Probabilistic ICA to dissect modality-specific and amodal constituents of sensory ERPs A. Mouraux 1,3 , G.D. Iannetti 2 1 FMRIB Centre, 2 Department of Physiology, Anatomy and Genetics, University of Oxford (UK). 3 Unité READ, Université Catholique de Louvain (BE)
19
Embed
Probabilistic ICA to dissect modality-specific and amodal constituents of sensory ERPs
Probabilistic ICA to dissect modality-specific and amodal constituents of sensory ERPs. A. Mouraux 1,3 , G.D. Iannetti 2 1 FMRIB Centre, 2 Department of Physiology, Anatomy and Genetics, University of Oxford (UK). 3 Unité READ, Université Catholique de Louvain (BE). Tactile - PowerPoint PPT Presentation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Probabilistic ICA to dissect modality-specific and amodal constituents of sensory ERPs
A. Mouraux1,3, G.D. Iannetti2
1FMRIB Centre, 2Department of Physiology, Anatomy and Genetics, University of Oxford (UK).3Unité READ, Université Catholique de Louvain (BE)
What is the respective contribution of modality-specific and amodal brain activity to these recorded ERPs?
Do we have methods to address this question?
The signals recorded at sensors are modelled as a linear mixture of the source signals by an unknown mixing matrix.
Can this method be addressed using a blind source separation algorithm?
note that u ≠ s because of scaling and permutations
Blind source separation aims at finding an unmixing matrix that would recover the original source signals
Can this method be addressed using a blind source separation algorithm?
Probabilistic ICA avoids the problem of overfitting by constraining ICA to an objective estimate of the dimensionality of the data, obtained through Bayesian analysis.
In this waveform, amodal and modality-specific responses will have distinct time courses.Provided that amodal and modality-specific responses project differently onto the scalp sensors,
PICA should separate amodal and modality-specific responses in distinct ICs
For each subject, auditory, somatosensory and visual ERPs were concatenated into a single waveform
ICs contributing to the time courses of all four ERPs were categorized as amodal.ICs contributing to the time course of a specific ERP were categorized as modality-specific.
ICs contributing to the time course of a nociceptive and tactile somatosensory ERPs were categorized as somatosensory-specific.
On average, amodal activity explained : - 61% of the auditory ERP waveform- 66% of the non-nociceptive somatosensory ERP waveform- 76% of the nociceptive somatosensory ERP waveform- 55% of the visual ERP waveform
Amodal activity
Auditory-specific activity explained 32% of the auditory ERP waveform
Somatosensory-specific activity explained - 34% of the non-nociceptive somatosensory ERP waveform- 25% of the nociceptive somatosensory ERP waveform
Visual-specific activity explained 36% of the visual ERP waveform
Conclusion
Amodal brain responses represent the bulk of auditory, somatosensory and visual vertex potentials.
Modality-specific brain responses represent only a fraction of the early part of auditory, somatosensory and visual vertex potentials.
Probabilistic ICA can be used to separate sensory ERPs into its amodal and modality-specific constituents
Subtracting the contribution of amodal activity(activity contributing to all four ERP waveforms)
Subtracting the contribution of visual-specific activity(activity contributing uniquely to the visual ERP)
Subtracting the contribution of auditory-specific activity(activity contributing uniquely to the auditory ERP)
Subtracting the contribution of somatosensory-specific activity(activity contributing to both the non-nociceptive and nociceptive somatosensory ERP)
Subtracting the contribution of somatosensory-specific activity(activity contributing to uniquely to the non-nociceptive or the nociceptive somatosensory ERP)
Thank You!
Acknowledgements. We thank Drs Christian Beckmann, Léon Plaghki and Meng Liang for their insightful comments. André Mouraux is a Marie-Curie post-doctoral Research Fellow, and a “Chargé de recherches” of the Belgian National Fund for Scientific Research (FNRS). Giandomenico Iannetti is a University Research Fellow for The Royal Society.