Analyzing huge pathology images with open source software

Abstract

BackgroundDigital pathology images are increasingly used both for diagnosis and research, because slide scanners are nowadays broadly available and because the quantitative study of these images yields new insights in systems biology. However, such virtual slides build up a technical challenge since the images occupy often several gigabytes and cannot be fully opened in a computer’s memory. Moreover, there is no standard format. Therefore, most common open source tools such as ImageJ fail at treating them, and the others require expensive hardware while still being prohibitively slow.ResultsWe have developed several cross-platform open source software tools to overcome these limitations. The NDPITools provide a way to transform microscopy images initially in the loosely supported NDPI format into one or several standard TIFF files, and to create mosaics (division of huge images into small ones, with or without overlap) in various TIFF and JPEG formats. They can be driven through ImageJ plugins. The LargeTIFFTools achieve similar functionality for huge TIFF images which do not fit into RAM. We test the performance of these tools on several digital slides and compare them, when applicable, to standard software. A statistical study of the cells in a tissue sample from an oligodendroglioma was performed on an average laptop computer to demonstrate the efficiency of the tools.ConclusionsOur open source software enables dealing with huge images with standard software on average computers. They are cross-platform, independent of proprietary libraries and very modular, allowing them to be used in other open source projects. They have excellent performance in terms of execution speed and RAM requirements. They open promising perspectives both to the clinician who wants to study a single slide and to the research team or data centre who do image analysis of many slides on a computer cluster.Virtual slidesThe virtual slide(s) for this article can be found here:http://www.diagnosticpathology.diagnomx.eu/vs/5955513929846272

Tumor Infiltration in Enhancing and Non-Enhancing Parts of Glioblastoma: A Correlation with Histopathology

“…First, images were opened using NDPI Tools[12] or split into smaller-sized images if opening failed and subsequently converted to 8-bit (grey-scale). Cell density was calculated semi-automatically with the ImageJ plugin ITCN in multiple steps: First, up to 8 representative regions of cells were selected.…”

Hepatic elastin content is predictive of adverse outcome in advanced fibrotic liver disease

“…Quantification of elastin immunopositivity was repeated independently at a second centre with the same raw whole‐slide images. Images were split by the use of ndpisplit into tiles of ×5 magnification before the application of a classifier that had been generated by a specialist liver histopathologist using the machine learning weka plugin in fiji . All analysis was undertaken blind to all clinical and histological data.…”

Cholera Toxin Enhances Vaccine-Induced Protection against Mycobacterium Tuberculosis Challenge in Mice

“…Files were provided as .ndpi files, which were converted to .tif files using the ndip2tiff software provided by Dr Christophe Deroulers (NDPITools; http://www.imnc.in2p3.fr/pagesperso/deroulers/software/ndpitools/) [22]. The k means clustering algorithm in MATLAB (MATLAB 7.13.0, The MathWorks Inc., Natick, USA http://www.mathworks.com.au/help/stats/kmeans.html) was used to determine exclusive k means clustering of three colours (white, pink and purple) in the haematoxylin and eosin-stained tissue.…”