Faster Reinforcement Learning for AI-based Diagnostic Medical Image Generation

MSc assignment

Artificial neural network models based on reinforcement learning [1, 2] have been increasingly used in diagnostic medical imaging and image processing questions. [3] These models do not require a large amount of images, and instead feature an “agent” that can learn from a randomly presented sequence of images through a goal-oriented process. However, this training scheme is faster when exemplary cases are presented adaptively in a meaningful order [4].

We recently developed a generalizable and interpretable model to rank diagnostic difficulty of medical images. This assignment focuses on this model’s extension to implement an adaptive reinforcement learning method for generative models designed to create new artificial “radiographic” images. In particular:

The goal of the masters project is to set up an adaptive reinforcement learning model to facilitate training of diffusion models to generate artificial medical radiologic images.

The project is suitable for master students from Biomedical Engineering and Electrical Engineering. Contact Can Ozan Tan if you are interested in this assignment or would like to have more information (


[1] Sutton RS, Barto AG (1987) A temporal-difference model of classical conditioning. In: Proceedings of the Ninth Conference of the Cognitive Science Society. pp 355–378
[2] Sutton RS, Barto AG (1981) Toward a modern theory of adaptive networks: Expectation and prediction. Psychol Rev 88:135–170.
[3] Hu M, Zhang J, Matkovic L, et al (2023) Reinforcement learning in medical image analysis: Concepts, applications, challenges, and future directions. J Appl Clin Med Phys 24:1–21.
[4] Winkel DJ, Brantner P, Lutz J, et al (2020) Gamification of electronic learning in radiology education to improve diagnostic confidence and reduce error rates. Am J Roentgenol 214:618–623.