A user-friendly tool for cloud-based whole slide image segmentation, with examples from renal histopathology

Abstract

Image-based machine learning tools hold great promise for clinical applications in nephropathology and kidney research. However, the ideal end-users of these computational tools (e.g., pathologists and biological scientists) often face prohibitive challenges in using these tools to their full potential, including the lack of technical expertise, suboptimal user interface, and limited computation power. We have developed Histo-Cloud, a tool for segmentation of whole slide images (WSIs) that has an easy-to-use graphical user interface. This tool runs a state-of-the-art convolutional neural network (CNN) for segmentation of WSIs in the cloud and allows the extraction of features from segmented regions for further analysis. By segmenting glomeruli, interstitial fibrosis and tubular atrophy, and vascular structures from renal and non-renal WSIs, we demonstrate the scalability, best practices for transfer learning, and effects of dataset variability. Finally, we demonstrate an application for animal model research, analyzing glomerular features in murine models of aging, diabetic nephropathy, and HIV associated nephropathy. The ability to access this tool over the internet will facilitate widespread use by computational non-experts. Histo-Cloud is open source and adaptable for segmentation of any histological structure regardless of stain. Histo-Cloud will greatly accelerate and facilitate the generation of datasets for machine learning in the analysis of kidney histology, empowering computationally novice end-users to conduct deep feature analysis of tissue slides.

A tool for federated training of segmentation models on whole slide images

“…This pipeline is deployed in the cloud for easy access for data viewing and annotation by each site's respective cons�tuents. This is a companion work to our recently published Histo-Cloud segmenta�on tool 8 ; it shows the feasibility for training Histo-Cloud in a federated setup. Histo-Cloud is a cloud-based tool for segmentation of WSIs.…”

“…This work is heavily based upon our previously published Histo-Cloud tool 8 , where we modified the DeepLab V3+ architecture 5 to work na�vely on WSIs and developed a series of plugins for running segmentation training and predic�on in the cloud. This work was based on the Digital Slide Archive (DSA) 24 an open source slide viewer and repository developed by Kitware Inc.…”

“…We use the requests python library 39 to send REST calls to the federated workers, which all have the DSA and Histo-Cloud installed and are hos�ng the respec�ve training datasets. 8 . The training job scheduling is handled by the DSA internally using slicer_cli_web, which uses Celery 40 for task queue management, and RabbitMQ 41 as a message broker.…”

“…Training patches with a size of 512x512 pixels were extracted from the training WSIs with various downsampled resolutions. Here we followed the training protocol in the Histo-Cloud work 8 using training patches randomly downsampled to 1, 2, 3, & 4 �mes smaller with respect to the na�ve WSI resolution.…”

“…The matura�on of convolutional neural networks (CNNs) 3 (a specialized subset of deep learning) for the analysis and segmentation of natural images has led to widespread adoption of this technology in the field of computa�onal pathology. CNNs have shown promising results for state of the art computa�onal pathology image analysis tasks including �ssue segmenta�on [4][5][6][7][8] , disease classifica�on [9][10][11][12][13] , and outcome prediction 14,15 . Training these networks is enhanced by access to diverse WSI datasets, as greater data variability is known to enhance model robustness 16 .…”