5th Illinois Health Data Analytics Summit
| Time | Item |
|---|---|
| 8:45 - 9:00 a.m. |
Welcome and Opening Remarks |
| 9:00 - 9:45 a.m. |
Keynote I | The Challenges of an N=1 Genetic Disease Paradigm in a Big Data World |
| 9:45 - 10:30 a.m. | Keynote II | The Measurement and Modeling of Social Behavior in Infancy James M. Rehg, Ph.D. - Professor School of Interactive Computing at the Georgia Tech, Co-Director Center for Health Analytics and Informatics Chair: Sheeba Arnold, Ph.D. - Assistant Professor at Carle Illinois Advanced Imaging Center |
| 10:30 - 10:40 a.m. | Break |
| 10:40 - 11:40 a.m. |
Session: AI & Data Science
|
| 11:40 a.m. - 12:00 p.m. |
Q&A Panel: Session AI & Data Science |
| 12:00 - 12:30 p.m. | Break |
| 12:30 - 1:00 p.m. |
Industry Talk - Predicting Knowledge: Finding the Unknown Knowns |
| 1:00 - 2:10 p.m. |
Session: Data Science and Small Data Sets
|
| 2:10 - 2:20 p.m. |
Q&A Panel: Session Data Science and Small Data Sets |
| 2:20 - 2:30 p.m. |
Break |
| 2:30 - 2:40 p.m. |
Guidance Breakout Session |
| 2:40 - 3:30 p.m. |
Breakout Sessions 1. Data Efficient Algorithms in Healthcare 2. Ensuring Security & Privacy in the Context of Small Data Sets 3. Rare Diseases and Individualized Medicine & Deep Phenotyping and Data Science |
| 3:30 - 3:45 p.m. |
Breakout Session Reports & White Paper Summary |
| 3:45 p.m. |
Closing Remarks |
Keynotes
Keynote I: Associate Professor Eric Klee, Mayo Clinic
Title: The Challenges of an N=1 Genetic Disease Paradigm in a Big Data World
Abstract: The scope of genetics in disease etiology continues to expand and create new paradigms of patient care, including significant advances in the treatment of rare genetic disease. Rare diseases affect less than 1 in 200,000 individuals, however, collectively are common, with approximately 30 million individuals in the United States impacted. Often medical diagnoses with Mendelian genetic cause are difficult to achieve based on clinical information alone and a patient experiences a diagnostic odyssey, defined as the patient’s and their family’s experience from onset to symptoms until the moment an etiologic diagnosis has been reached. This journey includes multiple clinical evaluations, imaging studies, and laboratory tests. It is well-known that such patients and families pay a heavy price during this journey, as it certainly results in delayed management, and often also in incorrect management along this journey. When genetic diagnoses are eventually identified, it is not infrequent that these patients are one of only a limited number of individuals with pathogenic defects in the identified gene, and very frequently are N=1 in the world for the specific variant identified.
While major advances in the care of these patients have occurred with the increase in clinical genome sequencing, there still exists major diagnostic and therapeutic challenges for this patient population. To sift through the thousands of genetic variants present in a patient to identify the causal defect requires systems that use structured phenotypes to identify putative genes of interest to focus the diagnostic inquiry. However, these structured phenotypes, using codified systems like the Human Phenotype Ontology (HPO), are often not readily available and natural language processing methods often fail to extract the appropriate information. Even when robust structured phenotypic and molecular data are present, there is a lack of tools available to enable “patient-like-mine” matching. This is very often a critical aspect of establishing new gene-phenotype or gene-disease associations or clarifying phenotypic expansions in known gene-disease relationships. However, the phenotypic variability, incomplete penetrance, and small N patient population with similar presentations have hindered computational methods for addressing this need.
Finally, there also exists a need for these types of systems to assist health care providers in identifying patients that will benefit from Clinical Genomics consultations and appropriate genetic testing. The lack of such a system is one of the elements leading to current patient diagnostic odysseys, as most health care providers are not extensively trained to recognize patients manifesting a rare genetic disease, which results in delayed referral to genetic specialty care.
In this talk, these challenges will be described, and case examples provided for contextualization. The overall objective will be to provide greater awareness of the challenges facing clinical genomics and hopefully stimulate engagement and discussion within the AI community, leading to novel innovation that will improve the quality of health care for patients suffering from rare genetic disease.
Keynote II: Professor James Rehg, Georgia Tech
Title: The Measurement and Modeling of Social Behavior in Infancy
Abstract: Beginning in infancy, individuals acquire the social and communication skills that are vital for a healthy and productive life. Children with an Autism Spectrum Disorder (ASD) face great challenges in acquiring these skills, resulting in substantial lifetime risks. As the neural basis for ASD is unclear, the diagnosis, treatment, and study of autism depends fundamentally on the analysis of social behavior. Standard methods for behavioral observation, which form the backbone of research, diagnosis, and treatment in young children at risk for ASD, are effective in providing a holistic characterization but are inherently coarse-grained and not easily scalable.
In this talk, I will describe our research agenda in Behavioral Imaging, which targets the capture, modeling, and analysis of social and communicative behaviors between children and their caregivers and peers. We are developing computational methods and statistical models for the analysis of video, audio, and wearable sensor data. I will present several recent findings, including a method for detecting eye contact using wearable cameras which has been shown to achieve human-level accuracy, and methods for measuring gaze shifts and hand movements from video. I will describe our recent work in collecting the first large scale publicly-available egocentric dataset of social behavior (Ego4D) and on-going work on developing latent state models for predicting trajectories of language development. I will describe potential applications of this technology to the diagnosis and treatment of autism and other developmental conditions. This is joint work with Drs. Nancy Brady, Rebecca Jones, Cathy Lord, Jenna McDaniel, and Agata Rozga, and Ph.D. students Eunji Chong, Miao Liu, Fiona Ryan, and Yun Zhang.
Planning Committee
Adam Cross
University of Illinois Chicago
David Liebovitz
Northwestern University
John Vozenilek
OSF HealthCare
Paul Arnold
Carle Hospital
Irfan Ahmad
University of Illinois Urbana Champaign
Colleen Bushell
University of Illinois Urbana Champaign
Maria Jaromin
University of Illinois Urbana Champaign
Antonios Michalos
University of Illinois Urbana Champaign
Lav Varshney
University of Illinois Urbana Champaign
Jared Rogers
OSF HealthCare
Jonathan Handler
OSF HealthCare
Ravi Iyer
University of Illinois Urbana Champaign
Jodi Schneider
University of Illinois Urbana Champaign
Rebecca Lee Smith
University of Illinois Urbana Champaign
Elizabeth T. Hsiao-Wecksler
University of Illinois Urbana Champaign
Lauren Laws
University of Illinois Urbana Champaign
Awais Vaid
Champaign Urbana Public Health District
Jimeng Sun
University of Illinois Urbana Champaign
May Wang
Georgia Tech
Roy Campbell
University of Illinois Urbana Champaign
Sridhar Seshasdri
University of Illinois Urbana Champaign
Yuxiong Wang
University of Illinois Urbana Champaign
Michelle Osborne
University of Illinois Urbana Champaign
George Heintz
University of Illinois Urbana Champaign
LEARNING THROUGH DATA
This year's Health Data Analytics Summit will be a virtual meeting via Zoom. A link will be provided after registration.
Are you interested in sharing your research at our summit? Please email us at hcesc@illinois.edu.