Motion measurement algorithms for MARS imaging

Abstract

The goal of the MARS molecular imaging team is to advance medicine by researching, developing, and commercialising spectral CT systems. This thesis presents the work I performed to facilitate live imaging with MARS scanners. This aim was achieved by developing a gating algorithm, designing and developing a mouse holder, and creating a motorised motion phantom. My gating algorithms will contribute to improving the image quality of human data obtained by human-scale MARS scanners.

I contributed to the design and development of a mouse holder with a temperature regulating system that is compatible with MARS scanners for the purpose of live animal imaging. This holder design provides simple animal handling, secure positioning, anaesthesia delivery, regulated temperature control, and physiological monitoring.

I developed a post-acquisition automatic gating method based on the acquired scan data over time. This method is capable of identifying various movement phases to sort the acquired exposure images into temporal bins. To reduce the undersampling noise due to gating, a weight-based reconstruction algorithm was introduced and implemented. Instead of binning the data, this method employed all images for the reconstruction of specific time points by assigning a weight to each. The result of applying this method showed that it can improve the undersampling artefacts compared to the temporal binning method.

To evaluate the gating method, a motorised motion phantom was designed and manufactured. The motion phantom could be programmed to produce periodic signals with a similar frequency and amplitude to that of a mouse or human breathing. The quantitative measurements showed that gating can reduce motion artefacts and blurring by 50% with a 1mm amplitude and 26% for a 5mm amplitude.

The effect of motion on the material decomposition process in MARS imaging systems was investigated. Known contrast agents were added to the motion phantom and then scanned with movements with the amplitude of 1 to 5 mm. No clear trend between the motion amplitude and the material decomposition accuracy was observed. The gated images had lower SNR compared to the non-gated data, resulting in more misidentified voxels. This suggests that noise properties are more important than motion blur.

In summary, the research documented in this thesis facilitates live imaging in MARS scanners in the future.