Eighth Clinical and Translational Science Ontology Workshop
The Clinical and Translational Science Ontology Group (CTSOG) invites you to join us March 16-18, 2022 in Orlando, Florida to discuss AI, Complex Systems in Biomedicine, and the role of ontology both in tempering the expectations of AI and advancing it to goals it can achieve. For example, we hear things all the time like “Google’s deep-learning program for determining the 3D shapes of proteins stands to transform biology, say scientists.” An optimism of this sort as to the potential of AI is shared by many working in the field of clinical and translational science. The purpose of this workshop is to explore the basis for this optimism, by looking at successes and failures of AI in different areas of biomedicine.
Bill Hogan, Jobst Landgrebe, Barry Smith
Bill Hogan (University of Florida College of Medicine, Gainesville, FL), firstname.lastname@example.org
Barry Smith (University at Buffalo, Buffalo, NY), email@example.com
- University of Florida Clinical and Translational Science Institute Biomedical Informatics Program
March 16 (Wednesday) - 18 (Friday), 2022
Embassy Suites by Hilton Orlando Lake Buena Vista South, Kissimmee, FL (20 minutes from Orlando airport).
Registration is free, but we absolutely need you to register for planning purposes.
To register for the meeting, click here.
Tuesday March 15th
Pre-Workshop Informal Meet & Greet: We will meet between 7pm and 10pm ...
Wednesday March 16th
Working session on the ontology of social determinants of health
9:00 Clint Dowland, "Social Categories as Cognitively Represented Person Aggregates"
10:45 Matt Diller, "What We Mean When We Talk About Occupations"
- We discuss two topics of ongoing work motivated by key SDoH use cases: (1) occupation and (2) social categories such as ethnic and gender categories
1:15-3:15 Jobst Landgrebe, AI and the Ontology of Complex Systems
- Abstract: “Google’s deep-learning program for determining the 3D shapes of proteins stands to transform biology, say scientists.” An optimism of this sort as to the potential of AI is shared by many working in the field of clinical and translational science. The purpose of this tutorial is to explore the basis for this optimism, by looking at successes and failures of AI in different areas of biomedicine.
- 1. Medicine: A Science of Complex Systems (0:00:00)
- The Cartesian (Mechanistic) View of Medicine (0:02:23)
- Cartesian Sciences Describe Simple (= Logic) Systems (0:05:15)
- Phenomena Underlying Medicine are Complex System Processes (0:14:10)
- 2. AI in Clinical Medicine: An Overview (0:28:59)
- Limits of AI in Clinical Medicine (0:42:36)
- 3. AI for Biomedical Research (1:00:37)
- AlphaFold (1:01:30)
- 4. Conclusion (1:26:15)
- AI Will Bring Many Benefits to Medicine, But it Will Not Change its Nature as a Heuristic Science (1:26:15)
- AlfaFold: One of the Greatest Achievements on the Part of Human Beings since Cologne Cathedral (1:32:2)
Jobst Landgrebe is the founder and managing director of Cognotekt, an AI company based in Cologne, Germany, focusing on the creation of structured data from natural language text. Dr Landgrebe is an MD with a background in biomedical informatics. He is the co-author, with Barry Smith, of Why Machines Will Never Rule the World. Artificial Intelligence without Fear, to be published by Routledge in summer 2022. This tutorial is co-sponsored by the University at Buffalo Working Group on Artificial Intelligence and Complex Systems
- Abstract: In an increasingly polychronic patient population, acute and/or new diseases (e.g. Covid) present a distressing reality of our limitations in providing high quality clinical care. A key challenge for Clinical Decision Support has been in building executable clinical guideline models that can interpret patient data and provide insights at the point of care. In this talk, the CPG-on-FHIR open standards based framework for creating shareable and reusable Clinical Practice Guidelines (CPG) is described.
5:30p Reception / Working Dinner
Thursday March 17th
- 9:00 Bill Hogan, Ontology for Social Determinants of Health with a Focus on Education Slides
- Educational attainment is a key social determinant of health. I will report on classes developed in the Ontology of Medically Related Social Entities (OMRSE), where we defend a view that education imparts both knowledge (as an ICE) and the neurologically-based skills (some motor, some more intellectual) to apply it.
- The 'replication problem' is a phrase to describe the inability of scientific communities to independently confirm the results of scientific work. It has plagued medicine as a positive science since its beginnings (Virchov and Pasteur). But it has become worse over the last 30 years and has massive consequences for healthcare practice and policy. This talk explains the reasons for the replication problem in medicine and why it is here to stay.
- 11:00 Coffee
- 11:15 Amelia Kahn, Definitions of Uncertainty Slides
- A preliminary survey of meanings assigned by government agencies to estimative words in the domain of probability and statistics.
- 12:15 Lunch
- Part 1 provides a general theory of capability as a universal intermediate between function and disposition in BFO. A capability is defined as a disposition in whose realisation some organism or group of organisms has an interest.
- Part 2 develops an ontology of language according to which a language is a capability of a linguistic community. The theory is tested in application to dialects phenomena.
- 14:00 Clint Dowland, Language Capabilities and Clinical Demographics Slides
- 15:00 Matt Diller, The Ontology of Money with a View towards Economic Determinants of Health Slides
- Matters of money constitute an entire category of social determinants of health. To understand them, we first wish to define 'money'.
- 16:00 Jihad Obeid, Phenotyping Using Deep Learning Text Classification: Can Ontologies Play a Role? Slides
- The use of electronic health records (EHR) to identify specific clinical phenotypes has gained significant momentum over recent years. A variety of natural language processing pipelines leverage ontologies for of clinical text annotation and information extraction. With the advent of vast computational power, significant strides have been made in deep learning approaches. During this presentation, we will discuss various use cases of clinical text classifiers, using both traditional machine learning algorithms and deep learning. We will examine the performance and utility of these models for both phenotyping and predictive tasks in a variety of clinical scenarios as well as the impact of pre-trained simple language models. Future approaches using more advanced language models and ontologies will be considered.
6:30pm Working dinner
Friday March 18th
9a-12p Working session and discussion of next steps, closing
- Discussion and next steps
Presentations by Jobst Landgrebe on explainable AI and by Barry Smith on healthcare economics ontologies
- Barry Smith, co-organizer, University at Buffalo
- Jobst Landgrebe, co-organizer, University at Buffalo
- William Hogan, co-organizer, University of Florida
- Sivaram Arabandi, Ontopro
- Ravi Bajracharya, datum.md
- Jiang Bian, University of Florida
- Sarah Bost, University of Florida
- Naomi Braun, University of Florida
- Matt Diller, University of Florida
- Clint Dowland, University of Florida
- Bill Duncan, University of Florida
- Yi Guo, University of Florida
- Hank Head, Optum Inc.
- Amelia Kahn, University at Buffalo
- Alex Loiacono, University of Florida
- Jihad Obeid, Medical University of South Carolina
- Samson Tu, Stanford University
- Donny Weinbrenner, University of Florida
- Pengfei Yin, University of Florida