Novel X-Ray Detection for Large Field of View (FOV) Very High Resolution Computed Tomography (CT) Inspection and Evaluation

Not Available We will prototype an original methodology for search and knowledge discovery in electronic databases. The methodology represents a quantum leap with respect to ongoing work on Latent Semantic Indexing (LSI), and is based on the Backus Gilbert method for inversion. This provides the framework for the implementation of a practical and scalable algorithm. We recast the measurement of the similarity between query and document projections in principal component space as a constrained optimization problem. The adoption of alternative optimization techniques brings a dramatic improvement in computational speed. It also provides important conceptual advantages over the unsupervised classification process implied by LSI. The first is a formalism to assess the process of fitting a noisy or imprecise query to the information in a database in a reduced transform space. The ability to introduce constraints in the optimization translates into new and appealing IR functionality. A new metric quantifies performance as a function of the order of relevance of the returned documents. The prototype addresses issues pertaining to scalability, portability and extensibility. It is based on the abstract services and distributed objects paradigm, where arrays of indexing and information retrieval modules reside on different platforms and communicate across an intranet or the internet.