Robust image-to-image translation tool for fibrotic quantification of whole-slide images

Fibrosis is a biological phenomenon characterized by the formation of excessive fibrous connective tissue, which can be visualized using whole-slide image (WSI) scanning techniques. However, the quantification of fibrosis in WSIs is presently a subjective, labour-intensive and time-consuming process. Accurate fibrosis quantifications are of vital importance to pathologists and researchers alike, as high fibrotic content has been linked to a variety of diseases in both animals and humans, and fibrotic prevention or reversal have become major endpoints in clinical trials. We present a novel, fully-automated software tool capable of performing image-to- Image (I2I) translation of picrosirius red (PSR) stained murine WSIs to a translated counterpart with all non-fibrotic pixels set to black. A dataset consisting of 32,652 PSR-stained source images paired with their manually translated counterpart was used to train a conditional conditional generative adverserial network. The source images consist of murine diaphragm, liver, and tibialis anterior sections, varying in lighting and staining conditions. Based on an extensive architecture search, the final machine learning model was identified and named Stacked u-nets with Extended Range Connections Generative Adverserial Network (SERGAN), which once trained on the supervised dataset was capable of generating translations more similar to the ground truth images (obtaining a test set mIOU=0.934) than previous I2I methods, such as Pix2Pix or u-net. After translation, the software tool calculates the collagen proportionate area of the WSI by dividing the number of fibrotic pixels by the total number of tissue pixels found during a separate tissue detection step. Fibrosis quantifications of source images were compared with biochemical assay results (quantitative PCR or RNA-Seq) from the same specimen where data was available, and SERGAN quantifications were found to have significant correlations with several known biomarkers of fibrosis (hydroxyproline concentration [rₛ = 0:52] and periostin gene expression levels [rₛ = 0:24]), whereas the manually translated ground truth quantifications did not. Overall, the final fully-automatic software tool performs multi-organ image segmentation of fibrotic pixels more accurately than other known I2I frameworks, and demonstrates potential usage for preclinical applications.Applied Science, Faculty ofBiomedical Engineering, School ofGraduat