Whitehead, Alexander Charles; (2025) Improved Quantification for Respiratory Gated PET/CT: Data-Driven Algorithms for Respiratory Motion Correction in PET/CT. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Whitehead_10212970_thesis.pdf Download (31MB) | Preview

Abstract

Respiratory motion is a significant general problem in Positron Emission Tomography (PET) imaging, affecting both image quality and quantitative accuracy. Respiratory motion not only blurs lesions and other anatomical features in the lungs, but also complicates the application of attenuation correction to the acquired PET data. While respiratory gating can help reduce motion artefacts, it extends acquisition time and reduces effective counts. A further challenge arises from the temporal mismatch between the attenuation map and the emission scan, as patients are typically instructed to hold their breath during the Computed Tomography (CT) acquisition but breathe freely during the PET scan of a combined PET/CT. As a result, the static Attenuation Map (μ-Map) often does not correspond to any specific respiratory phase of the PET data, introducing misalignment that degrades both qualitative and quantitative image accuracy. Motion correction methods exist which incorporate registration in order to attempt to improve upon non-motion corrected results. Often these methods involve separating the PET data into bins, where the respiratory motion is minimal within each bin. However, because of the high level of noise and low spatial resolution of PET data, few bins are often used, leading to a not insignificant amount of respiratory motion still being present in the resultant images. Additionally, this approach does not solve the mismatch of the μ-Map. Logically, the bin closest to the μ-Map could be used as the reference bin for registration but it is not guaranteed that this bin will be close to the position of the μ-Map and as such artefacts will remain in the final image. Furthermore, registration fails when used on dynamic PET data, where the signal from the aorta, at early time points, often leads to mis-registration. More complex motion correction methods exist, however, these methods, in general, tend to be more resource intensive in both the sense of computation time and computational resources. This thesis focuses on the development of a motion correction method, which seeks to rectify some of the issues above by using respiratory gated data in combination with motion modelling. Firstly, this thesis presents preliminary results, where the bounds of the problem were found. This includes experiments to discover the effectiveness of different motion correction techniques (focusing on motion modelling) in the case of Time of Flight (TOF) vs Non-Time of Flight (Non-TOF) PET data, especially where a high number of bins are used. Then, the thesis explores work related to attempting to solve the mismatch of the μ-Map by deforming it to the position during the gates. Furthermore, a comparison between the effects of the reconstruction and Motion Model (MM) fitting process is presented, including using Maximum Likelihood Activity and Attenuation Correction Factors Estimation (MLACF), to approximate attenuation correction, as well as fitting a MM on coarsely binned data and applying it to finely binned data. Finally, this thesis also presents work on the problems of Data Driven (DD) Surrogate Signal (SS) extraction methods applied to dynamic PET. A SS is imperative to the effectiveness of binning data as well as to MM fitting. By presenting work relevant to SS extraction from dynamic PET this work potentially opens the motion correction methods presented previously to the application of dynamic PET. The thesis concludes by critically reflecting on the work presented, highlighting both methodological advancements and areas for further refinement. By outlining key future directions, it sets the stage for continued development of clinically viable respiratory motion correction approaches for both static and dynamic PET imaging.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as CSVJSONXML

Downloads by country - last 12 months

Export as XMLJSONCSV

Archive Staff Only

View Item