Pathological glomerulus classification plays a key role in the diagnosis of nephropathy. As the difference between different subcategories is subtle, doctors often refer to slides from different staining methods to make decisions. However, creating correspondence across various stains is labor-intensive, bringing major difficulties in collecting data and training a vision-based algorithm to assist nephropathy diagnosis. This paper provides an alternative solution for integrating multi-stained visual cues for glomerulus classification. Our approach, named generator-to-classifier (G2C), is a two-stage framework. Given an input image from a specified stain, several generators are first applied to estimate its appearances in other staining methods, and a classifier follows to combine these visual cues for decision making. These two stages are optimized in a joint manner. To provide a reasonable initialization for the generators, we train an unpaired image-to-image translation network for each stain, and fine-tune them with the classifier. Since there are no publicly available datasets for glomerulus classification, we collect one by ourselves. Experiments reveal the effectiveness of our approach, including the authenticity of the generated patches so that doctors can hardly distinguish them from the real ones. We also transfer our model to a public dataset for breast cancer classification, and outperform the state-of-the-arts significantly
View on arXiv