Below are featured projects for the 2018 Summer Undergraduate Research Program in Neuro Computation (uPNC) project. Selecting the name of the mentor will take you to a short description the research project.
Mentor: Jana Kainerstorfer
University and Department:Carnegie Mellon University, Biomedical Engineering
Project Description: We have previously shown that hemodynamic changes in the brain can be measured by optical means via near-infrared spectroscopy (NIRS) and correlated to intracranial pressure (ICP) changes. While the ICP measurement is a highly invasive procedure, the optical measurement can be conducted with a non-invasive near-infrared spectroscopy (NIRS) system for hemoglobin concentrations and a non-invasive diffuse correlation spectroscopy (DCS) system for blood flow changes in the microvasculature. The goal of this project is to use regression models, feature extraction and machine-learning algorithms to estimate the absolute ICP based on non-invasive data.
As a first step, a non-linear kernel regression model after Xu et al.  will be implemented, that estimates ICP based on blood pressure and blood flow changes in the brain. Xu’s approach is based on cerebral blood flow velocity in large arteries, while the optical system used here measures the flow rate in the microvasculature, which is assumed to be more sensitive to pressure changes in the brain.
Depending on the success of aforementioned method, further approaches include feature-based methods such as k-means, random decision forests and deep learning algorithms. Features that need to be extracted for these methods include, but are not limited to, principle components, moving averages, derivatives, time delays and magnitude differences between non-invasive measurements at known frequencies. These frequencies induce sinusoidal ICP changes, and heart and respiration rates.
The goal for the summer is to implement the non-linear regression model of Xu et al and depending on its performance proceed to methods that are more complex.
 Xu P. et al (2010) Improved noninvasive intracranial pressure assessment with nonlinear kernel regression. IEEE Trans Inf Technol Biomed 14:971–978.
Mentor: Matthew Smith
University and Department: University of Pittsburgh, Ophthalmology
Project Description: Recent work in systems neuroscience has benefited from molecular tools that identify cell types and allow researchers to tease apart their roles of neural circuits important for sensation, perception and cognition. Extracellular electrophysiology has been unable to take advantage of many of these cellular distinctions, leaving us to only consider neuronal responses on average, and not taking into account separate inhibitory and excitatory neurons, or subtypes of each. Recent work combining electrophysiology and molecular tools in slice physiology has allowed us to identify spiking characteristics of neuronal subtypes, such as waveform shape, firing rate, inter-spike interval, and more. Accessing information about neuronal subtypes in extracellular neurophysiology would enable a host of new investigations into their role in neural circuits that support perception and cognition. We will apply classification methods from machine learning to describe large-scale electrophysiological recordings from non-human primates, using a multi-dimensional approach. We will then use these classifications to determine how different neuronal types play a role in network dynamics and stimulus response in data from behaving animals.
Mentor: Byron Yu
University and Department: Carnegie Mellon University, Electrical and Computer Engineering and Biomedical Engineering
Project Description: The Yu group works at the intersection of neuroscience, machine learning, and biomedical engineering. We study how large populations of neurons interact during sensation and behavior, using brain-computer interfaces and advanced statistical methods. Example projects include studying how neural activity changes during learning and developing dimensionality reduction methods. Applicants should have a strong math background, particularly in probability and linear algebra.