Artefact detection and measurement of surface change in stereophotogrammetry data

Abstract

The diagnosis and treatment planning of disease may require image data acquisition using a range of 3D medical imaging modalities. These modalities include computed tomography, magnetic resonance, single photon emission computed tomography and positron-emission tomography. In combining images from such modalities, clinicians are provided with aligned data sets that have structural and functional information from within the body. However, no external textural data of the body surface (skin, wounds, and surface disease) is collected. Stereophotogrammetry can produce high resolution topographical surfaces with texture information and is becoming more readily available in the clinical environment. Combining stereophotogrammetry surface data with conventional 3D medical imaging data has the potential to allow improved visualisation of the 3D image data and to provide information on how surface data relates to the patient’s internal anatomy. Medical image data is prone to outliersand artefacts due to physical limitations of the modalities involved and patient specific characteristics.

To combine a variety of image data into a universal co-ordinate frame an image registration method is required. Outlier-robust registration is required due to the presence of artefacts and noise within the surfaces and images. A range of registration methods are evaluated using phantom test objects in the presence of outliers, simulated as noise, to determine the performance of the registration algorithms.

A novel automated method is proposed to identify and examine artefacts at the surface edge of stereophotogrammetry data, using previously acquired registered volumetric image data as a reference. The largest stereophotogrammetry artefacts are observed at the surface edge and have a negative impact on the registration accuracy. Identification and removal of these surface edge artefacts is investigated using novel automated cleansing algorithms. The proposed cleansing methods are evaluated using quantitative and qualitative measures to assess the level of success for the implemented approaches. Results are presented in which stereophotogrammetry surface accuracy canbe increased whilst ensuring that the surface is suitable for visualisation.

Surface change can potentially indicate underlying anatomical change due to patient growth or disease. An automated algorithm is proposed to identify and measure regions of surface change in phantom stereophotogrammetry surfaces. These surface change regions are identified using information obtained from the registration of the stereophotogrammetry surfaces and the corresponding volumetric image. The preliminary results indicate that the proposed method can beused to locate and measure regions of surface change over time in stereophotogrammetry data.

Date of Award	Feb 2013
Original language	English
Sponsors	Department for Employment and Learning (DEL)
Supervisor	Philip Morrow (Supervisor), John Winder (Supervisor) & Bryan Scotney (Supervisor)