The increasing availability of high-quality digital scanners has enabled the generation of large collections of histology images, confocal/multichannel images, and accompanying metadata. However there is a dearth of robust open- source solutions to efficiently visualize, process and manage these ever growing imaging collections. Our goal is to open-source, document and further develop integrative technologies leveraging our experience with the Cancer Digital Slide Archive (CDSA), a tool we have developed to facilitate analysis of data provided by the NCI's the Cancer Genome Atlas (TCGA). This tool is NOT and will not be limited to the analysis of TCGA data, however by working backwards from public data already available we can ensure the informatics technologies developed are scaleable and usable by the cancer community. In our proposal, we will first go through a process of software engineering review to improve the ease of installation to facilitate distribution to other research groups. We have partnered with Kitware for this proposal allowing us to use their 15+ years of experience in building and maintain quality open source software. The rest of the proposal will focus on the testing and integration of new features such as the ability to perform image quantification (e.g. cell counting, cell profiling), image markup and labeling, as well as perform basic group level analysis allowing the correlation of imaging features with user defined variables of interest. As an example, a user may classify an individual slide based on the mean density of nuclei and correlate this imaging parameter with patient survival, or with tumor grade.
The goal of our proposal is to further develop an open source platform for the management, visualization and analysis of large image repositories focused around cancer imaging, specifically digital histology. The size of these image collections ( gigabytes to terabytes ) require new technologies and platform for the efficient analysis and dissemination of these images, as well as linking these images to other relevant information (patient diagnosis, age, gender, etc). Without such tools, this data will remain largely unavailable for better understanding and characterization of tumor biology.
Showing the most recent 10 out of 18 publications
