Which Pixel to Annotate: a Label-Efficient Nuclei Segmentation Framework
This work addresses the inefficiency of labeling all pixels in nuclei segmentation for pathology images, offering a label-efficient solution that could reduce annotation costs in medical imaging.
The paper tackles the problem of reducing annotation workload for nuclei segmentation in pathology images by proposing a framework that selects only a few image patches for labeling, augments training data with synthesized samples, and achieves segmentation in a semi-supervised manner, achieving performance comparable to fully-supervised baselines with less than 5% pixel annotation on some benchmarks.
Recently deep neural networks, which require a large amount of annotated samples, have been widely applied in nuclei instance segmentation of H\&E stained pathology images. However, it is inefficient and unnecessary to label all pixels for a dataset of nuclei images which usually contain similar and redundant patterns. Although unsupervised and semi-supervised learning methods have been studied for nuclei segmentation, very few works have delved into the selective labeling of samples to reduce the workload of annotation. Thus, in this paper, we propose a novel full nuclei segmentation framework that chooses only a few image patches to be annotated, augments the training set from the selected samples, and achieves nuclei segmentation in a semi-supervised manner. In the proposed framework, we first develop a novel consistency-based patch selection method to determine which image patches are the most beneficial to the training. Then we introduce a conditional single-image GAN with a component-wise discriminator, to synthesize more training samples. Lastly, our proposed framework trains an existing segmentation model with the above augmented samples. The experimental results show that our proposed method could obtain the same-level performance as a fully-supervised baseline by annotating less than 5% pixels on some benchmarks.