Guests and speakers will arrive at room B200 in the University Building. Tea/coffee will be available.

A few words before we begin.

Molecular probes for fMRI-based analysis of integrated brain function Understanding the neural basis of behavior and cognition requires determining how distinct processing elements interact to carry out brain function at an integrated level. In this talk, I will describe our laboratory’s efforts to address this goal using a combination of molecular sensors with noninvasive wide-field neuroimaging. To introduce our approach, I will give an overview of my lab’s work on functional and molecular MRI in animals. I will then discuss in detail a new approach designed to probe the engagement of genetically targeted neural circuit elements in brain processing. Finally, I will describe a form of ultrasensitive MRI sensor that may enable translational studies of neurochemical dynamics in the future. We expect that the experimental paradigms presented here could contribute “ground truth” data about functional connectivity and help inform mechanistic models of information flow throughout the brain.

Large-scale network dynamics and transient spectral events Functional brain networks display complex spatiotemporal dynamics that span multiple time-scales. In this talk, I will present computational methods that can access the dynamics of large-scale networks at the sub-second time-scales associated with the fastest of cognition. I will show how these approaches can be used in combination with MEG data to: a) infer fast dynamics of spectrally distinct large-scale phase locking networks in rest and task, b) infer transient spectral events (e.g. beta bursts) and link them to large-scale brain networks, c) provide a link between the spontaneous replay of learnt sequences and the spontaneous activity of resting state networks. Finally, I will finish with a look at a deep learning based approach, Dynemo, which we are developing as a new alternative to the HMM.

We will take a short break between speakers.

Changing Connectomes: Using network analysis and computer simulations to inform interventions for mental and brain health conditions Our work on connectomics over the last 20 years has shown a small-world, modular, and hub architecture of brain networks . Small-world features enable the brain to rapidly integrate and bind information while the modular architecture, present at different hierarchical levels, allows separate processing of various kinds of information (e.g. visual or auditory) while preventing wide-scale spreading of activation. Hub nodes play critical roles in information processing and are involved in many brain diseases. As neural systems are hierarchical networks, a modular architecture is visible at different levels of organisations, both for the macro-connectome of fibre tract connections between brain regions as well as the meso-connectome of fibre connections within brain regions. Changes in connectome organisation can be inform interventions for brain disorders. For epilepsy, for example, functional connectivity can predict whether surgery in patients will be successful. Using structural connectivity, network changes within brain regions increase with epilepsy disease duration and with the severity of seizures. Importantly, local information about connectivity within regions is a better predictor of surgery outcome than the standard approach of observing connectivity between regions [6]. Looking at clinical applications, I will finally outline how network analysis and computer simulations can inform how to change brain connectivity and dynamics through non-invasive focused ultrasound neuromodulation.

We will take a lunch break before our next speaker.

Freeform Stimulator and a path to more versatile neural implants. Conventional neural implants (a.k.a. Implantable Pulse Generators (IPGs)) use charge balanced biphasic pulses to evoke action potentials (APs) in target neurons. They are constrained to delivering these short pulses because otherwise violation of charge injection criteria at the metal-tissue interface will result in bubbles due to electrolysis, pH changes, and potentially toxic electrochemical biproducts. However, it has been shown in many previous experiments that delivery of long duration waveforms, including direct current, can significantly expand the range of neural control compared to short pulses. It can cause neurons to not only excite their activity in a natural stochastic fashion in contrast to phased-locked pulse-evoked APs, but also reduce spontaneous firing rate or inhibit it altogether, sensitize neurons to synaptic input, and even change synaptic weights by enhancing or reducing long-term potentiation (LTP). A neural implant that could deliver long duration non-pulsatile waveforms safely could therefore be beneficial to many applications, in which this wide range of neural control is desirable. I will present the Freeform Stimulator (FS), a novel implant that can deliver ultra-low frequency and direct electric fields to neurons safely and describe its principle of operation. I will also present current experimental results for using FS for suppressing chronic peripheral pain and as a method to triple the effectiveness of the vestibular prosthesis. Additionally, I will address the fundamental theoretical differences between neural modulation delivered by FS vs. an IPG. Finally, I will describe ongoing and future work toward human translation and experimental work toward biassing of cortical neural networks and decision making.

TBA

We will take a short break between speakers.

Good old fashioned generative models of image data This talk will cover some of the work I have been involved in recently, which mostly involves more classical generative modelling type approaches. This will include the Multi-Brain toolbox for SPM, which combines many of my previous works into a single framework intended for groupwise alignment of populations of brain (or other) images. In addition to achieving good alignment across subjects using a variety of imaging modalities, this same generative modelling framework has been applied to tissue classification, intensity nonuniformity correction and image synthesis. In addition, I will say a little about some ongoing work on computing denoised quantitative parameter maps from multi-echo MRI sequences. This work has been incorporated into the NiTorch Python library.

A few words before we close.