudinal
study with SIENA, then matching the scanner and sequence
is crucial, and it is the *changes* in distortion and other things,
rather than the base level which counts. However, in SIENAX
you need to be more careful, and the base level of distortion
would be important and needs to be minimised. It is extremely
difficult to give numerical information about how much things
change when you vary distortion or noise, as it depends on so
many aspects of the image quality. However, the advice is more
straightforward - do everything that you can to get the best
quality images.
As for different pathologies, there may be issues with segmentation
and masking if the pathologies are particularly severe. However,
you can check this manually by looking at the results and the
intermediate steps to assess the quality. As long as there is no
obvious problem then the edge-finding mechanism, which works
on the original grey-level data, should work fine. You can always
perform similar tests to those performed in the SIENA papers, by
testing repeat scans (which should show zero change) and seeing
what the detected change is. Also, the three point test: A-C vs A-B-C
is useful if you have the right kind of longitudinal data.
I hope this helps.
All the best,
Mark
Antonios - Constantine Thanellas wrote:
> Dear fsl users,
> I'm trying to find information regarding the assumptions that should be
> fulfilled for a valid siena/sienax analysis.
> One factor that can affect the validity of the results is the imaging
> acquisition characteristics [that has been mentioned in literature (Smith et
> al 2002) such as different slice thickness among subjects, different
> scanners, different sequences especially in case of Sienax which seems to be
> more vulnerable].
>
> What other issues besides the previous mentioned image characteristics can
> affect the validity of the results (In which cases, even if all the
> intermediate steps during the analysis seem to work ok, the results cannot
> be considered valid?).
> For example noisy images, distorted images ,or even physiological factors
> (hydrocephalus patients etc)
> Where can i find this information?
>
> Thank you
>
>
|