Diferencia entre revisiones de «GIC-experimental-databases/Cocaine feature vectors»
De Grupo de Inteligencia Computacional (GIC)
| (No se muestran 11 ediciones intermedias de 3 usuarios) | |||
| Línea 1: | Línea 1: | ||
== Datasets of features extracted from a database of 124 subjects: 61 healthy controls and 63 cocaine adicted patients == | == Datasets of features extracted from a database of 124 subjects: 61 healthy controls and 63 cocaine adicted patients == | ||
These features are based on two feature extraction pipelines: one based on Pearson's correlation and other based on voxel based morphometry (VBM) methodology. For the moment, we work over the GM voxel intensity values after image preprocessing. We apply different | These features are based on two feature extraction pipelines: one based on Pearson's correlation and other based on voxel based morphometry (VBM) methodology. For the moment, we work over the GM voxel intensity values after image preprocessing. We apply different FWHM Gaussian kernels (sigma) to the data: sigma=0; sigma=3; sigma=6; sigma=9; sigma=12. We also apply different thresholds to classify with different number of features. | ||
Every sigma (for Pearson correlation or VBM feature extraction) is accompanied by different thresholds to select a different number of features. To avoid circularity we use 10 fold cross validation: there are 10 folds for every number of features. | |||
If you use these databases for any publication, please refer to the following papers: | |||
;Maite Termenon, Manuel Grana, Barros-Loscertales Alfonso and Cesar Avila | |||
:"Extreme Learning Machines for feature selection and classification of cocaine dependent patients on structural " , | |||
:Neural Processing Letters (online) | |||
;M. Termenon, D. Chyzhyk, M. Graña, A. Barros-Loscertales, and C. Avila | |||
:Cocaine Dependent Classification on MRI Data Extracting Features from Voxel Based Morphometry | |||
:J.M. Ferrandez et al. (Eds.): IWINAC 2013, Part II, LNCS 7931, pp. 140--148. Springer, Heidelberg (2013) | |||
;M. Termenon, E. Fernández, A. Barrós-Loscertales, J.C. Bustamante, C. Ávila | |||
:Impact of Analysis Circularity: a Case Study in Cocaine Addiction detection on MRI | |||
:Advances in Knowledge-Based and Intelligent Information and Engineering Systems. | |||
:Frontiers in Artificial Intelligence and Applications (FAIA) series, Vol. 243, pages 2201 - 2209, 2012. | |||
:Eds: Manuel Graña, Carlos Toro, Jorge Posada, Robert J. Howlett and Lakhmi C. Jain. | |||
<h3> | <h3> | ||
Feature sets extracted applying Pearson's correlation | Feature sets extracted applying Pearson's correlation: | ||
</h3> | </h3> | ||
* | * sigma=0 - [[media:FeatVect-Pearson-sm0-10crossvalid.zip| download]] | ||
* | * sigma=3 - [[media:FeatVect-Pearson-sm3-10crossvalid.zip| download]] | ||
* | * sigma=6 - [[media:FeatVect-Pearson-sm6-10crossvalid.zip| download]] | ||
* | * sigma=9 - [[media:FeatVect-Pearson-sm9-10crossvalid.zip| download]] | ||
* | * sigma=12 - [[media:FeatVect-Pearson-sm12-10crossvalid.zip| download]] | ||
<h3> | <h3> | ||
Feature sets extracted applying VBM | Feature sets extracted applying VBM: | ||
</h3> | </h3> | ||
* | * sigma=0 - [[media:FeatVect-VBM-sm0-10crossvalid.zip| download]] | ||
* | * sigma=3 - [[media:FeatVect-VBM-sm3-10crossvalid.zip| download]] | ||
* | * sigma=6 - [[media:FeatVect-VBM-sm6-10crossvalid.zip| download]] | ||
* | * sigma=9 - [[media:FeatVect-VBM-sm9-10crossvalid.zip| download]] | ||
* | * sigma=12 - [[media:FeatVect-VBM-sm12-10crossvalid.zip| download]] | ||
Revisión actual - 15:49 14 oct 2013
Datasets of features extracted from a database of 124 subjects: 61 healthy controls and 63 cocaine adicted patients
These features are based on two feature extraction pipelines: one based on Pearson's correlation and other based on voxel based morphometry (VBM) methodology. For the moment, we work over the GM voxel intensity values after image preprocessing. We apply different FWHM Gaussian kernels (sigma) to the data: sigma=0; sigma=3; sigma=6; sigma=9; sigma=12. We also apply different thresholds to classify with different number of features. Every sigma (for Pearson correlation or VBM feature extraction) is accompanied by different thresholds to select a different number of features. To avoid circularity we use 10 fold cross validation: there are 10 folds for every number of features.
If you use these databases for any publication, please refer to the following papers:
- Maite Termenon, Manuel Grana, Barros-Loscertales Alfonso and Cesar Avila
- "Extreme Learning Machines for feature selection and classification of cocaine dependent patients on structural " ,
- Neural Processing Letters (online)
- M. Termenon, D. Chyzhyk, M. Graña, A. Barros-Loscertales, and C. Avila
- Cocaine Dependent Classification on MRI Data Extracting Features from Voxel Based Morphometry
- J.M. Ferrandez et al. (Eds.): IWINAC 2013, Part II, LNCS 7931, pp. 140--148. Springer, Heidelberg (2013)
- M. Termenon, E. Fernández, A. Barrós-Loscertales, J.C. Bustamante, C. Ávila
- Impact of Analysis Circularity: a Case Study in Cocaine Addiction detection on MRI
- Advances in Knowledge-Based and Intelligent Information and Engineering Systems.
- Frontiers in Artificial Intelligence and Applications (FAIA) series, Vol. 243, pages 2201 - 2209, 2012.
- Eds: Manuel Graña, Carlos Toro, Jorge Posada, Robert J. Howlett and Lakhmi C. Jain.
