Artificial Intelligence

Objective of the CRT meeting

Trustworthy Implementation of AI in Clinical Practice

Overall objective

To establish a comprehensive and actionable roadmap for the trustworthy integration of AI into clinical practice within cardiology, ensuring that AI tools are reliable, evidence-based, and contribute to improved patient outcomes. This roadmap will be developed through collaboration between the medical industry, regulators, notified bodies, professional societies, policy makers, payers and patient representatives.

Specific objectives

1. Developing an AI Implementation Roadmap:
   • Identify key stakeholders involved in AI implementation.
   • Outline the phases of implementation, from pilot projects to full integration.
   • Establish milestones and timelines for each phase.
• Objective: To create a clear, step-by-step plan for the adoption of AI technologies in clinical practice.
• Discussion points:
2. Ensuring Data Quality and Integrity:
   • Identify key data sources and types of data required for AI algorithms.
   • Develop strategies to minimise bias in data collection and processing.
   • Discuss data governance and management practices, including patient privacy concerns.
   • Address data harmonisation, common data models, federated learning and synthetic data generation for data sharing and evaluation of AI algorithms.
• Objective: To define standards and protocols for ensuring the quality, consistency, and reliability of data used in AI systems.
• Discussion Points:
3. Creating a Framework for AI Evaluation:
   • Define criteria for evaluating AI algorithms, including accuracy, transparency, and fairness.
   • Discuss the role of independent third-party evaluations and certifications.
   • Explore methods for continuous monitoring and post-market surveillance of AI tools within network of centres.
• Objective: To establish a robust framework for evaluating the safety, efficacy, and ethical considerations of AI tools before and after their deployment in clinical settings.
• Discussion Points:
4. Defining Computable Disease Definitions and Outcomes:
   • Identify diseases and outcomes that require standardised definitions.
   • Harmonisation with existing definitions and deposit in centralised ESC phenotype platform.
   • Discuss how these definitions can be used to support clinical decision-making and research.
• Objective: To develop standardised, computable definitions for diseases and clinical outcomes that can be consistently used across AI tools, guidelines, and clinical trials.
• Discussion Points:
5. Establishing Evidence Requirements for AI in Clinical Trials:
   • Discuss the need for rapid-cycle trials embedded within routine clinical care.
   • Explore short-term outcome definitions that are suitable for AI validation.
   • Identify the balance between rigorous evidence requirements and the need for timely innovation.
   • Discuss the role of HTA.
• Objective: To determine the appropriate levels of evidence needed to validate AI tools in clinical trials, ensuring that these tools are both (cost-)effective and safe for patient care.
• Discussion Points:
6. Engaging with Stakeholders:
   • Facilitate dialogue between cardiologists, AI developers, and patient representatives.
   • Address regulatory and policy challenges and opportunities related to AI in healthcare.
   • Discuss the role of professional societies in guiding the ethical and effective use of AI.
• Objective: To ensure that the perspectives of all relevant stakeholders, including regulators, policy makers, payers, professional societies, patients, notified bodies and industry, are incorporated into the roadmap.
• Discussion Points:
7. Planning Future Collaborations:
   • Set the agenda for the next CRT meetings, focusing on unresolved issues and new developments.
   • Identify working groups or committees to continue work on specific objectives.
   • Discuss potential partnerships or funding opportunities to support the roadmap’s implementation.
• Objective: To outline the next steps and future meetings required to continue progress on the AI implementation roadmap.
• Discussion Points: