Spectral-spatial filtering for efficient multi-material decomposition in x-ray CT

项目摘要/摘要

PROJECT SUMMARY / ABSTRACT Computed tomography (CT) scans is an extremely important diagnostic tool with ~80 million scans per year with over half of those scans using exogenous contrast agents to enhance visualization of various lesions and to provide functional information. Spectral CT uses data acquisition with different spectral channels to separate individual materials and to provide concentration maps of particular contrast agents. Contrast agent develop- ment is an active research area with many new compounds under investigation including gold, bismuth, gado- linium, xenon, and lanthanide-based contrast agents. Imaging of multiple agents simultaneously in a single ac- quisition has distinct advantages for multiphasic studies and multisystem (e.g., angiography/respiratory or an- giography/gastrointestinal) studies. Similarly, even if agents are estimated individually, flexibility to provide op- timal spectra for those agents is desirable in terms of dose utilization and/or raising low concentration visibility limits. However, current spectral CT systems are limited in their ability to image multiple contrast agents either due to the limited number of spectral channels available (e.g., two in dual-energy systems) or due to their lim- ited flexibility in shaping the x-ray spectrum (e.g., using one, or in dual-source system two x-ray filters). We propose a novel and relatively simple hardware modification applicable to both current diagnostic CT as well preclinical spectral CT prototypes that can either enable or enhance material decomposition capabilities, respectively. The concept borrows from an idea used in color optical imaging where spectral information is en- coded spatially using tiled filters. For x-ray imaging, these filters can be comprised of materials with k-edges in the diagnostic range to effect specific spectral shapes. Data acquisitions using these filters will be sparse in each spectral channel. Ordinarily this would present a challenge for traditional reconstruction and material de- composition; however, model-based material decomposition (MBMD) and compressed sensing allow for direct estimation of material concentration from the projection measurements enabling this unique data acquisition. The proposed joint hardware and software research effort has the following specific aims. Aim 1: Characterize and design spectral-spatial filters. We will develop models for spectral-spatial filters to enable simulation, MBMD reconstruction, and optimized filter design. Aim 2: Construct spectral-spatial filters and implement sparse data acquisition in a CT test bench. We will develop methods for fabrication and integration of spec- tral-spatial filters into a CT test bench. Aim 3: Evaluate spectral-spatial filtered CT in physical material de- composition experiments. Performance of the novel spectral CT designs relative to current methodologies will be assessed including low-dose and low-concentration contrast limits. Successful completion of these aims will establish the feasibility of the spectral-spatial filtering approach as a modification to both standard diagnos- tic CT (enabling spectral imaging with a relatively simple modification) and to existing spectral CT prototype (providing additional spectral shaping capabilities for better dose utilization and lower concentration limits).