Abstract
Photoacoustic computed tomography (PACT) is a hybrid imaging modality that combines the strong optical absorption contrast of photoacoustic imaging with the deep tissue penetration capabilities of ultrasound. It is well-suited for wide-field functional imaging of deep biological tissues. By utilizing endogenous absorbers such as hemoglobin or exogenous contrast agents, PACT allows for the visualization of various physiological parameters, including blood sO₂ and vascular density. This technique has been widely applied in tumor diagnosis, brain imaging, and cardiovascular monitoring. However, despite its high sensitivity to functional changes, PACT has intrinsic limitations in delineating soft tissue boundaries. In particular, when using a linear-array transducer, the limited view angle leads to incomplete reconstruction of vertically oriented blood vessels, resulting in missing information and reduced structural fidelity.In contrast, conventional B-mode ultrasound imaging offers high imaging speed, excellent soft tissue delineation, and robust depth penetration. These characteristics make ultrasound imaging highly suitable for providing structural references to guide photoacoustic imaging. Furthermore, ultrasound Power Doppler enables effective visualization of blood vessels based on flow signal amplitude, and the view angle of the linear array does not constrain its performance. As a result, integrating Power Doppler with photoacoustic imaging not only improves vascular localization but also has the potential to compensate for structural loss in limited-view photoacoustic imaging.
Given this complementary relationship between ultrasound and photoacoustic imaging, this study proposes an ultrasound-assisted photoacoustic imaging strategy focused on three significant aims:
(1) to achieve accurate anatomical localization via ultrasound imaging and validate the feasibility of cyanine-based photoacoustic probes for tumor diagnosis.
(2) to perform skin and muscle segmentation based on ultrasound images, facilitating functional parameter analysis of different vascular in both regions.
(3) to incorporate ultrasound Power Doppler as structural guidance to mitigate image loss due to limited view angles in linear-array PACT systems.
This research centers on the fundamental principles, algorithm development, and experimental validation of PA/US dual-modality imaging. A multifunctional and high-resolution imaging platform is established to improve the multidimensional characterization of biological tissues, including their anatomical structure, physiological state, and pathological changes.
On the theoretical level, the study first analyzes the physical basis and technical characteristics of photoacoustic and ultrasound imaging. Photoacoustic imaging leverages the photoacoustic effect, which converts absorbed optical energy into ultrasonic waves, enabling high contrast and molecular-specific imaging. In contrast, ultrasound imaging relies on the reflection of acoustic waves and excels in visualizing soft tissue morphology with strong penetration depth and real-time capabilities. The complementary nature of these two modalities offers a solid foundation for integrated dual-modality imaging systems.
For tumor imaging, a novel dual-wavelength photoacoustic probe based on a cyanine dye platform was designed. This probe exhibits opposite signal responses at 635 nm and 780 nm when interacting with tumor environments. By combining ultrasound-guided localization and dual-wavelength photoacoustic imaging, this strategy significantly improves the specificity and sensitivity of tumor detection. Experimental validation in mice models demonstrates that this probe enables functional discrimination of tumor diagnosis based on esterase or hydrogen peroxide concentrations.
In vascular imaging applications, the study introduces a vascular segmentation strategy based on the sO₂ performance during the vascular occlusion tests (VOT). Ultrasound images were used to separate the interesting layers, such as the skin and muscle layers. A Nine-Grid Segmentation method was proposed to divide the localized vessels of interest into a 3×3 classification, enabling dynamic monitoring of sO₂ changes in arteries, veins, and capillaries during the VOT. Additionally, this segmentation method supports quantitative analysis of type-specific vessel proportions, offering a valuable tool for studying microcirculatory function under different physiological and pathological conditions.
To further address the limited-view constraints of linear-array PACT, this study integrates ultrasound Power Doppler with deep learning–based image enhancement. An attention-based model was developed to learn vascular structural features from ultrasound Power Doppler images and guide the completion of missing photoacoustic signals. Phantom experiments verified the model’s ability to reconstruct tube structures that were missing due to angular limitations. When applied to in vivo imaging of rat and human tissues, the method demonstrated substantial improvements in feature structure, vascular continuity, and signal-to-noise ratio.
To this end, several promising research directions emerge. First, the ratiometric cyanine probe platform can be expanded by incorporating disease-responsive elements (e.g., ROS, pH, enzymes) and targeting ligands, enabling precise molecular imaging in conditions such as cancer, ischemia, or stroke. Second, the Nine-Grid Segmentation Strategy, introduced for classifying vascular responses during occlusion-reperfusion, should be validated in animal models and clinical settings to support personalized assessment of microvascular dysfunction. Finally, the DEPANet framework, originally designed to enhance PA imaging with ultrasound Power Doppler guidance, can be generalized to other imaging modalities—such as elastography, OCT, or PET—to improve structural and functional interpretation. Together, these directions will advance both the technological depth and clinical translation of photoacoustic imaging systems.
In conclusion, this research demonstrates a comprehensive and integrative approach to overcoming the technical limitations of linear-array PACT through ultrasound fusion and machine learning. By advancing photoacoustic and ultrasound foundations, this work contributes to the growing field of multimodal biomedical imaging and paves the way for future clinical translation in disease diagnosis, treatment planning, and personalized medicine.
| Date of Award | 16 Sept 2025 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Lidai WANG (Supervisor) |