Potential Supervisor: Tim Kilgus
Level: Bachelor / Master / both
Strategic Orientation: AI and Analytics

Background

The U.S. Food and Drug Agency does allow Software as a Medical Device, specifically Artificial Intelligence-Enabled Medical Devices. The authorization as such enables marketing in the United States, but also other countries globally. At the forefront of AI in medicine is the field of radiology. Of the 1245 authorized AI-enabled medical devices, 956 are in the Radiology panel.

As part of a collaboration with the Department of Radiology at the University of Cambridge, we aim to better understand the current state of authorized software. Specifically, we are interested in the technical (training data, ML-algorithms), and organizational (organizations, platforms) aspects related to the development and clinical implementation.

The list of authorized AI-enabled medical devices is publicly available.

Research Directions

• AI Landscape in Radiology
Radiology is at the forefront of AI adoption in medicine. Analyzing the spectrum of FDA-authorized AI-enabled medical devices can reveal dominant imaging modalities, algorithmic approaches, and clinical application areas. Such an overview helps to map technological focus areas and identify trends across disease domains.

• Organizational Ecosystems and Innovation
The development of AI tools in radiology involves diverse actors—from startups to multinational medtech firms. Exploring these organizational networks and partnerships can shed light on how innovation emerges and diffuses across the medical imaging industry.

• Data and Algorithm Transparency
AI systems in radiology rely heavily on the quality and representativeness of training data. Examining how companies disclose information about datasets, validation methods, and performance metrics provides insights into current transparency practices and potential biases.

References

Lebovitz, S., Levina, N., & Lifshitz-Assa, H. (2021). Is AI Ground Truth Really True? The Dangers of Training and Evaluating AI Tools Based on Experts’ Know-What. MIS Quarterly, 45(3), 1501–1526. https://doi.org/10.25300/MISQ/2021/16564

Lawrence, R., Dodsworth, E., Massou, E., Sherlaw-Johnson, C., Ramsay, A., Walton, H., O’Regan, T., Gleeson, F., Crellin, N., Herbert, K., Ng, P. L., Elphinstone, H., Mehta, R., Lloyd, J., Halliday, A., Morris, S., & Fulop, N. (2025). Artificial Intelligence for diagnostics in radiology practice: A rapid systematic scoping review. The Royal College of Radiologists Open, 3, 100218. https://doi.org/10.1016/j.rcro.2024.100218

Jacobides, M. G. (2024). Externalities and complementarities in platforms and ecosystems: From structural solutions to endogenous failures. Research Policy.