This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. Primary support for the subproject and the subproject's principal investigator may have been provided by other sources, including other NIH sources. The Total Cost listed for the subproject likely represents the estimated amount of Center infrastructure utilized by the subproject, not direct funding provided by the NCRR grant to the subproject or subproject staff. We will research algorithms for the unsupervised analysis of medical volume images based on texture properties. Texture properties of volume data are defined based on spatial frequencies and we use a bank of Gabor filters to determine these textural properties. Each Gabor filter in the bank is tuned to detect patterns of a specific frequency and orientation when convolved with a medical volume. The response image is added to the feature vector and once complete, the feature vector is passed into a classification/segmentation algorithm. The total number of Gabor filters needed is determined based on the size of the input volume. The feature vector grows exponentially with increasing volume because larger volumes require more Gabor filters. The sequential implementation for the unsupervised analysis algorithm was accurate, but resource requirements for the algorithm limits the size of input volumes. We have developed a parallel approach for processing larger medical volume images by efficiently managing the resources of multiple nodes on a cluster. Our goal is to use teragrid resources to analyze larger images and to gain better accuracy by increasing the size of our feature vector.
Showing the most recent 10 out of 292 publications
