Dynamic causal modelling of retino-cortical connectivity

TY - CONF

T1 - Dynamic causal modelling of retino-cortical connectivity

AU - Dietz, Martin

AU - Bailey, Christopher

AU - Dalal, Sarang S.

PY - 2018/8/27

Y1 - 2018/8/27

N2 - BackgroundThere is a rich literature on the primate visual system describing the anatomy and physiology of the retina and its projections to visual cortex via the lateral geniculate nucleus (LGN). While nearly all research in humans has focused on visual cortical regions using both invasive and non-invasive electrophysiology, surprisingly little is known about the relative timing and spectral behaviour of the human retinal network at the earliest stages of representing the visual world. This is important when we want to characterize how synaptic responses in the retina drive responses in primary visual cortex, via the LGN. More specifically, we want to identify whether this feedforward connectivity is realized in a frequency-specific manner or through cross-frequency coupling.MethodsIn order to address this question, we use a combination of non-invasive electro-retinography (ERG) of the retina and magnetoencephalography (MEG) of the visual cortex in a pilot dataset from a healthy human adult, presented with full visual field 1 ms light flashes at a 1 sec inter-trial interval (200 trials). Following source analysis of fused magnetometers and planar gradiometers (Elekta Neuromag Triux) using an Empirical Bayesian beamformer implemented in SPM12, the ERG and MEG source space data were analysed using Dynamic causal modelling (DCM) of induced spectral responses. This models the time-varying power as a distributed dynamical system of coupled regions that drive one another in a frequency-specific manner or through cross-frequency coupling. ResultsConventional time-frequency analysis of the retina and V1 time-series revealed transient high gamma (110-200 Hz) activity in the retina peaking at 20 ms post-stimulus time, accompanied by slower responses in the beta range (15-35 Hz). These retinal responses were followed by transient gamma (110-130 Hz) activity in V1 peaking at 105 ms post-stimulus time, again accompanied by slower responses in the beta range (15-35 Hz) and followed by gamma activity in the 40-75 Hz range and 85-95 Hz range at 200 ms post-stimulus time. Finally, DCM of the retina and V1 showed evidence that the beta activity in the retina drives beta in V1, while the characteristic high gamma activity in the retina drives gamma in V1 in the both the higher (110-130 Hz) and lower (40-75 Hz) frequencies.ConclusionRecordings from the human retina in response to particular visual contexts offer a unique window into the earliest level of synaptic signalling within the retinal microcircuit following phototransduction. Combining these early responses with electromagnetic recordings from primary visual cortex, the earliest cortical region observable with MEG, we were able to show that feedforward projections (via the LGN) are likely to operate in both a frequency-specific manner, as well as through cross-frequency coupling. Moreover, we were able to obtain an plausible estimate the conduction delay between retina and V1 using DCM for CSD. Our immediate future research will aim at including the LGN and pulvinar to model the subcortical pathway and characterize the role of feedback connections in visual receptive field modulation. This will be important in characterizing diseases that affect both early and late connectivity in the visual system, including optic neuritis and diabetic retinopathy, as well as blindsight and neglect syndromes.

AB - BackgroundThere is a rich literature on the primate visual system describing the anatomy and physiology of the retina and its projections to visual cortex via the lateral geniculate nucleus (LGN). While nearly all research in humans has focused on visual cortical regions using both invasive and non-invasive electrophysiology, surprisingly little is known about the relative timing and spectral behaviour of the human retinal network at the earliest stages of representing the visual world. This is important when we want to characterize how synaptic responses in the retina drive responses in primary visual cortex, via the LGN. More specifically, we want to identify whether this feedforward connectivity is realized in a frequency-specific manner or through cross-frequency coupling.MethodsIn order to address this question, we use a combination of non-invasive electro-retinography (ERG) of the retina and magnetoencephalography (MEG) of the visual cortex in a pilot dataset from a healthy human adult, presented with full visual field 1 ms light flashes at a 1 sec inter-trial interval (200 trials). Following source analysis of fused magnetometers and planar gradiometers (Elekta Neuromag Triux) using an Empirical Bayesian beamformer implemented in SPM12, the ERG and MEG source space data were analysed using Dynamic causal modelling (DCM) of induced spectral responses. This models the time-varying power as a distributed dynamical system of coupled regions that drive one another in a frequency-specific manner or through cross-frequency coupling. ResultsConventional time-frequency analysis of the retina and V1 time-series revealed transient high gamma (110-200 Hz) activity in the retina peaking at 20 ms post-stimulus time, accompanied by slower responses in the beta range (15-35 Hz). These retinal responses were followed by transient gamma (110-130 Hz) activity in V1 peaking at 105 ms post-stimulus time, again accompanied by slower responses in the beta range (15-35 Hz) and followed by gamma activity in the 40-75 Hz range and 85-95 Hz range at 200 ms post-stimulus time. Finally, DCM of the retina and V1 showed evidence that the beta activity in the retina drives beta in V1, while the characteristic high gamma activity in the retina drives gamma in V1 in the both the higher (110-130 Hz) and lower (40-75 Hz) frequencies.ConclusionRecordings from the human retina in response to particular visual contexts offer a unique window into the earliest level of synaptic signalling within the retinal microcircuit following phototransduction. Combining these early responses with electromagnetic recordings from primary visual cortex, the earliest cortical region observable with MEG, we were able to show that feedforward projections (via the LGN) are likely to operate in both a frequency-specific manner, as well as through cross-frequency coupling. Moreover, we were able to obtain an plausible estimate the conduction delay between retina and V1 using DCM for CSD. Our immediate future research will aim at including the LGN and pulvinar to model the subcortical pathway and characterize the role of feedback connections in visual receptive field modulation. This will be important in characterizing diseases that affect both early and late connectivity in the visual system, including optic neuritis and diabetic retinopathy, as well as blindsight and neglect syndromes.

KW - ERG

KW - MEG

KW - DCM

KW - Vision

M3 - Poster

ER -