El Rassi, Yara (2024) The Representation of Visual Naturalistic Stimuli in Resting State Activity: An Investigation in the Visual and Motor Areas Representations at Rest. Advisor: Ricciardi, Prof. Emiliano. Coadvisor: Betti, Prof. Viviana . pp. 108. [IMT PhD Thesis]
Text (Doctoral thesis)
thesis_reviewed_elrassi.pdf - Published Version Available under License Creative Commons Attribution Non-commercial Share Alike. Download (3MB) |
Abstract
Resting state is characterized as an offline period, during which the eyes may be either open or closed. In this disengaged state, one’s system operates independently of external input or feedback, and by definition, relies on an internalized model of the world. Literature shows that resting state activity may reflect the statistics of the natural environment, but also the unique individual biases, and is possible to be reshaped over time. This is highlighted by studies that show that resting state fluctuations maintain traces of everyday activity; but how are these representations extracted, how stable are they, and to what extent are they malleable? To answer that we need to understand: 1) How is our system structured to maintain regularities? 2) How are they integrated in an internalized model? 3) How do low frequencies fluctuate when detecting an error? The main aim of this thesis is to understand how natural information is represented in resting state. The working model is that (1) naturalistic information is processed along a hierarchy in time and space to code higher level information that is low dimensional and sparse (chapter 2). 2) This information is then maintained in resting state in a generic form (chapter 3). 3) This is achieved because low frequency fluctuations are adapted to naturalistic statics, and hence are altered in otherwise unexpected situations (chapter 4). We examined the functional connectivity (FC) of MEG signal changes in the visual (VIS) and dorsal attention (DAN) networks during the observation of naturalistic videos, by comparing them to a pretrained convolutional network. We reveal distinct temporal dynamics in processing low and high-level features. Low-level features are immediately and abundantly represented, while high-level features exhibit a delayed and scarce representation, potentially storing information in a generic form (chapter 2). For instance, we find that the BOLD multivoxel spatial representation of a still hand, controlled for low-level features, is coherent with the spatial representation of the resting somatomotor area, as opposed to another object such as a food item (chapter 3). We suggest that the representations during resting states may contribute to the goal of interacting with the environment. This is enriched by our final findings; the multivoxel spatial representation of observing common movements aligns more coherently with resting somatomotor patterns as opposed to uncommon (chapter 4).
Item Type: | IMT PhD Thesis |
---|---|
Subjects: | R Medicine > RC Internal medicine |
PhD Course: | Cognitive, Computational and Social Neurosciences |
Identification Number: | https://doi.org/10.13118/imtlucca/e-theses/412 |
NBN Number: | [error in script] |
Date Deposited: | 13 May 2024 07:59 |
URI: | http://e-theses.imtlucca.it/id/eprint/412 |
Actions (login required, only for staff repository)
View Item |