Research

Functional computed tomography (CT) can provide very high spatio-temporal information on the status of organs and lesions. The most complete method for performing this imaging is CT perfusion (CTP), in which the passage of iodinated contrast agent through tissue is continuously imaged for an extended period. The resulting images are then mathematically manipulated to estimate quantitative tissue properties that describe the tissue status. CTP has the potential to allow for personalized treatment in cancer, stroke, cardiac, and many other patients. However, currently CTP results in unacceptably high dose or low-quality images, resulting in CTP currently being of limited use. In this project, we will further develop and optimize a filter to reduce the noise in CTP images acquired at current clinical multi-phase CT doses, allowing both their qualitative interpretation and their quantitative analysis. This filter, denoted the 4D similarity filter, processes the whole CTP image at the same time, using the information present throughout the entire 3D image and the entire acquisition time sequence to reduce the noise present in each voxel. As a result, images that appear to contain almost no information are rendered of diagnostic quality, with quantitative accuracy. The largest step forward in the development that will be performed in this project is the creation of a deep learning-based optimization of the filter to maximize its impact on any CTP image from any part of the anatomy, irrespective of its acquisition and reconstruction parameters. Given its lack of dependence on the acquisition process, this development will also result in the future easy potential translation of the similarity filter to use in perfusion images of other modalities, such as MRI and PET.

4D similarity filter in CTP of the pancreas (left) without filtering, and (right) with filtering. (top) 4D CTA at 0.5 mm section, and (bottom) corresponding blood volume maps.

Researchers:

Sjoerd Tunissen

Koen Michielsen

Ioannis Sechopoulos

Luuk Oostveen

Key Publications:

• Sjoerd A. M. Tunissen, Luuk J. Oostveen, Nikita Moriakov, Jonas Teuwen, Koen Michielsen, Ewoud J. Smit, Ioannis Sechopoulos. "Development, validation, and simplification of a scanner‐specific CT simulator", 2023. Abstract. DOI.
• Sjoerd A. M. Tunissen, Nikita Moriakov, Mikhail Mikerov, Ewoud J. Smit, Ioannis Sechopoulos, Jonas Teuwen. "Deep learning‐based low‐dose CT simulator for non‐linear reconstruction methods", 2024. Abstract. DOI.