Event
Ph.D. Dissertation Defense: Vrishab Commuri
Monday, July 27, 2026
11:00 a.m.
AVW1146
Emily Irwin
301 405 0680
eirwin@umd.edu
ANNOUNCEMENT: Ph.D. Dissertation Defense
Name: Vrishab Commuri
Committee:
Professor Jonathan Z. Simon (Chair)
Professor Behtash Babadi
Professor Shihab Shamma
Professor Stefanie E. Kuchinsky
Professor Samira Anderson (Dean's Representative)
Date/Time: Monday, July 27th, 11:00 AM
Location: AVW 1146
Title: Band-Specific Analysis of Magnetoencephalography Data: Application to Clinical and Auditory Experimentation
Abstract:
Neural activity in the brain can be divided into several frequency bands: delta (1-4 Hz),
theta (4-8 Hz), alpha (8-12 Hz), beta (12-25 Hz), and gamma (25+ Hz). Historically, the delin-
eation of these bands derives from measurements within each that were observed to have charac-
teristic distributions over sensors placed on the scalp. Changes in signal power within each band
were also observed to correlate with behavioral states or tasks. Taken together, these findings
intimate the presence of distinct networks, specific in location and function, within the brain.
However, localizing these band-specific responses using non-invasive recording methods is still
a nascent area of research.
In this thesis, we conduct several band-specific analyses by combining two complemen-
tary approaches: Network Localized Granger Causality (NLGC), which characterizes directed
interactions between brain regions, and Temporal Response Functions (TRFs), which link neu-
ral activity to external stimuli. Together, these methods allow us to investigate both endogenous
network dynamics and stimulus-driven (exogenous) activity across multiple frequency bands.
We begin with an investigation of high-gamma (75-200 Hz) responses obtained from mag-
netoencephalography (MEG) recordings. These responses time-lock to (synchronize with) fea-
tures of an auditory speech stimulus. It is known that such responses are elicited in primary
auditory cortex, but it is unknown whether they are modulated by selective attention. We utilize
the TRF framework to show that these responses are indeed modulated by selective attention,
and that this attentional modulation happens with very low latency – earlier in the auditory pro-
cessing pathway than previously thought. Our findings indicate top-down interactions as early
as primary auditory cortex. Identifying which areas are modulated in the high-gamma range
represents an important step toward mapping attentional control networks involved in auditory
processing.
Next, we analyze MEG beta-band resting-state networks in a clinical population of minor
ischemic stroke patients. The beta band is strongly associated with motor-related neural activity,
which is known to be disrupted after stroke. In these individuals, cognitive slowing is observed
alongside encumbered motor planning and execution, suggesting diffuse network reorganization
involving areas that are linked to cognition. This cognitive degradation is observed irrespective
of vascular territory affected by the lesion, indicating nonfocal and distributed effects of minor
stroke. We introduce a graph-based statistical method for contrasting NLGC-derived connectivity
maps between cohorts. We apply this method to link network-level changes to behavioral out-
comes across multiple follow-up visits. We also characterize networks associated with favorable
outcomes across visits.
Finally, we apply network dynamic and stimulus-driven analysis methods in tandem to an
MEG data set recorded from older and younger listeners. Marked changes between older and
younger listeners have been previously observed in the delta and theta bands of the same data
using a temporal response function paradigm. We apply NLGC to the same recordings and per-
form a similar network analysis to contrast older and younger subjects’ listening networks. We
also fit TRFs to the NLGC source estimates to replicate the results from prior work. We eluci-
date how cortical circuits change as listening conditions become increasingly adverse, and we
reveal differences in regional connectivity between older and younger individuals. We provide
evidence for distributed network changes with aging, particular to the delta and theta bands, and
demonstrate the sensitivity of these network changes to listening conditions. Our findings high-
light age- and condition-dependent shifts in speech-related cortical connectivity across delta and
theta frequency bands. By integrating NLGC-derived networks with TRF analysis, we localize
hierarchical speech processing within these circuits.
This thesis proposes and validates network dynamic and stimulus-driven approaches for
analyzing neural activity across diverse datasets, spanning basic auditory processing, clinical
populations, and aging. It is shown how these analyses can be conducted in tandem to obtain
similar results to prior TRF-only studies. Additionally, this thesis makes methodological contri-
butions to the analysis of any sparse functional connectivity data, with broad applications across
clinical and cognitive neuroscience.
