Genetic algorithms based finite mixture modelling for neuroimaging applications
(These pages are under construction)

Jussi Tohka

Department of Signal Processing Tampere University of Technology, Finland

Evgeny Krestyannikov, Ulla Ruotsalainen
Department of Signal Processing Tampere University of Technology, Finland

Ivo D. Dinov, Allan MacKenzie-Graham,David W. Shattuck, Arthur  W. Toga
Laboratory of Neuro ImagingDepartment  of Neurology , UCLA  School of Medicine,  CA, USA

What is it?

MIXTUREGA is a collection of software for solving complex finite mixture model parameter estimation problems in neuroimaging. At the moment, the Matlab package for the FMM parameter estimation is available. In near future, also source code in the C language will be made available. However, I want to extend the C-code a bit to increase its applicability. If you simply cannot wait for C-code, then just send me e-mail. Also other additions are likely.

The algorithm is described in:

J. Tohka, E. Krestyannikov, I.D. Dinov, A. MacKenzie-Graham, D.W. Shattuck, U. Ruotsalainen, A.W. Toga.
Genetic algorithms for finite mixture model based voxel classification in neuroimaging.
IEEE Transactions on Medical Imaging , 26(5):696 - 711, 2007.

J. Tohka, E. Krestyannikov, I. Dinov, D. Shattuck, U. Ruotsalainen, A.W. Toga.
Genetic algorithms for finite mixture model based tissue classification in brain MRI.
In  Proc. of European Medical and Biological Engineering Conference, IFMBE Proceedings vol. 11, pp. 4077 - 4082, Prague, Czech Republic, 2005.

Where can it be applied?

FMM parameter estimation has many applications in neuroimaging. The papers above describe the potential applications to tissue classification in human as well as mouse brain MRI and tissue classification in parametric FDG-PET imaging. However, FMMs form a very general data analysis tool and we expect the algorithm to find other applications as well.


Matlab, version 5.3 or higher


  See the file gamixture_simple.m


Current version 1.1 including Matlab tools for solving parameter estimation problems within 1 and d-dimensional FMMs:

Documentation for the version 1.1

Version 1.2 including a tissue classification example with mouse MRI. Otherwise the same as the version 1.1. The documentation for the version 1.1 is to be used also with this version:

Version 1.2.1 containing fixes to make the toolbox compatible with the new versions of Matlab (see below for more info). Otherwise the same as the version 1.2:
More info on fixes: The unidrnd function of the matlab stopped working properly at some point (I noticed this with version (R2010b), but the same issue could occur with the earlier versions as well). To fix the issue, the unidrnd function calls have been replaced from mixturega_tournamentselection.m and mixturega_BLXxover.m. At the same time, calls to normpdf were also replaced in order to reduce (and hopefully remove) the dependency on statistics toolbox.
Please read the README-file before using the software.

The software can be downloaded also from the Web-site of the Laboratory of Neuro Imaging , University of California, Los Angeles.