[Insight-users] Re: question about demons

[Luis Ibanez]

Hi Alma,


     The computational cost of Demons is high.

                      but...

     so it is for *any* other deformable
     registration method.



If you are really interested in this subject, the right
way of asking the question is to select a set of methods,
specific images, specific constraints and then run
experiments on them, in order to determine which one
gives you reasonable results in an acceptable amont of
time.


At the end you will be able to conclude that method A with
the specific parameters (x,y,z,w...) is faster than method B
with the specific parameters (m,n,k,l...) for the specific
case of the images that you tested.



Any *generalization claims* are the evidence of *bad science*,
and *superficial thinking*. That's fine if your purpose in life
is to publish papers in vanity journals with the goal of
fulfilling your annual quota of "intellectual productivity",
but if you want to do honest technical work, then you must
stick to the scientific method, run experiments, and limit
your conclusions to the cases covered by the experiments.



For details on the correct way of doing reasoning in science,
please refer to the following textbooks:



   "The Logic of Scientific Discovery"    by Karl Popper


   "Conjectures and Refutations"          by Karl Popper



If what you want to do is to evaluate the number of mathematical
operations that are needed per iteration of the method, that's
a different technical questions, can also be easily addressed
by following the code. Note that at the end, it doesn't makes
much practical sense anyways, since the cost of one iteration
must be evauated in the context of how many iterations are needed
with this method in order to obtain satisfactory results.


Please let us konw if you have further questions,


    Regards,


      Luis



---------------------------------
Alma Rosas Rugerio Ramos wrote:
> Hello!
> 
> I would like to know about the computational cost of the demons deformable  
> registration and affine registration but in the paper of the Thirion doesn´t 
> mention it, can you help me please? 
> 
> thanks
> 
>