论文标题
Safron：在边境缝线以产生结直肠癌组织学图像
SAFRON: Stitching Across the Frontier for Generating Colorectal Cancer Histology Images
论文作者
论文摘要
合成图像可用于在数据可用性有限的情况下开发和评估深度学习算法。在计算病理学领域，组织学图像的大小和视觉上下文至关重要，通过生成建模综合大型高分辨率图像是一项艰巨的任务。这是由于记忆和计算约束妨碍了大图像的产生。为了应对这一挑战，我们提出了一个新颖的safron（在边境缝线）框架，以从地面真实注释中构建现实的高分辨率组织图像瓷砖，同时保留形态学特征并使用最小的边界伪像。我们表明，在相对较小的图像贴片上训练它后，提出的方法可以生成任意尺寸的逼真的图像图块。我们证明，我们的模型可以在视觉上和特征开始距离上产生高质量的图像。与其他现有方法相比，我们的框架在训练的内存需求以及构建大型高分辨率图像的计算数量方面是有效的。我们还表明，Safron生成的合成数据培训可以显着提高结直肠癌组织学图像中最先进的算法用于腺体分割的性能。使用Safron生成的示例高分辨率图像可在网址上找到：https：//warwick.ac.uk/tialab/safron
Synthetic images can be used for the development and evaluation of deep learning algorithms in the context of limited availability of data. In the field of computational pathology, where histology images are large in size and visual context is crucial, synthesis of large high resolution images via generative modeling is a challenging task. This is due to memory and computational constraints hindering the generation of large images. To address this challenge, we propose a novel SAFRON (Stitching Across the FRONtiers) framework to construct realistic, large high resolution tissue image tiles from ground truth annotations while preserving morphological features and with minimal boundary artifacts. We show that the proposed method can generate realistic image tiles of arbitrarily large size after training it on relatively small image patches. We demonstrate that our model can generate high quality images, both visually and in terms of the Frechet Inception Distance. Compared to other existing approaches, our framework is efficient in terms of the memory requirements for training and also in terms of the number of computations to construct a large high-resolution image. We also show that training on synthetic data generated by SAFRON can significantly boost the performance of a state-of-the-art algorithm for gland segmentation in colorectal cancer histology images. Sample high resolution images generated using SAFRON are available at the URL: https://warwick.ac.uk/TIALab/SAFRON
