Study on High-sensitivity Photoacoustic Microscopy and Optical Coherence Tomography Dual-modality Imaging Technology

Abstract

Photoacoustic imaging is a multiscale biomedical imaging technology. Endogenous or exogenous molecules absorb pulsed laser energy, undergo thermoelastic expansion, and subsequently generate ultrasound waves. Through ultrasound detection and signal reconstruction, three-dimensional images can be obtained. Leveraging the high light-absorbing property of hemoglobin in red blood cells, photoacoustic imaging enables label-free imaging of blood vessels, metabolism, and functional activities within organisms. This technology has various applications in early tumor diagnosis, stroke screening, diabetes assessment, and other related fields. Through tight focusing of the excitation laser and ultrasound field, photoacoustic microscopy (PAM) achieves micrometer-level resolution. However, PAM exhibits specific limitations in some applications. First, the contrast of PAM originates from specific absorbers such as hemoglobin. A variety of organs and tissues exhibit weak light absorption, making them hardly detected by PAM. Second, photoacoustic signals may also arise from myoglobin, melanin, and extravasated blood, potentially interfering with vascular imaging. Third, the sensitivity of PAM is related to laser pulse energy. Thus, PAM contrast is insufficient in some energy-constrained applications, such as high-speed imaging, exogenous probe imaging, and ocular imaging.

Optical coherence tomography (OCT) is a biomedical imaging technique based on optical scattering, serving as a complementary modality to PAM. OCT relies on light interference, enabling high sensitivity under safe laser exposure levels. Optical refractive index varies in biological tissues, allowing OCT to visualize detailed structure information, including multiple layers of the fundus, skin, and bone. OCT can also visualize vascular networks by detecting variations in backscattered light at the same location, which is termed optical coherence tomography angiography (OCTA). By utilizing near-infrared light and relying on blood flow, this method remains less susceptible to interference from bleeding or melanin-induced light absorption. However, OCT-based oxygen saturation measurements are less accurate than those obtained by PAM.

Leveraging optical absorption (PAM) and optical scattering (OCT), these modalities fully exploit the interaction between photons and tissue. However, existing dual-modality imaging methods suffer from suboptimal PAM sensitivity, which limits their broader applicability. This thesis focuses on the study of high-sensitivity PAM-OCT dual-modality imaging technology. Chapter 1 discusses the background and fundamental principle of biomedical imaging, photoacoustic imaging, and optical coherence tomography.

The objective of chapter 2 is to develop a confocal PA-OCT dual-modality microscopy system. Most existing studies on PAM-OCT dual-modality systems employ optical scanning via galvo-mirrors and unfocused ultrasound transducers. Such systems ensure good modality matching and fast imaging speed, while exhibiting relatively low and non-uniform sensitivity, leading to suboptimal photoacoustic imaging quality. This study proposes a confocal design to achieve simultaneous PAM and OCT imaging, providing high-contrast vascular imaging and high-resolution multilayer tissue structural information. To ensure high sensitivity and accurate alignment of the two modalities, the imaging probe is designed to precisely align the PAM laser beam, OCT laser beam, and ultrasound detection path. Animal experiments demonstrated that structural and functional imaging of skin, blood vessels, and bone with uniformly distributed sensitivity was achieved. This confocal PAM-OCT dual-modality microscopy system enables the acquisition of rich biological information with high resolution and sensitivity, serving as an effective tool for biomedical research.

Chapter 3 focuses on high-sensitivity PA-OCTA dual-modality microscopy system. Photoacoustic vascular imaging is susceptible to interference from blood leakage and melanin. OCTA relies on the decorrelation of blood flow, serving as a complementary modality. However, OCTA imaging requires fast scanning and exhibits low contrast for slow blood flow. By comprehensively considering the characteristics of both modalities, we proposed a switchable PA-OCTA dual-modality microscopy system. PAM and OCT/OCTA images are acquired sequentially and then co-registered in 3D space. This system enables the acquisition of tissue structural, vascular network, and flow information in biological samples. Animal experiments demonstrated that this system enables multi-parameter label-free functional imaging of the mouse ear and brain while mitigating potential image degradation.

Chapter 4 of this study is to enhance photoacoustic imaging sensitivity via a self-supervised learning algorithm. While several hardware-based approaches have been proposed to improve photoacoustic imaging sensitivity, high-contrast vascular imaging remains challenging at extremely low laser pulse energies, which minimize the photobleaching, phototoxicity, and heating hazards. Recent advancements in deep learning have demonstrated promising image enhancement capabilities, yet they often require high-quality ground truth data, which may not be available in photoacoustic imaging context. Therefore, we propose a self-supervised learning algorithm to enhance photoacoustic image signal-to-noise ratio (SNR) using only noisy data during the training process, which leverages intrinsic spatiotemporal redundancy in signals and the independent, random nature of noise. Animal experiments have demonstrated that this method significantly improves signal SNR, achieving structural and functional vascular imaging with extremely low laser pulse. This method enables high-quality photoacoustic imaging in laser-constrained applications.

Chapter 5 aims to further enhance PAM sensitivity via OCTA guidance. Even with advanced image enhancement algorithms, the information that can be extracted from PAM data alone remains limited. Co-registered OCTA images provide high-quality vascular structural information, which has been widely applied in clinical practice. We propose a multimodality fusion deep learning algorithm to collaboratively enhance PAM image quality. OCTA provides prior knowledge of vascular structure and location to recover corrupted PAM data. Animal experiments demonstrated that blood oxygen saturation imaging of mice at laser pulse energy below 1 nJ was achieved. This method expands the boundaries of ultra-high-sensitivity photoacoustic microscopy.

Chapter 6 presents conclusions and outlines future research directions. This Ph.D. thesis focuses on high-sensitivity PAM-OCT dual-modality imaging technology. By leveraging optical absorption and scattering in biological tissue, these two modalities have provide complement and correlated information. This study integrates the advantages of PAM and OCT to develop high-sensitivity dual-modality systems and image enhancement algorithms, providing a multi-contrast imaging platform for label-free in vivo biomedical imaging. This technology is anticipated to be applied in diverse biomedical areas, including neurovascular coupling, tumor diagnosis, and ophthalmic examinations, etc.

Date of Award	3 Sept 2025
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Lidai WANG (Supervisor)