Enhancing Photoacoustic Resolution Using Sparsity-Based Reconstruction

• Author / Creator

  Egolf, David M.

• Sound waves can be generated by shining a pulsed laser on an object of interest. The resulting sound waves can be observed with an ultrasound transducer, and then reconstructed to form an image of the object. The entire process is called "photoacoustic imaging". Medical photoacoustic imaging aims to form high-quality images of parts of the body, such as vasculature, to enable better diagnosis and treatment of diseases. The resolution - the ability to tell apart objects that are very close - is an important measure of the quality of a medical photoacoustic imaging system. In the absence of prior information, the resolution of a photoacoustic imaging system is limited by the (center) wavelength of the observed sound waves. However, it is desirable to reconstruct photoacoustic images where points closer than a wavelength-based resolution limit are still resolved, enabling the production of higher quality photoacoustic images with additional detail. One approach to do this is to incorporate prior information about the nature of the imaged object. In this thesis, we explore the particular case where the unknown object is known to be a weighted sum of a small number of simple objects of known form. We call such objects "sparse". Photoacoustic images can be reconstructed while incorporating sparsity information by using a convex optimization program, a process we refer to as "sparsity-based reconstruction". We hypothesized that sparsity-based reconstruction could enable resolution enhancement in a photoacoustic setting. As a challenge to this approach, the most straightforward implementation of sparsity-based reconstruction requires a very large amount of memory. If sparsity-based reconstruction is to become a practical photoacoustic reconstruction technique, strategies for reducing its memory requirements - while preserving enhanced image quality - are relevant.

  In this work, we experimentally test the ability of sparsity-based reconstruction to super-resolve points. Working first with a linear ultrasound receive array, and then with a ring array, we reconstruct successive cross-sections of a target consisting of two crossed wires. In these experiments, we found sparsity-based reconstruction was able to super-resolve point sources. Two approaches for reducing the memory requirements of sparsity-based reconstruction were also implemented, namely "random projection" and an "alternating descent conditional gradient" (ADCG) approach. These approaches significantly reduced the memory required for sparsity-based reconstruction, while preserving the ability to experimentally super-resolve point sources. We feel that the ADCG approach is also promising for enabling sparsity-based reconstruction of objects more complex than point sources. To our knowledge, this work is among the first work to experimentally demonstrate the ability of sparsity-based reconstruction to super-resolve photoacoustic point sources. In addition, as far as we are aware, this work represents the first experimental implementation of ADCG sparsity-based reconstruction for photoacoustic imaging.

  The work in this thesis experimentally demonstrates that sparsity-based reconstruction can be used to super-resolve photoacoustic point sources. In addition, it describes and implements two strategies for reducing the memory required by sparsity-based reconstruction, while preserving resolution enhancement. Long term, we hope this work will be a stepping stone to the development of photoacoustic imaging systems that are able to efficiently incorporate prior information to form higher-quality images of the body.

• Subjects / Keywords

  • photoacoustic imaging
  • resolution
  • compressive sensing
  • sparsity
  • medical imaging
  • image reconstruction

• Graduation date

  Fall 2023

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/r3-5a9q-6p43

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.