Ink removal from histopathology whole slide images by combining\n classification, detection and image generation models

Histopathology slides are routinely marked by pathologists using permanent\nink markers that should not be removed as they form part of the medical record.\nOften tumour regions are marked up for the purpose of highlighting features or\nother downstream processing such an gene sequencing. Once digitised there is no\nestablished method for removing this information from the whole slide images\nlimiting its usability in research and study. Removal of marker ink from these\nhigh-resolution whole slide images is non-trivial and complex problem as they\ncontaminate different regions and in an inconsistent manner. We propose an\nefficient pipeline using convolution neural networks that results in ink-free\nimages without compromising information and image resolution. Our pipeline\nincludes a sequential classical convolution neural network for accurate\nclassification of contaminated image tiles, a fast region detector and a domain\nadaptive cycle consistent adversarial generative model for restoration of\nforeground pixels. Both quantitative and qualitative results on four different\nwhole slide images show that our approach yields visually coherent ink-free\nwhole slide images.\n