DKI levenberg marquardt
in [Matlab]
|
From: Luis Lacerda on 17 Jul 2010 06:31 Hello! I'm current developping code in Matlab to receive DKI (Diffusional Kurtosis Imaging) images and do a parametric fitting with the levenberg marquard method, in order to discover the aparent diffusion coeficient and the kurtosis term, expressed in the paper: Diffusional Kurtosis Imaging - The Quantification of Non-Gaussian Water Diffusion by Means of Magnetic Resonance Imaging (page 4) by the equation - Sexp= {n^2 * (So *exp(-b*Dapp + 1/6*b^2*Dapp^2*Kapp)^2)}^1/2 where n is noise, S0 singnal intensity, b difusion coeficient, Dapp aparent difusion coeficient and Kapp kurtosis difusion coeficient. I started out to obtain information about a single voxel in the images with diferent values of coeficient of difussion, more particularly b=0; 500; 1000; 1500; 2000;2500, and i'm now trying to data fit with the levenberg marquardt method, but i'm having some dificulties. Therefore, i would like to ask you if you could explain me how should i do it, or please indicate me some resources that you find useful to this case. Thank you so much in advance. Luis Lacerda
From: Alan Weiss on 19 Jul 2010 08:23 lsqnonlin or lsqcurvefit from Optimization Toolbox have the option of using a Levenberg-Marquardt algorithm. Alan Weiss MATLAB mathematical toolbox documentation On 7/17/2010 6:31 AM, Luis Lacerda wrote: > Hello! > > I'm current developping code in Matlab to receive DKI (Diffusional > Kurtosis Imaging) images and do a parametric fitting with the levenberg > marquard method, in order to discover the aparent diffusion coeficient > and the kurtosis term, expressed in the paper: Diffusional Kurtosis > Imaging - The Quantification of Non-Gaussian Water Diffusion by Means of > Magnetic Resonance Imaging (page 4) by the equation - > Sexp= {n^2 * (So *exp(-b*Dapp + 1/6*b^2*Dapp^2*Kapp)^2)}^1/2 > > where n is noise, S0 singnal intensity, b difusion coeficient, Dapp > aparent difusion coeficient and Kapp kurtosis difusion coeficient. > I started out to obtain information about a single voxel in the images > with diferent values of coeficient of difussion, more particularly b=0; > 500; 1000; 1500; 2000;2500, and i'm now trying to data fit with the > levenberg marquardt method, but i'm having some dificulties. Therefore, > i would like to ask you if you could explain me how should i do it, or > please indicate me some resources that you find useful to this case. > > Thank you so much in advance. > Luis Lacerda
From: Luis Lacerda on 21 Jul 2010 04:56 Alan Weiss <aweiss(a)mathworks.com> wrote in message <i21g3n$f59$1(a)fred.mathworks.com>... > lsqnonlin or lsqcurvefit from Optimization Toolbox have the option of > using a Levenberg-Marquardt algorithm. > > Alan Weiss > MATLAB mathematical toolbox documentation > > On 7/17/2010 6:31 AM, Luis Lacerda wrote: > > Hello! > > > > I'm current developping code in Matlab to receive DKI (Diffusional > > Kurtosis Imaging) images and do a parametric fitting with the levenberg > > marquard method, in order to discover the aparent diffusion coeficient > > and the kurtosis term, expressed in the paper: Diffusional Kurtosis > > Imaging - The Quantification of Non-Gaussian Water Diffusion by Means of > > Magnetic Resonance Imaging (page 4) by the equation - > > Sexp= {n^2 * (So *exp(-b*Dapp + 1/6*b^2*Dapp^2*Kapp)^2)}^1/2 > > > > where n is noise, S0 singnal intensity, b difusion coeficient, Dapp > > aparent difusion coeficient and Kapp kurtosis difusion coeficient. > > I started out to obtain information about a single voxel in the images > > with diferent values of coeficient of difussion, more particularly b=0; > > 500; 1000; 1500; 2000;2500, and i'm now trying to data fit with the > > levenberg marquardt method, but i'm having some dificulties. Therefore, > > i would like to ask you if you could explain me how should i do it, or > > please indicate me some resources that you find useful to this case. > > > > Thank you so much in advance. > > Luis Lacerda i already found those, but i can't set the option 'Algorithm' to 'levenberg-marquardt'.. i have called the structure options with the parameter 'Algorithm' to 'levenberg-marquardt' but it won't "stuck". Can you tell me why? Thank you Luis
From: Alan Weiss on 21 Jul 2010 08:46 > i already found those, but i can't set the option 'Algorithm' to > 'levenberg-marquardt'.. > > i have called the structure options with the parameter 'Algorithm' to > 'levenberg-marquardt' but it won't "stuck". > > Can you tell me why? > > Thank you > > Luis Either your version of MATLAB is older than R2008b, when the 'Algorithm' option became available for lsqnonlin and lsqcurvefit, or you are not using the syntax (for optimset or for lsqnonlin) correctly. Alan Weiss MATLAB mathematical toolbox documentation
|
Pages: 1 Prev: maximizing of an objective function Next: Enumeration in Matlab. Please, need help. Thank You. |