CSL faculty-led projects receive funding from C3.ai DTI
The C3.ai Digital Transformation Institute (DTI) may have just been founded this spring, but it has already funded 26 research projects to help combat the COVID-19 pandemic. In addition to the funding, the researchers will also receive access to the C3.ai computing Suite and Microsoft Azure computing and storage to help complete their multi-disciplinary projects.
“I am delighted to see the interdisciplinary nature of the projects led by the CSL faculty members," said R. Srikant, CSL Professor and C3.ai DTI co-director. “The collaboration between engineers, computer scientists, epidemiologists and medical/healthcare professionals will produce both immediate and lasting benefits to society.”
In addition to Srikant being co-director, C3.ai DTI’s co-chief scientist is CSL Professor Tandy Warnow and five of the projects awarded funding are being led by CSL researchers.
"One of the wonderful benefits of C3.ai DTI is that it is supporting new collaborations between Illinois faculty and researchers at other institutions, around the world,” said Warnow.
Read on to see how they’re using high-level computing to make a difference in the fight against coronavirus.
COVID-19 Medical Best Practice Guidance System
One of the major issues in treating patients who have coronavirus is a shortage of staff trained to treat them. The goal of this project, led by CSL Professor Lui Sha, is to develop a guidance system that would provide real-time recommendations for patients based on the patient’s condition and current COVID-19 guidelines and research. Part of the project involves training medical professionals, which is why Sha is collaborating with physicians from OSF Children’s Hospital of Illinois and the University of Chicago Medical School.
The system will be backed by verifiable computations and models, which are aimed at improving the efficiency of
medical treatment. The first step is to develop a real-time guidance system module for physicians to use when patients have developed Acute Respiratory Distress Syndrome, which is the deadliest and most complex phase of COVID-19. The group then will create a guidance module for cardiopulmonary resuscitation.
Sha is the Donald B. Gillies Chair in Computer Science (CS). Co-PIs include Maryam Rahmanjheris and Grigore Rosu of Illinois; Paul Jezioczak of the University of Illinois College of Medicine Peoria; and Priti Jani of the University of Chicago.
Algorithms and Software Tools for Testing and Control of COVID-19
There have been a variety of approaches across the country for how to mitigate the spread of COVID-19. In order to understand which control strategies work, researchers need to analyze thousands of data points with little effort There is currently extensive testing of various populations for antibodies, which is expected to inform policymaking.
Several CSL faculty have come together to develop algorithms that will combine real-time testing data with epidemiological models to better inform decision makers on what effect control strategies like social distancing have on the spread of the coronavirus. Led by CSL Associate Professor Prashant Mehta, this research group also includes CSL’s Tamer Basar, Carolyn Beck, and CSL alum Philip Paré, now at Purdue, all of whom have spread of disease modeling, analysis, and control experience.
Mehta is a faculty member in mechanical science and engineering (MechSE), and an affiliate of electrical and computer engineering (ECE). Basar is the Swanlund Endowed Chair & CAS Professor in ECE. Beck is a professor in industrial and enterprise systems engineering. The other Co-PIs are Rebecca Smith and Matthew West of Illinois.
Mining Diagnostics Sequences for SARS-CoV-2 Using Variation-Aware, Graph-Based Machine Learning Approaches Applied to
SARS-CoV-1, SARS-CoV-2, and MERS Datasets
Rapid diagnostics is important for general medical care, but especially important during a pandemic like SARS-CoV-2 or coronavirus. CSL’s Nancy Amato and Lawrence Rauchwerger are co-leading a team attempting to identify potential genetic differences in SARS-CoV-2 that could throw off testing that relies on RNA-sequencing.
The group will compare the genomic information of SARS-CoV-2 to SARS-CoV-1, the viral agent that caused the SARS outbreak in 2002, as well as to Middle East Respiratory Syndrome, or MERS. These comparisons will allow them to determine biological differences that could have allowed the severity of COVID-19 to reach the pandemic-level.
Amato is the Abel Bliss Professor of Engineering as well at department head for CS at Illinois. Rauchwerger is also an Illinois CS professor. Rice University CS Assistant Professor Todd Treangen and his group are also collaborating on this project
Adding Audio-Visual Cues to Signs and Symptoms for Triaging Suspected or Diagnosed COVID-19 Patients
Nation-wide social-distancing efforts served to flatten the curve and not overwhelm hospitals with COVID-19 patients. Even with appropriate measures in place, some hospitals were still flooded with patients, resulting in some sick patients being discharged only to have to return to the hospital later, while some healthy patients were admitted unnecessarily. CSL’s Narendra Ahuja is leading a research team, which includes CSL’s Mark Hasegawa-Johnson, to prevent this from happening in the future by using audiovisuals coupled with artificial intelligence and machine learning algorithms.
The group is developing a tool that will read an audio-visual of a patient either from a doctor’s office or during an at-home telemedicine appointment and use it to predict the patient’s likelihood of deteriorating or improving. This information would then be uploaded to the patient’s medical record to better inform doctors.
Ahuja is affiliated with both the ECE and CS departments at Illinois. Hasegawa-Johnson is affiliated with ECE. Contributors include David Beiser, David Chestek, and Jerry Krishnan of the University of Chicago, and Arun Singh of All India Institute of Medical Sciences, Jodhpur.
Secure federated learning for clinical informatics with applications to the COVID-19 pandemic
There are a number of laws and programs in place to make sure medical data is secure, but this can be problematic when a wide-spread problem like coronavirus occurs and data is unable to be shared for privacy and intellectual property (IP) reasons. CSL’s Sanmi Koyejo is leading an effort, involving CSL's George Heintz, to use modern cryptography and distributed machine learning to enable clinical AI on data distributed across medical units while preserving patient privacy and IP rights.
The innovative cryptography and distributed machine learning techniques developed from this project will enable adaptive clinical AI, which can be deployed quickly when there is an outbreak. The group would like their technology to serve as a hub for the other c3.ai DTI projects to enable the secure distributed machine learning and help everyone improve their results.
Koyejo and collaborator Dakshita Khurana are both assistant professors in CS at Illinois. Heintz, is a Program Manager for the Health Data Analytics Initiative, within HCESC. They are joined by William Bond of Jump Simulation and OSF Healthcare; and Roopa Foulger, OSF Healthcare.
More information about each of these projects and the full listing of projects is available on the C3.ai DTI website.