Serial and comprehensive quantitative assessment of lung disease from CT
Respiratory medicine has a need for quantitative regional tools for disease assessment. Clinical decision-making is based on global functional measurements and qualitative image assessment, which are insensitive to the localised changes that are characteristic of many lung diseases. In particular, the quantitative assessment of routine serial CT images has the potential for more sensitive detection of disease progression and early detection of comorbid lung conditions.
This project will combine demographic and clinical CT data from existing databases and ongoing collaborations to develop a pipeline for deriving quantitative regional disease metrics from CT images and evaluating their change over time.
The project consists of the following components:
- Develop a robust non-rigid image registration method for relating CT lung time series, taking into account lung deformation and sliding at the lung and ideally lobar boundaries. This method will most likely build upon the NiftyReg toolkit.
- Develop a tool that allows clinicians to interactively define 3D objects (such as nodules or diseased regions) on a lung scan. This may exploit clinical knowledge about the appearance and shape of the regions to improve efficiency. This method will likely build upon existing tools such as Slic-Seg and NiftySeg.
- Develop a pipeline that exploits the above registration scheme to propagate defined regions forwards and backwards through the time series for comparison.
- Exploit the above pipeline to develop quantitative clinical measurements describing disease progression, for example:
a) the change in regional measurements such as tissue density and tissue textures;
b) the measurement of nodule volume changes following interactive refinement of the propagated regions
Within the clinical themes, this project directly aligns with the aims of the Cardiovascular Imaging and Cancer Imaging programmes. The aim is to create a pipeline that fits into the standard standard clinical workflow and is specifically focussed on the clinical need for assessing disease progression using standard follow-up imaging. The scheme design will therefore be highly driven by clinical input.
Within the methodological themes, this project fits in with the Image Analysis theme, building upon and further developing existing image registration work, and deriving new analysis techniques for detecting and assessing lung nodules or tissue characteristic of regional disease changes.