Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Quantifying the impact of the interpolation error after registration of time-lapsed high-resolution peripheral quantitative computed tomography images on static bone parameters.
Time-lapsed imaging provides valuable insight into changes occurring over time where image registration is used to align baseline with follow-up images. This alignment requires image interpolation which might affect the outcome of a subsequent bone morphometric and biomechanical analysis. The impact of the interpolation error after registration of time-lapsed human in-vivo bone images has not been quantified yet. Therefore, the purpose of the present study is to investigate different interpolation methods on static bone parameters.
Keywords: HR-pQCT Imaging, Image Registration, Data Analysis
Current high-resolution peripheral quantitative computer tomography (HR-pQCT) scanners allow for the assessment of human bone microstructure in vivo. To track temporal changes in the bone microstructure, e.g. when studying bone remodeling or the effect of a treatment, time-lapsed imaging of bone can be used. When scanning an object at different time-point, slight deviations in the object’s positioning cannot be avoided, which leads to spatial misalignment between the time-lapsed images. To overcome the problems associated with spatial misalignment, image registration can be used. The purpose of image registration is to find the most appropriate transformation with which one image can be re-oriented so that it fits optimally onto another image. The optimal alignment is found by iteratively transforming one image and assessing its alignment to the other image using a similarity metric. Transforming an image into a new voxel grid requires interpolation of the original voxel intensities. This interpolation error might not considerably interfere with the search for the optimal alignment during registration. However, the subsequent morphometric analysis of the re-aligned image might be affected by interpolation. A previous study quantified this effect using images of mouse vertebrae with a resolution of 10.5 microns. In the present study we want to investigate the effect of different interpolation methods on static bone parameters deduced from patient images with a resolution of 61 to 82 microns.
Repeated HR-pQCT-images should be used in this project to assess the reproducibility of different static bone parameters. Based on the literature, a relevant collection of static bone parameters should be chosen. The influence of registration and different interpolation methods on the reproducibility should then be investigated. Furthermore, the effect of the interpolation error of different interpolation methods on the static bone parameters should be quantified. The project can then be further expanded to study the influence of the order of the post-processing steps.
Current high-resolution peripheral quantitative computer tomography (HR-pQCT) scanners allow for the assessment of human bone microstructure in vivo. To track temporal changes in the bone microstructure, e.g. when studying bone remodeling or the effect of a treatment, time-lapsed imaging of bone can be used. When scanning an object at different time-point, slight deviations in the object’s positioning cannot be avoided, which leads to spatial misalignment between the time-lapsed images. To overcome the problems associated with spatial misalignment, image registration can be used. The purpose of image registration is to find the most appropriate transformation with which one image can be re-oriented so that it fits optimally onto another image. The optimal alignment is found by iteratively transforming one image and assessing its alignment to the other image using a similarity metric. Transforming an image into a new voxel grid requires interpolation of the original voxel intensities. This interpolation error might not considerably interfere with the search for the optimal alignment during registration. However, the subsequent morphometric analysis of the re-aligned image might be affected by interpolation. A previous study quantified this effect using images of mouse vertebrae with a resolution of 10.5 microns. In the present study we want to investigate the effect of different interpolation methods on static bone parameters deduced from patient images with a resolution of 61 to 82 microns. Repeated HR-pQCT-images should be used in this project to assess the reproducibility of different static bone parameters. Based on the literature, a relevant collection of static bone parameters should be chosen. The influence of registration and different interpolation methods on the reproducibility should then be investigated. Furthermore, the effect of the interpolation error of different interpolation methods on the static bone parameters should be quantified. The project can then be further expanded to study the influence of the order of the post-processing steps.
Quantification of the impact of the interpolation error after registration of time-lapsed high-resolution peripheral quantitative computed tomography images on static bone parameters.
Quantification of the impact of the interpolation error after registration of time-lapsed high-resolution peripheral quantitative computed tomography images on static bone parameters.
Gianna Marano (gianna.marano@hest.ethz.ch), Institute for Biomechanics, ETH Zurich, Professorship Ralph Müller
Gianna Marano (gianna.marano@hest.ethz.ch), Institute for Biomechanics, ETH Zurich, Professorship Ralph Müller