Ver el código de «GIC-experimental-databases/OASIS deformation feature vectors»

== Datasets of features extracted from the subset of 98 females from OASIS  ==

These features are based on deformation measures (displacement vector magnitudes and Jacobian determinant of gradient matrices) of a custom template made with all the 98 subjects registered to each subject.
Some feature sets doesn't exist because the correlation values were lower than the percentile limit.
<br/>

Reference paper with results on these datasets


:Alexandre Savio, Manuel Graña, Jorge Villanúa
:[[media:ASavio-HAIS-2011.pdf|Deformation based features for Alzheimer's disease detection with linear SVM]]
:Hybrid Artificial Intelligence Systems, 6th International Conference (HAIS 2011) - HAIS 2011, Part II, LNAI 6679 proceedings, p.336-343. Springer, Heidelberg (2011)

<h3>
Pipelines trying to explain how these features were extracted:
</h3>
* [[media:Registration_deformation_calculation.png | Obtaining the measures of the displacement vectors. ]]
* [[media:Correlation_measure_calculation.png | Obtaining the correlation values from the displacement measures. ]]

<h3>
Feature sets extracted from transformation displacement magnitudes (DM)
</h3>

* [[media:DM_prctile990.zip | Feature sets of Pearson, Spearman and Kendall correlation measures over a 0.990 percentile]]
* [[media:DM_prctile995.zip | Feature sets of Pearson, Spearman and Kendall correlation measures over a 0.995 percentile]]
* [[media:DM_prctile999.zip | Feature sets of Spearman and Kendall correlation measures over a 0.999 percentile]]

<h3>
Feature sets extracted from transformation gradient Jacobian matrices determinant (JD)
</h3>

* [[media:JD_prctile990.zip | Feature sets of Pearson, Spearman and Kendall correlation measures over a 0.990 percentile]]
* [[media:JD_prctile995.zip | Feature sets of Spearman and Kendall correlation measures over a 0.995 percentile]]
* [[media:JD_prctile999.zip | Feature sets of Spearman and Kendall correlation measures over a 0.999 percentile]]

Contact: Alexandre Savio.