FastPathology: An open-source platform for deep learning-based research and decision support in digital pathology

Deep convolutional neural networks (CNNs) are the current state-of-the-art\nfor digital analysis of histopathological images. The large size of whole-slide\nmicroscopy images (WSIs) requires advanced memory handling to read, display and\nprocess these images. There are several open-source platforms for working with\nWSIs, but few support deployment of CNN models. These applications use\nthird-party solutions for inference, making them less user-friendly and\nunsuitable for high-performance image analysis. To make deployment of CNNs\nuser-friendly and feasible on low-end machines, we have developed a new\nplatform, FastPathology, using the FAST framework and C++. It minimizes memory\nusage for reading and processing WSIs, deployment of CNN models, and real-time\ninteractive visualization of results. Runtime experiments were conducted on\nfour different use cases, using different architectures, inference engines,\nhardware configurations and operating systems. Memory usage for reading,\nvisualizing, zooming and panning a WSI were measured, using FastPathology and\nthree existing platforms. FastPathology performed similarly in terms of memory\nto the other C++ based application, while using considerably less than the two\nJava-based platforms. The choice of neural network model, inference engine,\nhardware and processors influenced runtime considerably. Thus, FastPathology\nincludes all steps needed for efficient visualization and processing of WSIs in\na single application, including inference of CNNs with real-time display of the\nresults. Source code, binary releases and test data can be found online on\nGitHub at https://github.com/SINTEFMedtek/FAST-Pathology/.\n