Decision-making is a complex behavior, partially innate and partially learned. Some aspects of our ability to make choices are highly conserved throughout evolution, and others are characteristically human. This process is influenced by various constraints, from our biological makeup to the information we have at the moment of making a choice. Neurally, it engages a wide network of densely interconnected brain regions that communicate at synaptic time scales (sub-second). To understand human cognition, we need to examine behavior and the underlying influencing factors, and the patterns of concurrent neural activity. In addition, we need to account for the human part - what unique decision strategies do humans use?
To understand how our brain makes decisions, we use a combination of computational models of behavior and invasive neurosurgical recordings.
We carry out deep brain recordings by leveraging intracranial clinical interventions. Patients receive surgery for the treat mentor diagnosis of brain disorders, opening a unique opportunity to study brain activity in a way that would otherwise not be possible.
In addition to recording brain activity, we utilize targeted electrical stimulation. This provides causal evidence for the involvement of individual brain areas in behavior, and can also be used for therapeutic interventions.
Using custom-made carbon fiber electrodes, we measure the concentration of neuromodulators intimately involved with cognition and disease such as dopamine and serotonin with high temporal (sub-second) resolution and anatomical accuracy.
To understand how decisions are made, we need to observe behavior. Using quantifiable and reproducible tasks (i.e.computer games) inspired by behavioral economics and game theory, we can elucidate how humans make decisions.
We analyze behavior through a computational modeling lens. This allows us to categorize and quantify individual behavior and examine the underlying factors of decision making.
To examine brain function, we record electrical activity during behavioral tasks. This electrophysiological response provides a rich characterization of neuronal activity and brain oscillations with high temporal resolution in the context of behavior.
Most decisions are carried out under uncertainty -contingencies are not known, probabilities shift through time, and the consequences of our actions are uncertain. This requires neural systems that can estimate the risks and potential rewards of available options, and are capable of learning. Using reinforcement-learning models, we characterize the underlying neural networks of decision and learning processes.
We examine the relationship between neural activity across frequency bands and brain regions (orbitofrontal, lateral prefrontal, cingulate cortices, etc.) and overt choice behavior using a combination of iEEG recordings and neuroeconomic probes of decision-making.
Brain regions are specialized, but are functionally organized in circuits and networks through dense interconnections. As a consequence, generating behavior requires the coordinated activity of multiple brain areas. During choice, decision information is highly distributed. We seek to understand the contribution of distributed and localized brain activity to these processes, and to build decoding models that can predict behavior from neural activity alone. In the future, we hope to generalize these models to be able to build individualized models that predict the mapping of brain activity to brain states with high accuracy.
We use linear dynamical systems to characterize time-varying neural dynamics across a variety of brain regions, both prefrontal (orbitofrontal, lateral prefrontal, cingulate cortices) and subcortical (striatum, hippocampus, amygdala) and construct patient-individualized decoding models that can predict trial-by-trial behavior with high accuracy.
Our current approaches for the treatment of depression are insufficient - there is a significant proportion of patients who don’t improve after therapy or pharmaceutical treatment. For these patients, a potential treatment avenue involves neurosurgical approaches where affected brain areas are directly stimulated using a chronically implanted electrode. Because depression is highly prevalent in intractable epilepsy patients undergoing invasive electrophysiological monitoring, we use this opportunity to study differences in their behavior and brain. We hope to combine the insights derived from these studies with invasive neurostimulation approaches to develop new patient-tailored therapeutic strategies.
In this project, we combine distributed iEEG recordings and reinforcement learning models of decision-making to study reward and mood processing across multiple brain areas. This approach allows for identification of the brain areas with aberrant activity during reward processing. Additionally, this approach uses an algorithmic targeting strategy to develop new stimulation therapeutic approaches.
In addition to the individual activation of specialized brain areas, behavior depends on the action of neuromodulator systems that regulate brain activity globally. The action of these neuromodulators, such as dopamine and serotonin, is essential for correct brain function in decision-making, motor control and mood. However, studying them directly in the human brain at the temporal resolution needed to study behavior is difficult.
In this project, we use custom-made carbon fiber electrodes during surgical deep-brain stimulation interventions to carry out fast (10 times per second) estimation of neuromodulator concentrations in deep brain areas. In this way we study the subsecond dynamics of dopamine and serotonin during behavior.
