| Abstract |
We have developed a United framework that integrates three self-supervised learning (SSL) ingredients (discriminative, restorative, and adversarial learning), enabling collaborative learning among the three learning ingredients and yielding three transferable components: a discriminative encoder, a restorative decoder, and an adversary encoder. To leverage this collaboration, we redesigned nine prominent self-supervised methods, including Rotation, Jigsaw, Rubik's Cube, Deep Clustering, TransVW, MoCo, BYOL, PCRL, and Swin UNETR, and augmented each with its missing components in a United framework for 3D medical imaging. However, such a United framework increases model complexity, making 3D pretraining difficult. To overcome this difficulty, we propose stepwise incremental pretraining, a strategy that unifies the pretraining, in which a discriminative encoder is first trained via discriminative learning, the pretrained discriminative encoder is then attached to a restorative decoder, forming a skip-connected encoder-decoder, for further joint discriminative and restorative learning. Last, the pretrained encoder-decoder is associated with an adversarial encoder for final full discriminative, restorative, and adversarial learning. Our extensive experiments demonstrate that the stepwise incremental pretraining stabilizes United models pretraining, resulting in significant performance gains and annotation cost reduction via transfer learning in six target tasks, ranging from classification to segmentation, across diseases, organs, datasets, and modalities. This performance improvement is attributed to the synergy of the three SSL ingredients in our United framework unleashed through stepwise incremental pretraining. Our codes and pretrained models are available at GitHub.com/JLiangLab/StepwisePretraining.
|
