Difference between revisions of "Machine learning based multimodal classification for Disorder of Consciousness"

Latest revision as of 09:17, 9 September 2013

[edit] People

• Francisco Gómez

[edit] Summary

Multimodal information has been proved to be a powerful tool to study brain diseases. Very challenging conditions such as disorder of consciousness (DOC) has been greatly benefited of the multidimensional perspective of these measurements. For instance, by measuring different information related with metabolic activity (PET), white and gray matter structural preservation (DTI and MRI) and brain dynamics (fMRI), an interesting set of biomarkers related with the necessary conditions to observe brain activity in these conditions has been proposed. In other hand, machine learning based analysis has also attracted the attention of the neuroimages community. This tool allows the study of multivariate patterns of occurrence in these signals. In general, machine learning analysis of neuroimages has been performed at single modality level. In this work we propose to use machine learning tools to study the multivariate relationships across multiple modalities related with the level of consciousness in DOC patients. By using this approach we expect two main advances: 1) improving our understanding of the relationships between modalities, in particular, commonalities and particularities and 2) improving the characterization capacity in these diseases by exploiting weighted information coming from different modalities.

[edit] Method overview

[edit] Data sources

[edit] Results (Expected)

• Method conference. Model proposal.
• Journal article. Complete model and application to clinical data.

[edit] References

"Relevance vector machine" consciousness classifier applied to cerebral metabolism of vegetative and locked-in patients. Phillips CL, Bruno MA, Maquet P, Boly M, Noirhomme Q, Schnakers C, Vanhaudenhuyse A, Bonjean M, Hustinx R, Moonen G, Luxen A, Laureys S, NeuroImage, 2011