Dear Gurus,

Any help on this is appreciated. I am trying to fit T2 relaxtion times (MRI) to a biexponential curve.
y=A.exp(x/TA)+B.exp(x/TB);

The points fit well on a single exponential curve. The starting values chosen are 1500 and 100, for A and TA, and they stabilize at 400 and 500 resp.
However, if I use a biexponential curve fit, with starting values of 200, 40, 1000 and 100 for A,TA,B,TB resp. they do not converge.

I am using a minimum of 5 and a maximum of 15 points for fitting the curve at each pixel in the MRI image. I wonder why the parameters do not converge.

Any help is appreciated!!!

Thanks in advance,
JInesh Jain

hi,
It is necessary to provide initial guess to levenberg-marquardt (LM) Method. My obsevation is that when the initial guess is too much closer to actual solution then only the convergence is possible, otherwise it is not. i think this is the drawback of this method. so to get good solution you should provide some good initial guess.
bye
dipak