Developments Towards Chromophore-Selectivity in Photoacoustic Remote Sensing Microscopy

Abstract

Medical imaging serves to diagnose and monitor illnesses in clinical settings. Several modalities such as X-Ray and Ultrasound have become ubiquitous due to their high utility. Photoacoustic microscopy, based on the photoacoustic effect originally discovered by Alexander Graham Bell, is sensitive to optical absorption contrast. This is highly useful in biomedical applications where the endogenous absorption contrast of tissue can directly be imaged, enabling label-free microscopy. Applications include histological assessment of tissue in support of cancerous tissue resection surgeries, as well as the functional imaging applications of blood oxygen saturation (sO2) and metabolic rate (MRO2) imaging. Recently, Photoacoustic Remote Sensing Microscopy (PARS), an all-optical implementation of photoacoustic microscopy, was pioneered by Parsin Haji Reza.

This thesis makes three main contributions supporting the development of PARS microscopy. The first is the creation of an inverse model designed to solve for the concentrations of individual chromophores when imaged using several excitation wavelengths. To achieve this, constraints and considerations specific to PARS are designed and employed. The inverse model allows for the concentrations of oxy- and deoxyhemoglobin to be solved for, enabling sO2 estimation. This was performed in-vivo in an ocular setting, demonstrating the first non-contact photoacoustic measurement of sO2 in the eye. The second contribution is an in-depth experimental study of Stimulated Raman Scattering (SRS) in single-mode optical fiber as a means of generating multi-wavelength light from a conventional single-wavelength laser source. Effects associated with several laser parameters and properties of optical fiber are studied. Results of this study will find use in the apparatus design for nearly any multi-wavelength PARS application, typically where imaged absorbers must be unmixed, and in particular, in sO2 estimation. The final contribution is the development of a novel unsupervised time-domain feature-learning algorithm, designed to learn characteristic signal shapes. This allowed tissue sub-components to be discerned in PARS imagery of both unstained human breast tissue on slides and freshly resected murine brain tissue without the need to use multiple excitation wavelengths nor have any prior knowledge of the time-domain characteristics associated with individual components.

The contributions made in this thesis represent significant steps towards the use of PARS for a broad range of applications where unmixing, or more specifically, discerning underlying components of the imaged target is required — beyond sO2 estimation or emulation of standard histological techniques. Furthermore, the improved understanding of how SRS can be used to generate additional excitation wavelengths opens the door to imaging an abundance of bio-molecules, thus broadening the scope and richness of the gamut of targets that PARS is capable of imaging.