THEME: THE FUTURE OF GRADUATE PROGRAMS IN BIOMEDICAL ENGINEERING��PREPARING STUDENTS FOR ACADEMIC POST-PHD CAREERS IN THE AI ERA

“Skate to where the puck is going, � not where it has been.”

​

Wayne Gretzky

SESSION CHAIRS

Wawrzyniec “Wawosz” Dobrucki

Associate Professor

Department of Bioengineering

University of Illinois at Urbana-Champaign

2

Chris Geiger

Instructional Associate Professor

J. Crayton Pruitt Family Department of Biomedical Engineering

University of Florida

SESSION’S AGENDA

Session Objectives and Sub-Topics

• An overview of the AI landscape in BME
• Challenges and opportunities presented by AI advancements (Case Studies in Research and BME Education)
• Best practices (breakout session, 20 min)
• Strategies to equip current graduate students with the skills and knowledge needed to thrive in an AI-driven environment
• Evolving Landscape (breakout session, 20 min)

​

3

CASE STUDY 1: AI IN RADIOLOGY - DETECTING LUNG � CANCER

Application

Google developed a deep learning algorithm to detect lung cancer from CT scans with a performance comparable to that of radiologists.

4

https://research.google/blog/computer-aided-diagnosis-for-lung-cancer-screening

Challenges

Data Quality: The AI system's performance depends heavily on the quality and quantity of the training data.

Integration: Integrating AI systems into existing clinical workflows can be complex and requires significant changes in infrastructure.

Regulatory Approval: Ensuring the AI system meets regulatory standards for clinical use is a lengthy and rigorous process.

AI in Research

CASE STUDY 2: AI IN GENOMICS - PREDICTING PROTEIN � STRUCTURES

Application

AlphaFold, an AI system developed by DeepMind, predicts protein structures with remarkable accuracy, solving a 50-year-old challenge in biology.

5

https://deepmind.google/technologies/alphafold/

Challenges

Complexity: Understanding and interpreting the AI’s predictions can be complex, requiring specialized knowledge.

Computational Resources: Training and running such sophisticated AI models require extensive computational resources, which may not be accessible to all research institutions.

Validation: While AlphaFold’s predictions are highly accurate, experimental validation is still necessary to confirm the structures in a biological context.

AI in Research

CASE STUDY 3: AI IN PERSONALIZED MEDICINE – � PREDICTING TREATMENT OUTCOMES

Application

IBM Watson uses AI to analyze medical literature and patient data to provide oncologists with treatment recommendations tailored to individual patients.

6

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6231834/

Challenges

Data Integration: Integrating diverse data sources (e.g., electronic health records and genomic data) into the AI system is complex and requires standardization.

Trust and Acceptance: Gaining the trust and acceptance of clinicians in AI recommendations is crucial for widespread adoption.

Bias and Fairness: Ensuring that the AI system provides unbiased recommendations and addresses disparities in healthcare is essential.

AI in Research

CASE STUDY 1: AI IN PERSONALIZED LEARNING – � INTELLIGENT TUTORING SYSTEMS (ITS)

Application

AI-driven ITS provides personalized learning experiences by adapting to individual students' needs and learning styles.

7

https://www.youtube.com/watch?v=IvXZCocyU_M

AI in Education

CASE STUDY 2: AI IN ASSESSMENT - AUTOMATED GRADING � SYSTEMS

Application

AI systems can grade multiple-choice questions, short answers, and even essays, providing timely feedback to students.

8

https://www.princetonreview.com/ai-education/how-ai-is-reshaping-grading

Challenges

Accuracy: Ensuring the AI system accurately grades complex and open-ended responses can be challenging.

Interpretability: Students and educators need to understand how the AI system arrives at its grading decisions.

Acceptance: Gaining trust in AI grading systems from both students and educators is crucial for their adoption.

AI in Education

CASE STUDY 3: AI IN RESEARCH SKILL DEVELOPMENT – � AI-ENHANCED LITERATURE REVIEW TOOLS

Application

AI tools can assist students in searching, screening, and summarizing vast amounts of scientific literature.

