Seminar – Functional neuroimaging data characterisation via tensor representations
When: Wednesday December 5, 14h-15h
Location: Room B00.35, Department of Electrical Engineering, Kasteelpark Arenberg 10, 3001 Leuven
Speaker: Christos Chatzichristos, National and Kapodistrian University of Athens, Greece
Title: Functional neuroimaging data characterisation via tensor representations.
Abstract:
Brain tasks involving action, perception, cognition, etc., are performed via the simultaneous activation of a number of brain regions, which are engaged in proper interactions in order to effectively execute the task. In functional Magnetic Resonance Imaging (fMRI), brain activity is captured by detecting associated changes in blood flow within the brain. The obtained data stream comprises a mixture of the source signals which carry the valuable information required by the neuroscientists in understanding the brain functions. Extracting information from fMRI data commonly relies on simplifying assumptions and is mainly based on matrix-based approaches that fail to exploit the inherent multi-way structure of brain data. The main topic of my research is the investigation of tensor (multi-way) models and associated algorithms that are adapted to the fMRI problem structure and assess their performance as compared to matrix-based schemes.
In the first part of the presentation, we will focus on the possible gains from exploiting the 4-dimensional nature of the brain images, through a higher-order tensorization of the fMRI signal, and the use of less restrictive generative models. In this context, the higher-order Block Term Decomposition (BTD) and the PARAFAC2 tensor models are considered for the first time in fMRI blind source separation. A novel PARAFAC2-like extension of BTD (BTD2) is also proposed, aiming at combining the effectiveness of BTD in handling strong instances of noise and the potential of PARAFAC2 to cope with datasets that do not follow the strict multilinear assumption.
On the latter part of the presentation, the problem of jointly analyzing electroencephalography (EEG) and functional Magnetic Resonance Imaging (fMRI) data will be considered. Analyzing both EEG and fMRI measurements is highly beneficial for studying brain function because these modalities have complementary spatiotemporal resolutions: EEG offers good temporal resolution while fMRI offers a good spatial resolution. The fusion methods reported so far ignore the underlying multi-way nature of the data in at least one of the modalities and/or rely on very strong assumptions concerning the relationship among the respective data sets. As a first attempt to address those two points, the use of tensor models for both modalities and a soft coupling approach, has been introduced as a solution to the fusion problem. To cope with the subject variability in EEG, the PARAFAC2 model is adopted. The main disadvantage of the method proposed is that in multisubject case the multinearity is a strict assumption. A more flexible model which will allow differences per subject shall be sought.