tipiX: Rapid Visualization of Large Image Collections

We present a novel approach for fast and effective visualization of large image collections in population studies. The key insight is to collapse inherently high-dimensional imaging data onto an interactive two-dimensional canvas native to a computer screen in a way that enables intuitive browsing of the image data. Increasingly, medical image computing research involves exploring large image sets with high intrinsic dimensionality. This includes three dimensions for each medical volume, and many meta-dimensions such as subject index, modality type in multimodal studies, time in longitudinal studies, or parameter choice in parameter sweep experiments. Current visualization tools generally display one or few 2D slices or 3D renderings at a time, and do not provide a natural way to explore the meta-dimensions. Instead, population statistics are often employed to summarize large cohorts. We propose a novel visualization approach that enables rapid interactive visualization of high dimensional image data, bridging the gap between single-volume viewers and large dataset statistics. Our approach allows users to identify important patterns in the data or anomalies that might otherwise be overlooked. We demonstrate that our platform can be used for quick and effective evaluation and analysis, and we believe it will improve research workflow and facilitate novel method development. Our tool is freely available at http://tipix.csail.mit.edu, where we also provide a video and live demonstration.