Analyzing brain networks in language and social tasks using data-driven approaches
| dc.contributor.advisor | Aazhang, Behnaam | |
| dc.creator | Yellapantula, Sudha | |
| dc.date.accessioned | 2021-08-16T19:25:20Z | |
| dc.date.available | 2023-08-01T05:01:12Z | |
| dc.date.created | 2021-08 | |
| dc.date.issued | 2021-08-13 | |
| dc.date.submitted | August 2021 | |
| dc.date.updated | 2021-08-16T19:25:20Z | |
| dc.description.abstract | Humans are innately social, and we express ourselves primarily through language. We effortlessly articulate 2-3 words per second in fluent speech, yet this deceptively simple task is a highly complex multistage process in our brains. Unfortunately millions are affected by disease or brain-disorders leading to language and social dysfunction, with devastating consequences to their quality of life. The goal of this work was to improve our understanding of higher-order cognitive processes in the domain of language and social behavior, using a multitude of recording modalities and data-driven approaches. Three main questions were studied in this work - (1) Are language specific cognitive functions discretely computed within well-localized brain regions or rather by distributed networks? (2) When the brain receives ambiguous stimuli from the outside world, do more brain regions need to be involved to resolve this ambiguity, versus when the stimuli are unambiguous? (3) What is the effect of visual features on social cooperation, and its dependence on social context? These questions were studied using a variety of cognitive tasks and recording modalities - ECoG, EEG and spike recordings. To study the first question, we used intra-cranial electrocorticogram (ECoG) recordings from a picture naming task, to analyze the network phenomena of distributed cortical substrates supporting language. We estimated causality among brain regions with Directed Information, followed by a graph theoretic framework to extract task related dynamics from the causal estimates. Finally, we validated these functionally defined networks against the gold standard for causal inference - behavioral disruption with direct cortical stimulation. We demonstrate that the network measures combined with power have greater predictive capability for identifying critical language regions than discrete, regional power analyses alone. For the second question, we quantified the ambiguity in speech perception using network phenomena of distributed cortical substrates supporting language. We estimated statistical dependence among brain regions with Mutual Information. Using innovative baseline normalization, time-varying graphs were derived from the EEG data. These measures were tested in healthy subjects by comparing network measures derived from both ambiguous and clear stimuli. Finally, to further validate the hypothesis, we also evaluated the brain networks of an aphasic subject, who perceived all stimuli as ambiguous. We demonstrate that the network measures could clearly distinguish the aphasic patient's processing from the healthy subjects, providing evidence to support the increased brain activity to process more ambiguous stimuli. This allows for better understanding of the cognitive processes to measure patient impairment. For the third question, multi-unit recordings from freely moving non-human primates were obtained, while they performed a social cooperative task. They were equipped with wireless recording system, and scene and eye cameras to capture the field of view. We identified fixations, and the objects within receptive fields during the fixations. We tested the classification accuracy of the neural data to distinguish visual features identified during fixations, and find the social learning is captured by the improvement in distinguishing socially relevant objects with time. We test these effects in two brain regions - dorsolateral pre-frontal cortex and the higher visual area V4, and test the effects of attention on the task, lack of attention and effect of the social behavior of the partner monkey. Understanding the dynamics underlying these higher order language and social tasks can advance our understanding of cognitive disorders that can occur due to traumatic injury or disease, and could yield better remedial treatments or therapies in the future. | |
| dc.embargo.terms | 2023-08-01 | |
| dc.format.mimetype | application/pdf | |
| dc.identifier.citation | Yellapantula, Sudha. "Analyzing brain networks in language and social tasks using data-driven approaches." (2021) Diss., Rice University. <a href="https://hdl.handle.net/1911/111198">https://hdl.handle.net/1911/111198</a>. | |
| dc.identifier.uri | https://hdl.handle.net/1911/111198 | |
| dc.language.iso | eng | |
| dc.rights | Copyright is held by the author, unless otherwise indicated. Permission to reuse, publish, or reproduce the work beyond the bounds of fair use or other exemptions to copyright law must be obtained from the copyright holder. | |
| dc.subject | language | |
| dc.subject | speech perception | |
| dc.subject | ECoG | |
| dc.subject | EEG | |
| dc.subject | information theory | |
| dc.subject | directed information | |
| dc.subject | mutual information | |
| dc.subject | graph theory | |
| dc.subject | social cooperation | |
| dc.subject | neural encoding | |
| dc.subject | non-human primates | |
| dc.subject | freely moving | |
| dc.title | Analyzing brain networks in language and social tasks using data-driven approaches | |
| dc.type | Thesis | |
| dc.type.material | Text | |
| thesis.degree.department | Electrical and Computer Engineering | |
| thesis.degree.discipline | Engineering | |
| thesis.degree.grantor | Rice University | |
| thesis.degree.level | Doctoral | |
| thesis.degree.major | Computational neuroscience | |
| thesis.degree.name | Doctor of Philosophy |
Files
Original bundle
1 - 1 of 1
Loading...
- Name:
- YELLAPANTULA-DOCUMENT-2021.pdf
- Size:
- 44.88 MB
- Format:
- Adobe Portable Document Format