Announcing 2019 Investigators Awards
From preventing postpartum hemorrhage to developing a “Precision Counselor,” the seven projects funded through Precision Health’s 2019 Investigators Awards exemplify the breadth of approaches precision health research encompasses, as well as the diverse expertise of University of Michigan researchers across the School of Public Health, College of Engineering, and Michigan Medicine.
We are pleased to announce the Investigators Awards recipients, who will be funded with $300,000 over two years to pursue their research:
“Developing an Early Warning System for Treatment of Postpartum Hemorrhage Using Time-Series Machine Learning Models”
Thomas Klumpner, MD, Clinical Assistant Professor of Anesthesiology and Obstetrics and Gynecology; Karandeep Singh, MD, MMSc, Assistant Professor of Learning Health Sciences, Internal Medicine, Urology, and Information
“While the maternal mortality rate has decreased throughout the developed world, in the US, the maternal mortality rate is increasing,” says Klumpner. “Postpartum hemorrhage is one of the leading causes of preventable maternal death. The existing systems to detect postpartum hemorrhage rely heavily on clinicians to be vigilant and do not detect hemorrhage early enough. As a team of experts in managing postpartum hemorrhage, machine learning, and systems design at the University of Michigan, we plan to develop a machine learning model that will help us build a better early warning system for postpartum hemorrhage and improve the safety of maternal care.”
“Characterizing and Understanding Time-varying Functional Connectivity States via Network Science and Deep Neural Networks”
Says Koutra, “The field of dynamic functional connectivity (dFC) is a newly booming and exploratory branch of neuroscience aimed at understanding how the connectivity (i.e., relationships across distinct regions of the brain) in functional brain networks may change over time, even in a ‘resting state.’ The overarching aim of our work is to shake up the existing statistical paradigms with entirely new approaches to tackle the complex problem of detecting dynamicity in noisy fMRI data with data-driven techniques that take maximal advantage of both the spatial and temporal granularity of fMRI.
“By taking a multidisciplinary approach that combines neuroscientific principles with cutting-edge computational methods like deep learning and network science, we will contribute a principled, data-driven framework that will provide much needed generalizability, interpretability and reproducibility to the field of dFC. Our work will bring the tenets of precision medicine to the critically important study of dFC. By enabling the robust and precise identification of baseline time-varying FC, our work may lead to better and earlier non-invasive diagnostics for various psychiatric disorders and may help to better understand their functional origins in the brain.”
“Precision Counselor: Natural Language Processing for Enhanced Behavior Counseling”
Rada Mihalcea, PhD, Professor of Electrical Engineering and Computer Science; Kenneth Resnicow, PhD, Professor of Health Behavior & Health Education; Veronica Perez-Rosas, PhD, Assistant Research Scientist, Electrical Engineering and Computer Science
“Patient-centered behavioral counseling for substance abuse, medication adherence, and other behavior changes is a cornerstone of our health care delivery system,” says Mihalcea. “We are very excited to embark on this collaboration involving expertise in both Artificial Intelligence and Motivational Interviewing to start developing computational methodologies that can assist counselors in their interventions and allow them to receive timely and cost-effective feedback.”
“Synthesizing Tumor Infiltrating Lymphocyte Patterns with Genomic Measurements for Head and Neck Cancer Survival”
Laura Rozek, PhD, Associate Professor of Environmental Health Sciences; Maureen Sartor, PhD, Associate Professor of Computational Medicine and Bioinformatics and Biostatistics; Arvind Rao, PhD, Associate Professor of Radiation Oncology and Computational Medicine and Bioinformatics
This project will apply a newly developed machine learning algorithm to quantify the spatial distribution of tumor infiltrating lymphocytes (TILs) in the microenvironment of head and neck squamous cell carcinomas (HNSCCs). Researchers will then evaluate disease-specific survival prediction performance of spatial TIL architecture measurements, alone and in combination with genomic and/or clinical prognostic factors. The aim of the project is to develop a clinically feasible protocol to classify patients into risk-stratified groups, based on TIL spatial infiltration, for precision therapy.
“Personalized Data-Driven Balance-Training Instruction and Assessment for Older Adults”
“We are very excited to work with an interdisciplinary team to develop and test algorithms that automatically progress older adults through home-based balance exercises for both preventative and therapeutic applications,” says Sienko. “The precision health approaches that we will use enable us to personalize balance training recommendations on an individual-by-individual basis. We hope that our findings will inform the design of a smartphone balance trainer that can complement clinic-based balance rehabilitation programs.”
“Precision Diagnosis in Patients with Acute Dyspnea by Linking Imaging and Clinical Data”
“Our work investigates how best to develop algorithms that combine data from digital images with clinical data from the electronic health record to support medical diagnosis,” Sjoding says. “Our focus is on the diagnosis of acute respiratory and cardiac conditions in hospitalized patients. I am a practicing pulmonary and critical care physician at Michigan Medicine, and My Co-PI, David Fouhey, is an expert in computer vision at the College of Engineering. This work leverages our joint expertise and the extensive clinical data resources available through Precision Health.”
“Short Tandem Repeats in Precision Health and Human Disease”*
(*co-funded by Precision Health and the A. Alfred Taubman Medical Research Institute)
Peter Todd, MD, PhD, Associate Professor of Neurology; Ryan E. Mills, PhD, Associate Professor of Computational Medicine and Bioinformatics and Human Genetics; Alan Boyle, PhD, Assistant Professor of Computational Medicine and Bioinformatics and Human Genetics
“About half of the human genome is made up of repetitive elements. For most of these repeats, however, we know almost nothing about whether they have normal roles in neurobiology or whether they contribute to human disease,” says Todd. “To explore these questions, we assembled an interdisciplinary, multi-departmental team of a physician scientist expert in neurological disease models, a bioinformatics expert on genome analysis, and a genetics expert who has designed tools to capture and sequence large pieces of DNA. Together, we have designed an innovative set of studies that will define variation in this ‘missing’ half of the genome and link it to Precision Health resources, to allow us to discover how repeat variation contributes to human disease.”