Hyperspectral Imaging for Diffuse Optical Tomography

Diffuse optical tomography (DOT) has emerged in the last decade as a new and exciting tool for functional medical imaging with applications in a range of areas including breast cancer detection and diagnosis. DOT employs observations of near infrared (NIR) light that has propagated through tissue to reconstruct the spatial distribution of various chromophores present in the region of interest. In the case of breast cancer, oxygenated and de-oxygenated hemoglobin are of particular interest in identifying and characterizing tumors. It is well known that the DOT reconstruction process can be quite sensitive to noise and other un-modeled effects due to the diffusive nature of the underlying physics as well as the limited aperture over which data can be acquired in many practical systems. While there exist a wide array of mathematical techniques for stabilizing the reconstruction, ideally one would like a richer data set. Most DOT instruments employ no more than five NIR wavelengths to probe the tissue; however recent work in the diffuse optical imaging group in the Biomedical department has led to the development of a hyperspectral system in which hundreds of wavelengths can be acquired. With the increase in data however comes an associated rise in the complexity of the image formation process. In this thesis, we explore the development and performance of algorithms for hyperspectral DOT. We detail an efficient method for forming the images based on the use of iterative algorithms applied to a linearized measurement model. Simulation and experimental results will be provided which show the advantages of hyperspectral imaging.