SpaCEM3- Spatial Clustering with EM and Markov Model




                          



About the Software

The SpaCEM3 software is dedicated to Spatial Clustering with EM and Markov Models. It proposes a variety of algorithms for supervised and unsupervised classification of multidimensional and spatially-located data. The main techniques use the EM algorithm for soft clustering and Markov Random Fields (MRF) for spatial modelling. The learning and inference parts are based on recent developments in mean field-like approximations. Its applications range from image segmentation (e.g. tissue detection in MRI, retrieval of planet surface properties from hyperspectral satellite images...) to gene clustering (e.g. biological module detection), remote sensing and mapping epidemics of ecological species.

The main functionalities of the program include:

  • Model-based unsupervised segmentation including the standard EM algorithm for mixtures and Hidden Markov Random Field models
  • Model selection for the Hidden Markov Random Field model
  • Simulation of commonly used Hidden Markov Random Field models
  • Simulation of independent Gaussian noise for noisy images
  • New Markov models including various extensions of the Potts model and triplet Markov models
  • Additional treatment of very high dimensional data using dimension reduction techniques within a classification framework
  • Models and methods allowing supervised classification with new learning and test steps
  • Integrated treatment of missing observations
  • Summary statistics of the data and visualization


Information / News

    A new version spacem3-2.0 is available.
This version includes a graphical user interface and is available for Windows and Linux environments (MacOS version should be available in a close future).


Publications

The methods using segmentation implemented in this software are explained in the following publications:

  • N. Peyrard. Approximation de type champ moyen des modèles de champ de Markov pour la segmentation de données spatiales. Phd Thesis, Université Joseph Fourier - Grenoble 1, 2001. [in French]
  • F. Forbes and N. Peyrard. Hidden Markov Random Field selection criteria based on Mean Field-like approximations. IEEE trans. on Pattern Analysis and Machine Intelligence, 25(8), 2003.
  • G.Celeux, F.Forbes and N.Peyrard. EM procedures using Mean Field-like approximations for Markov model-based image segmentation. Pattern Recognition, 36(1), pp. 131-144, 2003.
  • J. Blanchet. Modèles Markoviens et extensions pour la classification de données comples. Phd Thesis, Université Joseph Fourier - Grenoble 1, 2007. [in French]
  • J. Blanchet and F. Forbes, Triplet Markov fields for the supervised classification of complex structure data. IEEE PAMI, 30(6), pp. 1055-1067, 2008.
  • J. Blanchet, F. Forbes, S. Chopart and L. Azizi, Le logiciel SpaCEM3 pour la classification de donnĂ©es complexes. La Revue MODULAD, 40, pp.147-166, 2009. [in French]
  • J. Blanchet and M. Vignes, A model-based approach to gene clustering with missing observation reconstruction in a Markov Random Field framework. Journal of Computational Biology, 16(3), pp. 475-486, 2009.
  • M. Vignes and F. Forbes, Gene clustering via integrated Markov models combining individual and pairwise features. IEEE/ACM trans. on Computational Biology and Bioinformatics, 6(2), pp.260-270, 2009.

About the Team

SpaCEM3 is the result of several years of research ; first in the former IS2 project and then in the MISTIS project at INRIA Rhône-Alpes. It is also supported by members of SaAB team (Statistics and Algorithms for Biology) within BIA Unit at the INRA Toulouse.

Our research focus on developments of statistical methods to deal with complex systems and complex data. Our applications mainly consist mainly in image processing and spatial data problems with applications in Biology and Medicine. Our approach is based on the statement that complexity can be handled by working up from simple local assumptions in a coherent way, defining a structured model.This is the key to modelling, computation, inference and interpretation. The methods we consider involve mixture models, Markovian models, and more generally hidden structure models on one hand, and semi and non-parametric methods on the other.


Documentation

SpaCEM3 allows the user to perform supervised or unsupervised segmentation and simulation. The new graphical interface provides a user-friendly environment and allows the user to display data and results of segmentation or simulation. Documentation is available on the use of SpaCEM3: Documentation


Example Data sets

Some data sets are available for illustration, in particular those mentioned in the Documentation for unsupervised and supervised segmentation examples: Data sets. This file includes the data sets for the IEEE TCBB and Journal of Computational Biology papers mentioned above. Some other toy example datasets come with the sofware.
A presentation of the software was given at ModGraphII satellite of the JOBIM 2010 conference (Montpellier, France). Slides can be downloaded here.


License

SpaCEM3 is distributed under the licence CeciLL-B (http://cecill.info/ ).


Download

Please take some time to complete the form below before downloading the software; we won't make any commercial use of the details you provide. Should you wish to download SpaCEM3 directly, follow the links below the form.

Registration form for SpaCEM3 (not mandatory).
First Name:
Last Name:
Institution:
Country:
Email:
Use of SpaCEM3:
Type of system:


Contact

If you have questions or comments on the software (installation, use, included features), feel free to contact us at: SpaCEM3-help@lists.gforge.inria.fr



Last modification : November 2010

 

 


Powered By GForge Collaborative Development Environment