Statistical modeling and path-based iterative reconstruction for X-ray computed tomography
Abstract/Contents
- Abstract
- X-ray computed tomography (CT) and tomosynthesis systems have proven to be indispensable components in medical diagnosis and treatment. My research is to develop advanced image reconstruction and processing algorithms for the CT and tomosynthesis systems. Streak artifacts caused by metal objects such as dental fillings, surgical instruments, and orthopedic hardware may obscure important diagnostic information in X-ray computed tomography (CT) images. To improve the image quality, we proposed to complete the missing kilovoltage (kV) projection data with selectively acquired megavoltage (MV) data that do not suffer from photon starvation. We developed two statistical image reconstruction methods, dual-energy penalized weighted least squares and polychromatic maximum likelihood, for combining kV and selective MV data. Cramer-Rao Lower Bound for Compound Poisson was studied to revise the statistical model and minimize radiation dose. Numerical simulations and phantom studies have shown that the combined kV/MV imaging systems enable a better delineation of structures of interest in CT images for patients with metal objects. The x-ray tube on the CT system produces a wide x-ray spectrum. Polychromatic statistical CT reconstruction is desired for more accurate quantitative measurement of the chemical composition and density of the tissue. Polychromatic statistical reconstruction algorithms usually have very high computational demands due to complicated optimization frameworks and the large number of spectrum bins. We proposed a spectrum information compression method and a new optimization framework to significantly reduce the computational cost in reconstructions. The new algorithm applies to multi-material beam hardening correction, adaptive exposure control, and spectral imaging. Model-based iterative reconstruction (MBIR) techniques have demonstrated many advantages in X-ray CT image reconstruction. The MBIR approach is often modeled as a convex optimization problem including a data fitting function and a penalty function. The tuning parameter value that regulates the strength of the penalty function is critical for achieving good reconstruction results but is difficult to choose. We have developed two path seeking algorithms that are capable of generating a path of MBIR images with different strengths of the penalty function. The errors of the proposed path seeking algorithms are reasonably small throughout the entire reconstruction path. With the efficient path seeking algorithm, we suggested a path-based iterative reconstruction (PBIR) to obtain complete information from the scanned data and reconstruction model. Additionally, we have developed a convolution-based blur-and-add model for digital tomosynthesis systems that can be used in efficient system analysis, task-dependent optimization, and filter design. We also proposed a computationally practical algorithm to simulate and subtract out-of-plane artifacts in tomosynthesis images using patient-specific prior CT volumes.
Description
| Type of resource | text |
|---|---|
| Form | electronic; electronic resource; remote |
| Extent | 1 online resource. |
| Publication date | 2015 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Associated with | Wu, Meng |
|---|---|
| Associated with | Stanford University, Department of Electrical Engineering. |
| Primary advisor | Fahrig, Rebecca |
| Primary advisor | Pauly, John (John M.) |
| Thesis advisor | Fahrig, Rebecca |
| Thesis advisor | Pauly, John (John M.) |
| Thesis advisor | Pelc, Norbert J |
| Advisor | Pelc, Norbert J |
Subjects
| Genre | Theses |
|---|
Bibliographic information
| Statement of responsibility | Meng Wu. |
|---|---|
| Note | Submitted to the Department of Electrical Engineering. |
| Thesis | Thesis (Ph.D.)--Stanford University, 2015. |
| Location | electronic resource |
Access conditions
- Copyright
- © 2015 by Meng Wu
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...