Standalone or Continuous Learning Modules
Building Trustworthy Medical Imaging AI
Medical imaging AI relies on more than algorithms. Designed for professionals and students, this course explores how data quality, annotation, variability, governance, uncertainty, documentation, and validation shape robust AI development and support informed, defensible R&D decisions.
Praktische info:
12
uur
Online
Engels
Doelgroep: Medical imaging AI professionals, R&D teams, researchers, and graduate students.
Leertrajecten
Inschrijven?
- Voorwaarden: Basic understanding of AI or machine learning concepts. No prior expertise in medical imaging AI is required.
- Prijs: Price on request
georganiseerd door:
Building Trustworthy Medical Imaging AI
Medical imaging AI relies on more than algorithms. This course explores how data quality, annotation, variability, governance, uncertainty, documentation, and validation shape robust AI development and support informed, defensible R&D decisions.
Key topics
- Framing Medical AI Projects: Aligning clinical needs, intended use, technical feasibility, and R&D objectives.
- Understanding & Qualifying Medical Data: Building the data understanding required for robust R&D decisions.
- Bias in Medical AI: Identifying how data and development choices can introduce or amplify bias.
- Ground Truth & Annotation Quality: Understanding label validity, disagreement, and uncertainty.
- Model Reliability & Uncertainty: Assessing model confidence, uncertainty, and operational boundaries.
- Robust & Defensible Validation: Designing evaluation strategies that reveal performance limits and support defensible conclusions.
- Safe-by-Design Medical AI: Anticipating failure conditions and integrating safeguards into AI system design.
- Governance & Defensible Documentation: Building traceability and documentation that support robust R&D and regulatory requirements.
Learning objectives
Through these topics, participants develop the ability to:
- frame medical AI projects by connecting clinical needs, intended use, technical feasibility, and R&D objectives;
- critically understand and qualify medical imaging data for R&D;
- identify potential sources of bias and understand how development choices can amplify them;
- assess annotation quality, label validity, disagreement, and uncertainty;
- reason about model reliability, confidence, uncertainty, and operational boundaries;
- design and critically assess validation strategies that reveal performance limitations;
- anticipate failure conditions and consider safeguards throughout AI system development;
- strengthen traceability and documentation to support robust and defensible R&D decisions.
Why it matters
Trustworthy medical imaging AI depends on the quality of the reasoning and evidence behind the system, from project framing and data understanding to validation and documentation. These capabilities help strengthen R&D decisions, reveal limitations earlier, and support more robust and defensible AI development.
Lesgever/spreker
I am the founder of VeraDP, a training, advisory, and research practice specializing in medical imaging AI. Drawing on a PhD in Computer Vision and over 10 years of applied MedTech experience, I help organizations build defensible R&D, with a focus on data readiness for AI and risk analysis, experiment and evaluation design, state-of-the-art positioning, and technical assessment. My work also includes skill transfer through specialized training and masterclasses, translating research and technical expertise into actionable knowledge. My current research explores how data understanding and qualification can better inform R&D decisions.
