Recently, we held another of our webinar and were delighted to welcome Dr Ira Ktena from Google DeepMind to explore some discussions around some discussions around understanding how we can use generative models in healthcare to address underrepresentation of data and improve performance of machine learning models.. Thank you to all who managed to attend the webinar series and made it an interactive platform with lots of questions and ideas discussed. You can catch the full video recording below 👇
The talk was on "Diffusion models for medical imaging: the path to fairer, more robust and private models". Dr Ira discussed the use of generative models to address the challenge of domain generalisation in healthcare machine learning. She highlighted that model performance can suffer in real-world scenarios due to disparities between deployment data and training data, often stemming from underrepresented groups or conditions during model development. It is suggested that generative models can offer a solution by generating synthetic examples that help bridge these gaps. Dr Ira emphasised the models' ability to learn realistic data augmentations efficiently. Additionally, her talk explored the application of differential privacy to protect these synthetic images from privacy risks, all while maintaining their usability for downstream tasks
Don't want to miss out on any future Statistical Methods theme activities?
Stay up to date by joining the #theme-statistical-methods channel on our Slack workspace.