PhD Thesis Defenses

High-resolution biomedical phase-contrast tomography


Improved three-dimensional biomedical imaging can give a better understanding of tissue structure, growth and diseases. Most present imaging techniques that provide cellular spatial resolution are based on visible or infrared light. These methods cannot image deeper than a millimeter into tissue. Consequently, larger samples cannot be completely imaged without sectioning. Techniques that are typically used to image larger samples don’t provide sufficient contrast and resolution to image cellular-sized features in soft tissues. There is a need for new imaging methods that can fill the gap between present methods. For practical reasons, compact equipment is preferred, to enable close connection to other research and applications. Furthermore, minimized sample preparation both reduces the work needed and the time until results are ready.

In this Thesis, propagation-based phase-contrast tomography with liquid-metal-jet x-ray sources has been investigated for high-resolution three-dimensional biomedical imaging. By using phase contrast, the contrast for cellular-sized features in soft tissue is vastly increased compared to absorption, also in larger samples. The high resolution relies on using an x-ray source with small emission spot, but also with high power to keep exposure times reasonable.

This Thesis is about developing and optimizing experimental methods and image reconstruction algorithms. A new method to remove ring artifacts was developed and tested, and a comparison of multi-material phase-retrieval algorithms was made. The improvements provide better contrast and resolution, as well as reduce noise and artifacts. The improved image quality is demonstrated in a few biomedical applications. It is shown that the method can image 5 µm large myofibrils in whole-body zebrafish, despite the small size and low contrast of myofibrils. A high-resolution tomography of a mouse can be done fast by using a specialized high-power source. The image quality in tomographies of both human coronary arteries and a mummified human hand is sufficient to analyze the tissues and cellular-sized features, which is something that could be called virtual histology.