Research Projects

Data Classification and Image Segmentation

Mentor: Ekaterina Rapinchuk

Students working on this project will perform image segmentation on hyper-spectral imagery and execute data classification on high dimensional data sets. They will adapt and modify existing models to produce accurate results using their data. Students will also implement their models in Matlab or Python.
At the end of this project, students will become more proficient in programming, learn about data processing, and gain more knowledge about machine learning and optimization.

Keywords: Machine Learning & Optimization

Topological Data Analysis of 3D Images Using X-ray Computed Tomography

Mentors: Liz Munch & Dan Chitwood

Structure—whether inorganic matter, the morphology of living things, or objects crafted by culture—is data rich. Embedded within materials are structures that drive innovations in engineering and technology; encoded within living organisms is information sculpted by genetic and environmental forces; and within culture, the objects we craft are imparted with functionality and arise through cultural history and evolution. Using X-ray Computed Tomography, we are capturing exquisite 3D models (and 4D time lapses of dynamic objects) as volumetric, voxel-based images. Topology Data Analysis provides a method to quantify and compare these images against each other, providing a way to pursue a new, hypothesis-driven science that asks questions about the complex morphologies we see around us.
Students participating in this project will become familiar with X-ray Computed Tomography technology and producing their own 3D reconstructions of objects. Students will also learn the fundamental ideas underlying Topological Data Analysis (TDA), and be able to innovate new ways of applying TDA to scientific questions.  

Keywords: Topological Data Analysis, X-ray Computed Tomography

Computational Modeling of Cardiac Growth

Mentor: Tong Gao

This project aims to develop a computational modeling framework to simulate the abnormal growth of the human heart, taking into account some key physics occurring in the beating heart. We will focus on one chamber of the heart, the left ventricle, to study how blood flow interacts with soft muscle walls by coupling with physiological models which determine the tissue growth at the microscale.

At the end of this project, students will (1) gain basic knowledge of cardiovascular mechanics, (2) learn to use basic computation software/packages in cardiovascular modeling and simulation, (3) be able to perform engineering analysis to interpret numerical data and make connections to some abnormal mechanical behaviors.

Keywords: Computational Modeling & The Heart

Big Data that Shakes the Earth

Mentor: Min Chen

Monitoring and predicting the environment and climate change of our planet Earth requires advanced techniques to analyze and model big datasets. These datasets can take the form of field seismic signals that record the ground shakes during earthquakes or airborne and satellite images that contain rich information about our environment (ex. groundwater distribution in agricultural regions). In this project, students will learn how to analyze and model geospatial and temporal datasets in Earth Sciences, ex. using seismic signals and airborne and satellite images to identify and locate earthquakes, landslides, volcanic activities, and nuclear tests. Calculus I and II,  and some basic programing skills in Python would be desired. 

At the end of this project, students will be able to program in Python, R, and Fortran, to perform data analysis aided by machine learning and computational modeling, to create powerful scientific images through data visualization tools, and to communicate more effectively through both oral presentations and writing

Keywords: Big Data & Earth Sciences

Machine Learning Approaches for Automatic Detection of Change in Pediatric ICU Applications

Mentors: Adam Alessio

Pediatric patients in intensive care settings commonly receive multiple chest radiographs during their stay to assess changes in the patient's state.  These exams require time-consuming interpretation and there is a pressing need for strategies to automatically assess change.  This project will develop conventional and deep learning methods to automatically assess change using methodology previously developed for autonomous vehicles and digital photography. Additional programming experience or advanced programming is required for this project. Exposure to digital signal processing; image processing; machine or statistical learning is a bonus but is not required.

Students will be able to develop custom software solutions for basic image processing tasks including image enhancement and segmentation. Students will be able to apply machine learning image classification methods (from Keras and/or Tensorflow) to chest x-ray images.

