Areas of Focus

  • Software (Controls Estimation, Machine Learning, Computational Modeling)
  • in vitro (Brain Slicing, Patch Clamping, Electrophysiology)
  • in vivo pre-clinical experiments (novel electrode design, stimulation parameter optimization, real time modeling, closed loop stimulation systems)
  • in vivo clinical trials with neuromodulation industry leaders and clinical University of Minnesota labs

Computational Modeling

Different regions of the brain produce different kinds of epileptiform activity. One region of the hippocampus, the CA3, produces short bursts of activity which we liken to the clinical interictal bursts while a neighboring region, the CA1, can produce full fledged long lasting seizures. This work can be seen on my Small World Networks page. We have hypothesized that the difference in the wiring of the two regions results in the differences in their two behaviors. By generating large networks of model neurons and adjusting the connectivity of the network, we could reproduce both behaviors. We are now trying to extend this work to understand how inhibitory cells change or even control the dynamics of seizure onset.


We study seizures in slices of brain tissue using patch-clamp recording and dynamic clamp techniques as well as optical imaging. We supplement our experiments with modeling of large networks of neurons to test our hypotheses in a controlled experimental condition.

Patch Clamp Recordings

Using a micropipette of glass pulled to a very fine tip, we can record from a neuron with such high fidelity we can measure the synaptic inputs impinging on the neuron. By patching two cells simultaneously and recording seizures, I discovered that synchrony between cells decreased during seizures contrary to the common dogma that seizures are "hypersynchronous" neuronal activity.

Dynamic Clamp

This is a relatively new technique where you use a computer to calculate the current to be injected into a cell based on the voltage recorded at the cell. This allows us to:

  • create very complicated protocols
  • simulate synaptic conductances (rather than currents that are independent of the voltage of the cell)
  • couple two neurons using virtual synapses
  • simulate a neuron on the computer and connect it to a real neuron
  • simulate ion channel conductances

We use the dynamic clamp to measure phase response curves of neurons (PRCs). This is a dynamical measure of a cell where we measure how much a synaptic input modulates the timing of an action potential in a periodically firing neuron. PRCs are extremely useful for predicting how a change, such as the increase or decrease in a particular ion channel density, could affect a network of neurons.


Patch clamp is limited to recording from one or two cells at a time. Optical imaging allows for the recording from hundreds of cells simultaneously. Particular classes of calcium dyes allow us to load cells with the dye and measure the change in intracellular calcium concentrations that occurs with cellular activity. We have been able to measure activity of the slices during seizures. In this movie you can see a large activation of the superficial layers of the hippocampus just at the onset of the seizure. We think this reflects the large activation of the inhibitory population at the onset of the seizure.

Seizure Prediction

In conjunction with Dr. Keshab Pahri in Electrical Engineering, we are developing robust seizure prediction algorithms using support vector machines. This method is a very robust way to classify data into ictal or interictal categories.