Source Identification: A Self-Supervision Task for Dense Prediction
This addresses the bottleneck of annotation costs in data-driven methods for medical image analysis, though it appears incremental as it builds on existing self-supervision paradigms.
The paper tackles the problem of self-supervised representation learning for dense prediction tasks by proposing a new source identification task that reconstructs original images from fused synthetic ones, showing it outperforms traditional self-supervision methods on medical image segmentation tasks like brain tumor and white matter hyperintensities segmentation.
The paradigm of self-supervision focuses on representation learning from raw data without the need of labor-consuming annotations, which is the main bottleneck of current data-driven methods. Self-supervision tasks are often used to pre-train a neural network with a large amount of unlabeled data and extract generic features of the dataset. The learned model is likely to contain useful information which can be transferred to the downstream main task and improve performance compared to random parameter initialization. In this paper, we propose a new self-supervision task called source identification (SI), which is inspired by the classic blind source separation problem. Synthetic images are generated by fusing multiple source images and the network's task is to reconstruct the original images, given the fused images. A proper understanding of the image content is required to successfully solve the task. We validate our method on two medical image segmentation tasks: brain tumor segmentation and white matter hyperintensities segmentation. The results show that the proposed SI task outperforms traditional self-supervision tasks for dense predictions including inpainting, pixel shuffling, intensity shift, and super-resolution. Among variations of the SI task fusing images of different types, fusing images from different patients performs best.