>>> Proposition #6 from my thesis: the use of FFT to extract clinically 
>>> relevant parameters should be forbidden in the field of cardiology.
>>
>> Uh... I was under the impression that most of the useful information 
>> in a cardiagram is in the time-domain anyway...?
> 
> Spot on.

Uh... so you're telling me I know more about cardiagrams than the people 
attempting to interpret them? o_O

> I once saw this procedure to find a phenomenon that occurs every other 
> beat.: Find the time instant in every beat where you expect the change. 
> Create a time series of that. Do an FFT. look at the amplitude of the 
> center frequency. Ok, I admit it works and you can include a spectrum in 
> your paper...

I have a vague recollection that somebody designed an ANN that worked on 
ECG data in the frequency domain and was reputedly quite accurate. OTOH, 
the details were sparse. (It was an introductory text on ANNs, basically 
it had a section on "hey, look how useful these are! You should learn 
about them.")

Post a reply to this message