> Could it be the Holy-Graal-of-morphometry?
Personally, I think not. If we knew how to generate accurate
simulations of brain disorders, then we would be using the same
generative models to fit our data. Evaluations based on simulations
will generally show that the method closest to the one used for
generating the data will appear to be the most accurate. It's all a
bit circular.
Best regards,
-John
On 1 June 2012 16:02, MS Al-Rawi <[log in to unmask]> wrote:
> Thank you all for the kind responses.
>
>> I think the best so far include SVM and other statistical classifiers (perhaps PRoNTo for SPM in fMRI), but not quite what you want I think.
> Yes Joao, not quite , I am looking for something that realistically deteriorate an image of a healthy brain. Could it be the Holy-Graal-of-morphometry? It could be, since we'll have higher flexibility in generating new data and test (within large scales) our morphometry approaches. In the end, however, real-world experiments may have the upper hand.
>
>> You mean software that can be used to simulate atrophy? If so, then
>> maybe NiftySim would do the job:
>
> Thank you John.
> Nifty Sim is a high-performance nonlinear dynamic finite element (FE) solver, that's what the manual says, and the rest of the details seems unrelated.
>
>
> I am not sure if NiftySim can do that? I couldn't find anything useful, am I missing something?
>
> The simulator I have in mind is one that converts a healthy structural MRI volume into one with some kind of neuorodegeneration. This simulator could be based on (neurodegenerative diseases, ie be it based on atrophy or other forms) probabilistic maps + few other chance parameters, and may uses morphology operators, eg dilation and/or erosion, to convert a healthy brain to non-healthy. How much neurodegeneration could be tuned according to a set of factors , e.g., the age of the non-healthy brain. For example, if the original healthy brain is 25 yrs, the simulator will be asked to produce a non-healthy output-brain at the age of 65, or as the user desires. It would be nice that such simulator can also produce some form(s) of scanner-noise, and affine(ly) transform the resultant non-healthy brain.
>
> Regards,
> - Rawi
|