Name: Vrishab Commuri
Committee:
Professor Jonathan Z. Simon (Chair)
Professor Behtash Babadi
Professor Shihab Shamma
Professor Stefanie E. Kuchinsky
Professor Samira Anderson (Dean's Representative)
Date/Time: Monday, July 27th, 11:00 AM
Location: AVW 1146
Title: Band-Specific Analysis of Magnetoencephalography Data: Application to Clinical and Auditory Experimentation
Abstract:
Neural activity in the brain can be divided into several frequency bands: delta (1-4 Hz),
theta (4-8 Hz), alpha (8-12 Hz), beta (12-25 Hz), and gamma (25+ Hz). Historically, the delin-
eation of these bands derives from measurements within each that were observed to have charac-
teristic distributions over sensors placed on the scalp. Changes in signal power within each band
were also observed to correlate with behavioral states or tasks. Taken together, these findings
intimate the presence of distinct networks, specific in location and function, within the brain.
However, localizing these band-specific responses using non-invasive recording methods is still
a nascent area of research.
In this thesis, we conduct several band-specific analyses by combining two complemen-
tary approaches: Network Localized Granger Causality (NLGC), which characterizes directed
interactions between brain regions, and Temporal Response Functions (TRFs), which link neu-
ral activity to external stimuli. Together, these methods allow us to investigate both endogenous
network dynamics and stimulus-driven (exogenous) activity across multiple frequency bands.
We begin with an investigation of high-gamma (75-200 Hz) responses obtained from mag-
netoencephalography (MEG) recordings. These responses time-lock to (synchronize with) fea-
tures of an auditory speech stimulus. It is known that such responses are elicited in primary
auditory cortex, but it is unknown whether they are modulated by selective attention. We utilize
the TRF framework to show that these responses are indeed modulated by selective attention,
and that this attentional modulation happens with very low latency – earlier in the auditory pro-
cessing pathway than previously thought. Our findings indicate top-down interactions as early
as primary auditory cortex. Identifying which areas are modulated in the high-gamma range
represents an important step toward mapping attentional control networks involved in auditory
processing.
Next, we analyze MEG beta-band resting-state networks in a clinical population of minor
ischemic stroke patients. The beta band is strongly associated with motor-related neural activity,
which is known to be disrupted after stroke. In these individuals, cognitive slowing is observed
alongside encumbered motor planning and execution, suggesting diffuse network reorganization
involving areas that are linked to cognition. This cognitive degradation is observed irrespective
of vascular territory affected by the lesion, indicating nonfocal and distributed effects of minor
stroke. We introduce a graph-based statistical method for contrasting NLGC-derived connectivity
maps between cohorts. We apply this method to link network-level changes to behavioral out-
comes across multiple follow-up visits. We also characterize networks associated with favorable
outcomes across visits.
Finally, we apply network dynamic and stimulus-driven analysis methods in tandem to an
MEG data set recorded from older and younger listeners. Marked changes between older and
younger listeners have been previously observed in the delta and theta bands of the same data
using a temporal response function paradigm. We apply NLGC to the same recordings and per-
form a similar network analysis to contrast older and younger subjects’ listening networks. We
also fit TRFs to the NLGC source estimates to replicate the results from prior work. We eluci-
date how cortical circuits change as listening conditions become increasingly adverse, and we
reveal differences in regional connectivity between older and younger individuals. We provide
evidence for distributed network changes with aging, particular to the delta and theta bands, and
demonstrate the sensitivity of these network changes to listening conditions. Our findings high-
light age- and condition-dependent shifts in speech-related cortical connectivity across delta and
theta frequency bands. By integrating NLGC-derived networks with TRF analysis, we localize
hierarchical speech processing within these circuits.
This thesis proposes and validates network dynamic and stimulus-driven approaches for
analyzing neural activity across diverse datasets, spanning basic auditory processing, clinical
populations, and aging. It is shown how these analyses can be conducted in tandem to obtain
similar results to prior TRF-only studies. Additionally, this thesis makes methodological contri-
butions to the analysis of any sparse functional connectivity data, with broad applications across
clinical and cognitive neuroscience.
