In this session, organised by TRUSTroke project, we will explore the crucial aspects of Artificial Intelligence (AI) that make predictions in stroke care both understandable and reliable. We will discuss how AI models can be designed and evaluated to ensure transparency and trustworthiness in their predictions, ultimately enhancing patient care and medical decision-making. Finally, we will delve into the ethical issues inherent in the process of building, validating, deploying and making decisions through algorithmic systems.

Agenda

Introduction to AI in Healthcare: What is AI/ML/DL? Importance of data in healthcare AI applications.

Types of AI Problems in Stroke Prediction: Classification, Regression, Forecasting, etc. in the context of stroke prediction.

Evaluation of AI Models: Key metrics for evaluating AI models in the context of healthcare, Explanation of train/test splits and their importance in model validation.

Trustroke AI Algorithms: Overview of AI algorithms planned for the project (Tree-based models & Forecasting models)

Ethical Considerations I: Explainability, Techniques for making AI models interpretable (LIME, SHAP, etc.)

Ethical Considerations II: Transparency, Fairness & Empowerment of stakeholders

Q&A