The validation and optimization of CT perfusion for stroke assessment
Abstract/Contents
- Abstract
- Each year, approximately 800,000 people in the United States have a stroke. Early diagnosis and tailored treatment are critical for optimal patient outcomes. Physicians utilize cerebral perfusion maps (e.g., cerebral blood flow, cerebral blood volume, transit time) derived from dynamic magnetic resonance (MR) or computed tomography (CT) scans of the brain to prescribe the optimal plan of care for stroke patients. However, noise and variability in scanning techniques and post-processing software can impact perfusion maps. To determine which techniques are acceptable for clinical care, it is important to validate the accuracy and reproducibility of the perfusion maps. Validation using clinical data is challenging due to the lack of substantial clinical data sets in which the pathology is well known. Furthermore, it is impractical to scan patients multiple times with different scanning techniques to optimize protocols. In contrast, simulated data from a realistic digital phantom of the cerebral perfusion in acute stroke patients would enable studies to validate and optimize the scanning and post-processing techniques. This work presents a complete framework to simulate and evaluate CT perfusion studies for stroke assessment. To generate realistic CT perfusion test data, the framework includes (1) a digital brain phantom composed of NURBS surfaces whose material properties can be dynamically set to model spatially varying dynamics; (2) a model of contrast agent enhancement in the brain starting from first principles by using the phantom geometry and tissue perfusion parameters; (3) an efficient dynamic CT simulator that reduces simulation times by 95% compared to traditional simulation methods; and (4) an accurate method to insert noise in the image domain with computation time on the order of seconds, rather than minutes, to generate unique noise realizations. We use this simulated data to generate perfusion maps with four perfusion post-processing software packages. We first perform a visual comparison against the known underlying ground truth and propose a method to quantitatively assess the accuracy of computer-generated perfusion maps. We segment the brain tissue and plot the computer-generated perfusion map pixel values versus the ground truth pixel values, perform a linear fit, and calculate the r-squared value. Through this analysis, we observe differences in r-squared values achieved for each package and each perfusion map, with dependence on dose and smoothing. This comprehensive framework and quantitative assessment, therefore, enables a robust assessment of CT perfusion techniques and post-processing methods for stroke imaging. Ultimately, this has the potential to improve diagnostic tools for stroke patients and increase physicians' ability to prescribe a plan of care driven by quantitative data
Description
| Type of resource | text |
|---|---|
| Form | electronic resource; remote; computer; online resource |
| Extent | 1 online resource |
| Place | California |
| Place | [Stanford, California] |
| Publisher | [Stanford University] |
| Copyright date | 2020; ©2020 |
| Publication date | 2020; 2020 |
| Issuance | monographic |
| Language | English |
Creators/Contributors
| Author | Divel, Sarah Elaine |
|---|---|
| Degree supervisor | Pelc, Norbert J |
| Thesis advisor | Pelc, Norbert J |
| Thesis advisor | Lansberg, Maarten, 1971- |
| Thesis advisor | Nishimura, Dwight George |
| Degree committee member | Lansberg, Maarten, 1971- |
| Degree committee member | Nishimura, Dwight George |
| Associated with | Stanford University, Department of Electrical Engineering |
Subjects
| Genre | Theses |
|---|---|
| Genre | Text |
Bibliographic information
| Statement of responsibility | Sarah Elaine Divel |
|---|---|
| Note | Submitted to the Department of Electrical Engineering |
| Thesis | Thesis Ph.D. Stanford University 2020 |
| Location | electronic resource |
Access conditions
- Copyright
- © 2020 by Sarah Elaine Divel
- 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...