Keywords: Machine Learning & Chest Radiograph

From Numbers to Colors

Mentors: Alexei Bazavov

Solving linear systems of the type M*X=B (M is a matrix, X, B - vectors) is a very generic numerical linear algebra problem that shows up across many fields. Often we employ iterative solvers that make a guess for the solution X and then progressively refine it until some predefined accuracy is reached. The task in this project is to study some simple iterative solvers, program one of them for a specific model of M and then make a visualization of how the convergence of X to the exact solution happens. Such visualization can be presented as a “temperature” plot where the colors represent, say, difference between X and X_exact: towards blue where it is smaller and towards red where it is larger. Prior background in Physics would be a bonus but is not required for this project.

At the end of this project, students will acquire understanding of some algorithms that are used across many fields for solving systems of linear equation, learn about how such algorithms can be programmed and gain skills related to data visualization.

Keywords: Lattice QCD

Using Deep Learning to Identify Combinations of Genetic Variants Important for Predicting Traits

Mentors: Shinhan Shiu

One grand challenge in science is how differences in DNA sequences between individuals (genotypes) translate into differences in phenotypes, i.e. how individuals differ in their characteristics (e.g.  height). The goal of the project is to apply deep learning methods on existing genotype and phenotype data to determine which DNA differences and their combinations are important to explain the phenotypes of interest.
At the end of this project, the students will be able to design experiments, critically assess data quality, and address questions using computational and data science approaches, particularly deep learning methods.

Keywords: Computational Biology

Computationally-driven Quantum Chemistry

Mentors: Angela Wilson, Prajay Patel and Lucas Aebersold

Ab initio quantum chemistry methods provide the theoretical means to predict physical properties of molecules, from structures to energies.  A drawback of these methods is that their mathematically complexity require extraordinary computer time, memory, and disk space.  Students participating in this project will become engaged in problems where routes to reduce these computational bottlenecks are addressed. Efforts will be focused on transition metal and heavy element species, where the chemistry becomes very complicated (and exciting!) due to the significant number of electrons encountered. Experience with Python, C, C++, or Fortran are strongly preferred.

Students will become familiar with a number of computational chemistry software programs,  fundamentals of quantum mechanics, and the computational approaches required to solve the challenging mathematical equations. Students will be exposed to many areas of computational chemistry during research group meetings, learning about many diverse areas of computational chemistry, as well as become involved in building a many-node computer cluster which the research group does each summer (and then uses for our lowest end calculations).  Students will become familiar with a wide variety of computer platforms, and will have the opportunity to use/learn several computer languages.

Keywords: Computational Chemistry

Computational Modeling of the Ailing Heart

Mentors: Jason Bazil, Quynh Duong

This project uses an integrated dynamical model of myocardial energetics to study the impact of mitochondrial dysfunction on heart contractile performance, calcium handling, and free radical homeostasis. The model will be used to computationally test hypotheses generated to explain the experimentally observed decrease in cardiac function of diseased and damaged hearts. Some experience with Matlab is desired, but not necessary.

At the end of this project, students will be able to use Matlab to simulate and analyze dynamical models.

Keywords: Mitochondrial Metabolism, Enzyme Kinetics, Free Radical Generation

Modeling turbulence from supermassive black holes in galaxy clusters

Mentors: Brian O'Shea, Philipp Grete and Deovrat Prasad

Galaxy clusters are the most massive objects in the universe. In addition to being home to hundreds of galaxies, they are filled with hot, diffuse plasmas. The massive central galaxy in a cluster is home to a supermassive black hole that creates a massive jet of magnetized plasma, hundreds of thousands of light-years long, that stirs this hot diffuse plasma and affects the behavior of the galaxy cluster. In this project, we will use simulations to model the ways that this plasma jet interacts with the cluster it lives within, and ultimately understand how this interaction affects the observable properties of the galaxy cluster. Introductory level physics (mechanics and electromagnetism) background is required for this project. Experience programming in Python and/or C++ is desired but not necessary.

At the end of this project, students will be able to run simulations on a supercomputer, analyze those simulations using a Python-based analysis tool, and be able to communicate the results of this analysis. In addition, they will have learned about the behavior of plasmas, about cosmological structure formation, and about large-scale computational physics simulations.

Keyword: Astrophysics simulation