9

https://tamu.libguides.com/c.php?g=1289555

Challenges

Reliability: Ensuring the accuracy and reliability of AI-generated summaries and recommendations is crucial.

Learning Curve: Students may need training to use AI literature review tools effectively.

Dependency: Over-reliance on AI tools may limit the development of traditional literature review skills.

AI in Education

BEST PRACTICES�PREPARING FUTURE ACADEMICS AND PROFESSIONALS

10

BEST PRACTICES�PREPARING FUTURE ACADEMICS AND PROFESSIONALS

Interdisciplinary curriculum development

• Integration of AI and data science (foundational courses in machine learning, data analytics and programming languages)
• Project-based learning (apply AI tools/techniques to solve real-world biomedical problems)
• Collaborative courses (combined programs with departments of computer science, i.e. UIUC CS-BIOE program)

​

Emphasis on practical experience

• Internships and co-op programs (encourage students to participate in internships and co-op programs with companies and research institutions that focus on AI applications)
• Hackathons and competitions (Organize or promote participation in AI and biomedical hackathons and competitions to provide hands-on experience and foster innovation)
• Research opportunities (facilitate student involvement in AI-related research projects through labs, research assistantships, and collaborative projects with faculty)

​

​

11

BEST PRACTICES�PREPARING FUTURE ACADEMICS AND PROFESSIONALS

Skill development workshops

• Technical skills (AI frameworks and tools such as TensorFlow, PyTorch, and Keras)
• Soft skills (Offer workshops on communication, teamwork, and project management skills, which are essential for interdisciplinary collaboration and academic success)
• Ethical AI practices (Provide training on ethical considerations in AI, including data privacy, algorithmic bias, and responsible AI usage)

Ethical and responsible AI use

• Ethics courses (addressing the implication of AI in healthcare and biomedical research)
• Case studies (to discuss real-world scenarios and ethical dilemmas related to AI in biomedical engineering)
• Regulatory knowledge (educate students about the regulatory frameworks and standards governing AI in healthcare and biomedical research)

​

12

BREAKOUT SESSION I (20 MIN)

Discuss opportunities and threats and how AI might shift current practices moving forward. Also, think about how current or potential graduate student training can help communicate/prepare students for these. Break out into the same 4-6 groups. At the end of 20 minutes, each group will spend 1-2 minutes talking about an identified opportunity, threat, and/or paradigm shift, as well as any potential training implementations. Please use Jambord (use QR code) to jot down your ideas.

EVOLVING LANDSCAPE

14

EVOLVING LANDSCAPE

Opportunities

• Improving the language quality of papers for non-native English speakers.
• Aid in comparing multiple documents for similarities.

​

Threats

• Plagiarism.
• Hallucinations (factual errors), method of AI training (sources of training datasets, how to sieve out true information from a fake one).

​

Paradigm shifts

• Where/how information is gathered (information behind paywalls, companies withholding proprietary information from training models, speed of research/discovery).
• How material is covered in class and how students are tested.

15

EXISTING POLICIES TOWARD THE USE OF AI

Stanford University

• Policy: Stanford Human-Centered Artificial Intelligence (HAI) Guidelines
• Description: Stanford HAI has established guidelines to ensure that AI research and education are aligned with ethical principles and societal values.
• https://hai.stanford.edu/

MIT

• Policy: MIT Schwarzman College of Computing Ethical AI Guidelines
• Description: MIT’s Schwarzman College of Computing has guidelines for ethical AI research and education, focusing on the societal implications of AI.
• https://computing.mit.edu/ai-policy-briefs/

University of California, Berkeley

• Policy: Center for Human-Compatible AI (CHAI) Ethical Guidelines
• Description: The Center for Human-Compatible AI at UC Berkeley has established ethical guidelines for AI research and education.
• https://humancompatible.ai/

16

BREAKOUT SESSION II (20 MIN)

Discuss how your institution is currently helping students prepare for the use of AI (in the context we are describing) moving forward in BME. Break out into 4-6 groups with a recorder/reporter. At the end of 20 minutes, each group will spend 1 minute reporting to the group the most innovative idea currently being implemented. Please use Jambord (use QR code) to jot down your ideas.