Electroencephalogram (EEG) assessments noninvasively detect electrical activity using electrodes easily attached to one’s scalp, providing a rich neural signal that captures many aggregate neural phenomena. While its ease of use makes it an important diagnostic tool, EEG data lacks cellular resolution. As such, it has been difficult to identify the specific neural features that underlie EEG abnormalities observed in, and potentially contributing to, the severity and progression of autism spectrum disorders (ASDs).
In the current project, Stefano Panzeri proposes to address this fundamental problem by developing rigorous mathematical methods to analyze EEG data and directly interpret it in terms of neural phenomena. This proposal will focus first on the development of biophysically plausible models of neural networks that mathematically explain the relationship between EEG signatures and the underlying neural processes. These models will then be turned into EEG-analysis tools by fitting the models to the EEG data and then translating the EEG data back into an estimate of important neural parameters, such as the spike rate of excitatory and inhibitory neurons and the strength of their connections. These neural parameters will be inferred as the model’s parameter values that best explain the EEG data.
These methods will be validated using EEG and neural activity simultaneously recorded from the visual cortex in a mouse model of Rett syndrome (RTT), a neurodevelopmental disorder caused by de novo mutations in MECP2. By developing these methods in a mouse model, the mathematical model’s accuracy can be directly tested by comparing the neural parameters and their RTT-related longitudinal variations, obtained from intracranial recordings to those estimated from surface EEG data. Once the methodology has been validated in a mouse model, further validation will be undertaken in individuals with RTT by estimating changes in neural parameters taken from scalp EEG recordings in these individuals and age-matched controls.
This work will generate and validate a set of solid and highly credible mathematical tools that can be used to predict and then interpret, in terms of underlying neural processes, EEG features collected in individuals with RTT and related neurodevelopmental disorders. This will ultimately allow a deeper understanding of how genetic risk variants for ASDs affect neural processes.