Patrick Møller Jensen: The Big Picture from Small Details
To understand the whole, it is helpful to first study its parts. As an example, analyzing the 3D micro-structure of bones may lead us to early indicators for osteoporosis. Furthermore, studying the size and shape of textile fibers could help us understand how laundry detergent affects our clothes. One way to do this is via volumetric images (or 3D images), since these allow us to directly look at the micro-structure we want to examine.
The problem is that the differences between two (or more) samples may be elusive, even for the hu-man eye. Furthermore, in order to make quantitative statements about samples, it is important that analysis approaches consider the entire volumetric image as a whole. However, current methods have not kept up with the rapid advancements in imaging techniques, which makes large-scale quantitative analysis of volumetric images difficult and time consuming.
This project will attempt to narrow the gap between image acquisition and analysis by developing effi-cient analysis methods for volumetric images. A main focus will be the fitting of geometric models – e.g. triangle meshes – to the image data. Using geometric models allows us to incorporate our prior knowledge about the structures of interest, which can both improve the fit itself, and make it easier to perform subsequent analysis. Another major focus area will be computational efficiency. As volumetric images are often quite large, a great amount of computation is required for processing. Thus, it is nec-essary to carefully consider the efficiency of the implemented algorithms, since naive approaches are usually not feasible.
The methods, algorithms, and tools developed in this project will be applicable to a wide range of im-age analysis tasks, and will ultimately benefit the broader research community – both at DTU and elsewhere.