Texture-based 3D brain imaging

Abstract

Different modalities in biomedical imaging, like CT, MRI and PET scanners,, provide detailed cross-sectional views of the human anatomy. The imagery obtained from these scanning devices are typically large-scale data sets whose sizes vary from several hundred megabytes to about one hundred gigabytes, making them impossible to be stored on a regular local hard drive. San Diego Supercomputer Center (SDSC) maintains a high-performance storage system (HPSS) where these large-scale data sets can be stored. Members of the National Partnership for Advanced Computational Infrastructure (NPACI) have implemented a Scalable Visualization Toolkit (Vistools), which is used to access the data sets stored on HPSS and also to develop different applications on top of the toolkit. 2D cross-sectional images are extracted from the data sets stored oft HPSS using Vistools, and these 2D cross-sections are then transformed into smaller hierarchical representations using a wavelet transformation. This makes it easier to transmit them over the network and allows for progressive image refinement. The transmitted 2D cross-sections are then transformed and reconstructed into a 3D volume. The 3D reconstruction has been implemented using texture-mapping functions of Java3D. Sub-volumes that represent a region of interest are transmitted and rendered at a higher resolution than the rest of the data set.