Balancing innovation, privacy, and trust in healthcare

Health data hold enormous potential for improving medical research, healthcare delivery, and regulation. At the same time, they are among the most sensitive data we generate. Privacy concerns continue to limit their use - not only in public research, but also in the rapidly growing health-tech sector. How, then, can we enable data-driven innovation without compromising individual rights?

In his lecture Lessons from Synthetic Health Data Generation: Fidelity, Privacy, Augmentation & Time, Prof. Allan Tucker addresses this challenge by focusing on synthetic health data. Synthetic data aim to reproduce the statistical structure of real patient data while avoiding direct privacy risks. Drawing on a long-term collaboration between Brunel University London and the UK Medicines and Healthcare products Regulatory Agency, Tucker presents how high-fidelity synthetic datasets can be generated using probabilistic models with complex latent variable structures.

This work has already resulted in publicly available synthetic datasets for diseases such as COVID-19 and cardiovascular conditions, enabling state-of-the-art AI research without exposing sensitive individual records. However, the talk also highlights critical limitations. Even when based on comprehensive national datasets, synthetic data can reproduce or amplify bias. In addition, healthcare data evolve over time, leading to concept drift - situations where new data no longer align with existing models, raising important questions for regulation, validation, and long-term deployment.

Tucker will discuss practical lessons from this work: how bias can be detected and mitigated, how synthetic data can better reflect the true underlying population, and how AI models can be adapted to healthcare data that change over time. These insights are highly relevant for researchers developing medical AI, policymakers and regulators concerned with safety and accountability, students entering the field, and a broader public interested in the responsible use of AI in health.

Event details

📅 Date: 4 February 2026

⏰ Time: 9:30–10:30 AM

💻 Online participation via Zoom:

tu-dortmund.zoom.us/j/92778021899

Meeting ID: 927 7802 1899

Code: 793838

AI Colloquium
The AI Colloquium is a series of lectures dedicated to cutting-edge research in the field of machine learning and artificial intelligence, coorganized by the Research Center Trustworthy Data Science and Security (RC Trust), the Lamarr Institute for Machine Learning and Artificial Intelligence (Lamarr Institute), and the Center for Data Science & Simulation at TU Dortmund University (DoDas).

Graduate School
The lecture is also part of the Graduate School programme. The Graduate School offers early opportunities for independent research, close supervision, and access to a strong interdisciplinary and international network. With flexible structures and individual guidance, it provides an inspiring environment for developing original research directions.

Participation is open to all interested audiences.