Registration is open - make plans to attend
Get pricing and details
>
Conference 12037 > Paper 12037-3
Paper 12037-3

ST-FL: style transfer preprocessing in federated learning for COVID-19 segmentation

In person: 22 February 2022 • 2:40 PM - 3:00 PM PST

Abstract

Chest Computational Tomography (CT) scans present low cost, speed and objectivity for COVID-19 diagnosis and deep learning methods have shown great promise in assisting the analysis and interpretation of these images. Most hospitals or countries can train their own models using in-house data, however empirical evidence shows that those models perform poorly when tested on new unseen cases, surfacing the need for coordinated global collaboration. Due to privacy regulations, medical data sharing between hospitals and nations is extremely difficult. We propose a GAN-augmented federated learning model, dubbed ST-FL (Style Transfer Federated Learning), for COVID-19 image segmentation. Federated learning (FL) permits a centralised model to be learned in a secure manner from heterogeneous datasets located in disparate private data silos. We demonstrate that the widely varying data quality on FL client nodes leads to a sub-optimal centralised FL model for COVID-19 chest CT image segmentation. ST-FL is a novel FL framework that is robust in the face of highly variable data quality at client nodes. The robustness is achieved by a denoising CycleGAN model at each client of the federation that maps arbitrary quality images into the same target quality, counteracting the severe data variability evident in real-world FL use-cases. Each client is provided with the target style, which is the same for all clients, and trains their own denoiser. Our qualitative and quantitative results suggest that this FL model performs comparably to, and in some cases better than, a model that has centralised access to all the training data.

Presenter

Antonios Georgiadis
JP Morgan Chase (United Kingdom)
Antonios Georgiadis is a Senior Machine Learning (ML) Researcher at the CTO Applied Research team of J.P. Morgan Chase. In his current role, he specializes in the topics of distributed ML (e.g. Federated Learning, privacy-preserving A.I.), Computer Vision (e.g. GAN-augmentation, advanced deep feature representations) and ML on software code, with a number of successful patents on these fields. Prior to this, he worked extensively in electronic communications surveillance for Credit Suisse, delivering multi-award winning (Financial Times PWC, FCA, British Computing Society etc.) solutions for the firm. Antonios is passionate about the application of advanced A.I. techniques in fields with a direct positive impact to society, such as medical imaging, environmental preservation and education.
Presenter/Author
Antonios Georgiadis
JP Morgan Chase (United Kingdom)
Author
JP Morgan Chase (United Kingdom), Univ. of Cambridge (United Kingdom)
Author
JP Morgan Chase (United Kingdom)
Author
JP Morgan Chase (United Kingdom)
Author
Rob Otter
JP Morgan Chase (United Kingdom)