Modeling Urgency Signals in Dorsal Premotor Cortex
My name is Liam Sennott, and I am a rising senior at Boston University. I am a neuroscience major and a computer science minor. I worked in the Chand Lab remotely in the Summer of 2020 through Boston University’s UROP program.
For my project, I created a mathematical model to help explain “urgency” signals in the dorsal premotor cortex (PMd). I did all of my work using MATLAB.
The dorsal premotor cortex (PMd) is a brain area implicated in perceptual decision making. Prior work has shown light on how this brain region processes sensory evidence related to decision making. However, how an internal state such as urgency affects the decision making process and how it manifests in this brain region is currently unclear.
In order to tackle this project, my goal was to build a model in MATLAB based on a recent study of decision-making done by Peter R. Murphy and his collaborators (Nat. Comm. 2016); this model would be adapted to test possible sources of urgency in PMd. To start, I implemented a simple “Lorenz attractor” system in MATLAB, which taught me the fundamentals of solving differential equations in MATLAB. The following figure displays the output of the solved Lorenz attractor system.
From there, I began working on building my model. The model assumes that decisions emerge from a competition between two accumulators that integrate evidence for each choice over time. The accumulator that wins the competition is the decision generated by the model. The model contains a baseline input as well as a gain multiplier on the evidence for each choice. I chose to investigate two hypotheses on the source of urgency in PMd: 1) urgency is related to initial input on this brain region 2) urgency is related to level of gain.
After testing the effects of altering baseline input and gain levels, I found that the latter had a more dramatic effect on the choice selective signal. When baseline gain increased, the choice selective signal increased more rapidly (see figure below).
These two figures show that a notable difference between the decision accumulator arises as gain levels increase (right), but this isn’t the case for an increasing current (left). This suggests that the gain parameter could be emulating an urgency signal.
During my two months of work this summer, I learned much about the process of using MATLAB to solve differential equations as well as using a model to experiment with a simulated system. I also refined my MATLAB programming knowledge and ability to produce quality plots. Finally, this project was met with some difficulty, due to it being conducted remotely. However, I am grateful to have had the opportunity to perform a research project from home.