Deconvolution IRF from fluorescence decay curves

[sunny bob]

Hi There,

I have been working on fluorescence decays of the fluorophores whose lifetimes are comparable to the instrument response function(IRF) of my device. The technique I use is based on Time-correlated single photon Counting(TCSPC). I have been fitting the decay curves using multiple exponentials in Origin 8, which fits it on the basis of Non-linear least square fitting(NLSF) method. However, the fit is independent of any information about IRF.

My perception was that though the NLSF method doesn't take IRF into account, but because it fits only to the exponential forms, and indirectly it implies that it would not fit the non-exponential forms of IRF. Thereby fitting the data along with deconvolution, in which case it would never need to know the IRF.

I was just wondering if Iam right in thinking so or do I have to go for deconvolution first followed by fitting. Any source of code for deconvolution and fitting would be very much appreciable.

Thanks all,
Sunny.

 

[Rajini]

Hi,
In fitting to an experimental curve one normally use NLSF or Marquardt-Levenberg (ML is also called least-squares menthod) procedure. These two methods are just fitting..
If your curve is Lorentzian or Gaussian or exponential decay or any other function...you can always use any of these two methods. That means method used for fitting is never dependent on any theory. But you should use the right function to fit. For example, if you know that your experimental data is a Lorentzian function, then you should use the Lorentzian function to fit (you should not use Gaussian or exponential). But the fitting method can be NLSF or ML. NLSF or ML is just a tool to fit experimental data with some theory (You should choose the right theory). Got it?

 

[sunny bob]

Hi Rajini,

Thanks for responding. I got your point. But I think I did not frame my question well. Let me try to put it again. I could be able to fit the exponential decay data with exponential function. However the concern is the deconvolution of the Instrument response function(IRF, due to device) from the experimental decay data before fitting it. The NLSF technique(used Origin 8 package) fits the curve but it doesn't ask for any input of IRF. So obviously it appears like it doesn't do any deconvolution thing.
However, my perception is that because i have chosen an exponential curve to fit an exponentially decaying data points, the fit should obviously be circumventing the non-exponential form of IRF(assuming it is some kinda gaussian). So even if the fitting procedure doesn't ask for IRF data input, it is (unknowingly) doing the deconvolution procedure too while fitting.
Do you think that is not right?

Thank you,
Sunny.

 

[Rajini]

Hi Sunny,

First you deconvolute the experimental data with a instrument function (i guess you know the instrument function, Lorentzian or Gaussian ?)
And then fit the deconvoluted data by exponential function.
Fitting is just a fitting using NLSF or ML. So in the fitting routine the convolution or deconvolution are not considered.
I am not a expert in Origin (I use it only for fitting especially using peak analysis). I assume deconvolution can also be performed using Origin.
Hope this helps.
cheers

 

[sunny bob]

Thanks Rajini! That was a helpful discussion. Thanks again.

 

[sunny bob]

Hi Rajini,

By any chance, do you have or know any free source of program written for deconvolution using NLSF? I have seen a code in O' Connors's book on "Time-correlated single photon counting" but as Iam a beginner in programming, I fear if it might take ages to finish it. Do you have any recommendation?

Thanks,
Sunny.

 

[Rajini]

Hi Sunny,

I don't have any idea for deconvolution software.
But you can try to find some information in Matlab or Mathematica (both are not free).
I think if you have origin you can easily perform deconvolution.
But may be you can find ready made c codes or fortran in the book: Numerical recipes The Art of Scientific Computing.

 

[Cthugha]

If you have Origin 8, you can do it easily in Origin. There is even a tutorial on how to do deconvolution if you know the IRF on their web page:
http://www.originlab.com/index.aspx...aAnalysis/SignalProcessing/Convolution&pid=67

I am not sure about Origin 7.5, but I think something similar should be possible in that version, too.

 

[Rajini]

Origin is not free. But for few days you can use the evaluation version.

 

[sunny bob]

Thanks you guys! I tried the FFT deconvolution method, in Origin, before. However, it didn't give satisfying result.