Ultrafast structural and functional photoacoustic microscopy towards clinical applications

Abstract

Photoacoustic imaging (PAI), a hybrid imaging modality that integrates advantages of optical contrast and ultrasonic resolution, has been investigated in biomedical fields in recent years widely. Optical-resolution photoacoustic microscopy (OR-PAM) is an important branch of photoacoustic imaging and it can provide subcellular or cellular resolution of absorption targets, which is meaningful to interpret and grasp microscale mechanism for biology. To increase the detection accuracy and reduce the misalignments of physical activities, ultrafast OR-PAM is more suitable in capturing in vivo structural and functional information. In the past four years, my research mainly focuses on ultrafast structural and functional OR-PAM and its relevant biomedical applications via revision and improvement of current system or algorithm. Accordingly, the thesis is divided into three major aspects as follows. In the first aspect, we present an ultrafast wavelength switching approach to achieve both high pulse energy and clear separation of multiple photoacoustic signals. A short fiber, e.g., 10-m long, allows the stimulated Raman scattering to generate high-energy pulses at demanded wavelengths. A frequency-domain method is developed to separate two photoacoustic signals that are partially overlapped in time. Numerical simulation and phantom experiments have validated the signal unmixing method. The results show that 50-ns delay between the two A-lines induces partial overlap, these two signals can be separated with 98% accuracy with 1.4% reduction in signal-to-noise ratio. This Fourier-domain separation method is further applied in in vivo functional OR-PAM. The oxygen saturation values can be reliable obtained based the two A-lines separated at each position. This ultrafast wavelength switching technique allows fiber-based stimulated Raman scattering method generates high pulse energy with negligible misalignment among different wavelengths for photoacoustic imaging. In addition, the short wavelength switching time increases the highest A-line rate of multi-wavelength photoacoustic imaging. In the second aspect, three biomedical applications based on the ultrafast OR-PAM system developed in the first study or its modification are investigated. The first application is tissue dichroism measurement. Dichroism is a kind of material property that leads to anisotropic optical absorption with different light polarizations. It has a close relationship to types and alignments of excitation target and thus the dichroic tissues can be distinguished based the dichroism. In this study, we advance the dichroism photoacoustic imaging technique in terms of low noise and potentially high imaging speed. We develop a single-shot system that the three laser beams are splitted from one laser pulse and they are separated in time for dichroic optical-resolution photoacoustic microscopy. The three pulses have identical pulse energies, different polarizations, and a sub-microseconds time delay. In such a way, the dichroism photoacoustic imaging is insensitive to pulse energy fluctuation and can maintain a low detection noise in fast scanning. We not only measure the dichroism degree of target but also calculate the orientation angle of it. The new dichroism can be served as a contrast to image dichroic tumor or probe-conjugated tumor detection in the future. The second application is to monitor the outcomes of cupping therapy from a quantitative imaging perspective. Cupping therapy, thousands of years widely used traditional Chinese medicine practice, uses cups to create suction onto the skin, aiming to stimulate blood circulation and ease the symptoms of some diseases, such as fatigue, tension, and muscle pain. However, due to lack of scientific evidence, the actual effect has always been controversial. In order to objectively and quantitatively evaluate the cupping therapy effect, here OR-PAM is introduced to observe microenvironment parameters changes including structural and functional parameters in animal models before and after cupping through facial cups. It is found that within a short time and tiny pressure after cupping (5 minutes and -20 kPa), more capillaries appear, with some associated with slight blooding. We also quantify all the results and find that the total haemoglobin concentration increases 64% and 40% in veins and arteries, respectively. The oxygen saturation declines about 17% and 3% in veins and arteries, respectively. The elevation in haemoglobin concentration and the decrease in oxygen saturation recover to their original levels two hours later, indicating that the cupping therapy only last a short period. This study is the first time to quantitatively investigate the microenvironment changes of cupping therapy through imaging prospective, which indicates the evaluating ability of OR-PAM. These capabilities, if further engineered, can be extended for wide applications of cupping treatment monitoring and guidance. The third application is to measure medium viscosity via Grueneisen relaxation effect. Viscosity measurement is important in many areas of biomedicine and industry. Traditional viscometers are usually time-consuming and require huge sample volumes. Microfluidics overcomes the challenge of large sample consumption but takes a long time to measure it. Moreover, the microfluidic device needs to design a special supporting microstructure to measure the viscosity, which may be complex and costly. Here, we propose photoacoustic viscometry to measure the liquid viscosity in generic microfluidic devices based on dual-pulse photoacoustic flowmetry. The new viscometer method embraces fast detection speed, low fluid consumption, and high detection sensitivity, offering a new tool for efficient viscosity measurement in a broad range of microfluidic devices. In the last aspect, we propose to achieve ultrafast "OR-PAM" via deep learning. Here, the "OR-PAM" imaging performance is obtained by learning acoustic-resolution photoacoustic microscopy (AR-PAM)' scanning data but having OR-PAM's resolution. Conventional AR-PAM can image deeper than OR-PAM but with low image resolution. Besides, AR-PAM has a faster raster scanning speed as much fewer scanning steps are involved due to the limitation of resolution. How to achieve deep depth imaging while keeping high resolution and fast speed is essential in photoacoustic microscopic applications due to the trade-off between depth and resolution. Imaging quality such as contrast and signal-to-noise ratio has been improved through deep learning in recent years. Here, we propose to use that to realize ultrafast "OR-PAM" by learning the AR-PAM data so that we are able to obtain "OR-PAM" resolution and AR-PAM penetration depth simultaneously. Groups of mice ear images are collected to train the neural network. Feasibility and the efficiency of the network is validated with additional experimental mice brain AR-PAM data that are excluded in training. In summary, this thesis aims on ultrafast structural and functional OR-PAM through system modification and algorithm development. Moreover, three different biomedical and clinical applications are investigated based on the ultrafast system or its derivation. All these provide new knowledge of the structural and functional OR-PAM and constitute critical steps to further the development of photoacoustic imaging towards in-depth biomedical research and applications.