MICCAI 2022 Daily – Wednesday

When investigating biological problems of interest, combining data across multiple sites and scanners is necessary to increase statistical power and the breadth of biological variability . However, doing this presents two problems: the harmonization problem and data privacy concerns. The harmonization problem occurs because different scanners give different signals. The same subject acquired on different scanners will look subtly different due to the MRI scanner itself, rather than anything interesting in the person’s biology. When data is combined across scanners, this will increase noise . Data privacy is an issue because medical imaging data is inherently personal information, so sharing this across sites could be a contravention of privacy legislation. “ Our approach is trying to overcome these two barriers to enable us to answer the question of interest while removing scanner effects and protecting individual privacy, ” Nicola tells us. “ The idea is that you should be able to take data from different hospital sites and learn from it without ever having to move it. This method should allow us to combine data across different institutions. ” The paper proposes a federated framework in which the data doesn’t move. However, for harmonization, data must be compared between sites to be able to remove the scanner information. To solve this, it proposes domain adaptation in a federated setting while reducing the amount of shared information . FedHarmony: Unlearning Scanner Bias with Distributed Data 14 DAILY MICCAI Wednesday Poster Presentation Nicola Dinsdale is a postdoctoral research associate at the University of Oxford under the supervision of Ana Namburete. Her paper proposes a novel solution to the MRI data harmonization problem in a distributed setting. She speaks to us ahead of her poster this afternoon.