Learning General-Purpose Biomedical Volume Representations using Randomized Synthesis
This addresses the challenge of creating general-purpose biomedical volume representations for researchers and practitioners in medical imaging, enabling robust performance across diverse tasks and conditions, though it is incremental as it builds on existing contrastive learning and synthesis techniques.
The paper tackles the problem of limited generalization in volumetric biomedical foundation models due to small and non-diverse public 3D datasets by developing a representation learning method that uses randomized synthesis to anticipate domain shifts, resulting in new state-of-the-art performance in multimodality registration and few-shot segmentation without training on real images.
Current volumetric biomedical foundation models struggle to generalize as public 3D datasets are small and do not cover the broad diversity of medical procedures, conditions, anatomical regions, and imaging protocols. We address this by creating a representation learning method that instead anticipates strong domain shifts at training time itself. We first propose a data engine that synthesizes highly variable training samples that would enable generalization to new biomedical contexts. To then train a single 3D network for any voxel-level task, we develop a contrastive learning method that pretrains the network to be stable against nuisance imaging variation simulated by the data engine, a key inductive bias for generalization. This network's features can be used as robust representations of input images for downstream tasks and its weights provide a strong, dataset-agnostic initialization for finetuning on new datasets. As a result, we set new standards across both multimodality registration and few-shot segmentation, a first for any 3D biomedical vision model, all without (pre-)training on any existing dataset of real images.