Fred Prior

PhD

University of Arkansas for Medical Sciences

Professor

NLST (Learn more about this study)

NLST-114

Jan 21, 2015

QIN Pilot Project: Object Oriented Data Analysis of NLST Results

One of the aims of our Quantitative Imaging Network U01 is to utilize NLST data available from the Cancer Imaging Archive (TCIA) in a pilot project to explore the use of novel graph analysis and statistics tools for identification of candidate multi-parametric biomarkers. The goal is to extract the full set of anonymized trial data from TCIA's clinical trials database and using the Eureka! tool from our collaborators at Emory University, convert this data to a database of graphs. These graphs will be created and stored in a secure database at Washington University (WU) that is hosted in the same protected cloud computing environment as TCIA. Graphs will then be analyzed using a set of graph statistics tools developed at WU and implemented on the WU center for high performance computing cluster. The goal of the project is to identify statistically significant subgraphs that distinguish between participants who developed cancer and those who did not. The hypothesis being that these subgraphs represent candidate biomarkers. This type of graph statistical analysis is one form of Object Oriented Data Analysis and represents a state-of-the-art approach to data mining. The results of this pilot project will be published and the tools (but not the NLST data) will be made available to other NCI-funded, QIN researchers. In the second phase of this project, a subset of CT images will be analyzed using a feature extraction pipeline to characterize lung nodules and the full set of pathology images will be automatically segmented to identify all cellular nuclei which will then also undergo feature extraction and characterization. These quantitative image features will be added to the Phase 1 graphs for a more detailed analysis.