Low dose CT perfusion: filtering the way to the clinic
Aim and outline of thesis
The aim of this thesis was to characterize, validate, and, if possible, improve the 4DSF for its application in CT perfusion imaging. Our research efforts are divided into two distinct parts. The first part focuses on the development of CT image simulators that can generate realistic test data. The second part focuses on the detailed characterization of the 4DSF, exploring its behavior and limitations, ultimately leading to the development and evaluation of strategies for improving its performance. In the first part, two tools for CT simulations are developed. Chapter 2 presents a scanner-specific CT simulator capable of generating synthetic CT scans from digital phantoms. On top of this, we validated simplifications of the simulation to reduce computation time. The second simulation tool is presented in Chapter 3 and focuses on the simulation of low dose CT images from clinical dose patient CT images when knowledge and access on the CT system and reconstruction are unavailable. The method circumvents the use of projection domain data, which is often hard to obtain, by working directly in image domain. Therefore, the method will also work for non-analytical methods. The second part of this thesis is focused on the analysis and optimization of the 4DSF. In Chapter 4 we present a digital phantom study in which the 4DSF is characterized, and its use in combination with the TIPS filter is investigated. Finally, a suggestion on a possible implementation for 4D liver CT is given, focusing on lesion detection. Chapter 5 introduces a modified version of the 4DSF for GPU implementation. Since GPUs are much faster than CPUs, the use of the former is intriguing to accelerate the 4DSF. Chapter 6 presents another modified version of the 4DSF tailored for stroke imaging. By incorporating the unique perfusion characteristics of ischemic brain tissue, which receives blood later than healthy brain tissue, this modified version enhances the quantitative accuracy of the functional maps. The modified version is validated using a phantom study and patient cases to demonstrate its potential.