CellVisualizer : exploring hierarchical, multi-dimensional data with applications to high-throughput microscopy

Abstract

In this thesis, we present a system for visualizing hierarchical, multi-dimensional, memory-intensive datasets. Specifically, we designed an interactive system to visualize data collected by high-throughput microscopy and processed by CellProfiler, an open-source system jointly developed by researchers at MIT CSAIL and the White-head Institute. A typical high-throughput microscopy experiment produces thousands of images, with thousands of objects in each image. CellProfiler then measures hundreds of features for each cell, nuclei, and cytoplasm. In contrast to previously demonstrated visualization software, our system visualizes datasets that are on the order of hundreds of gigabytes, datasets too large to store in physical memory. We also implement tools to link the dataset to available resources such as online genetic databases and the actual images acquired by the microscope. Finally, we demonstrate how the system was used to highlight interesting genes for more detailed analysis in real biological studies.