相位敏感型光学相干层析成像中相位泄漏问题的分析与消除

Phase-sensitive optical coherence tomography (PhS-OCT) is an important functional extension of optical coherence tomography. It is also a common technological foundation for several other functional imaging modalities and is widely adopted in both research and clinics. However, current PhS-OCT technique is prone to the effect of phase leakage: the depth-resolved phases from adjacent layers manifest mutual coupling and crosstalk, which lead to a contamination in the phase measurements. This phenomenon will not only limit the further improvement on the accuracy of phase measurements, but will also give erroneous results under certain circumstances. To address this issue, the applicant intends to study the effect of spectral leakage on the imaging model by incorporating the spectral sampling and develop a novel imaging model based on spatial shift-variant system. The proposed project is planned to study the cause and mechanism of phase leakage and explore the possible ways to mitigate the phase leakage effect by means of constrained optimization. To conclude the project, an experiment on the system will be conducted to verify and prove the proposed imaging model and the algorithm. The applicant believes that the successful execution of the project could help construct a novel theoretical framework for OCT imaging, and help break the bottleneck imposed on the accuracy of the phase measurement of PhS-OCT. It will also help lay a solid foundation for the future development of novel functional imaging techniques that caters to the needs of medical diagnosis.

相位敏感型光学相干层析成像（PhS-OCT）是传统光学相干层析成像的一种重要扩展，也是多种功能性成像的共性技术基础，不论是在科研上还是在临床上都有着重要的应用。现有的PhS-OCT技术容易受到相位泄漏的影响，相邻待测反射面的相位之间存在着耦合和串扰。该效应的存在不但制约了PhS-OCT相位测量精度的进一步提高，而且在某些特定条件下甚至会给出错误的测量结果。针对这一问题，申请人拟通过研究频域采样和频谱泄漏在OCT成像系统中的作用，构建一个基于空间移变成像系统的新模型。本项目拟以新模型为核心，分析相位泄漏的成因和机理，探索新模型下基于约束优化方法的相位泄漏消除算法。最后，在模型系统上实验验证所提出之新模型和新算法的有效性。本项目的顺利完成将有助于构建一个新的OCT成像理论体系，突破PhS-OCT在相位测量精度上的技术瓶颈，为发展适用于医学诊疗的新型功能性成像技术奠定基础。

相位泄漏是困扰光学相干层析成像领域的一个重要的问题。由于相位泄漏的存在，导致FD-OCT在获取超过轴向分辨率极限的样品结构信息时，信号存在巨大的失真，使得准确地提取相位或者结构信息变得异常困难。本项目从成像理论、重建算法、图像增强、系统应用等多个层次对相位泄漏问题展开了全方位的研究。在成像理论上，我们对相位泄漏的原因进行了严格地理论分析和数理建模，首次对相位泄漏引发的coherence beating现象进行了理论解释，并精确预测了实验中的现象；在重建算法上，提出了一种全新的、单步的、融合图像处理与图像重建的、相位泄漏消除算法，通过使用GPU并行处理对算法进行加速，达到超过1000倍的速度提升，同时提高了重建后图像的SNR和CNR等；在图像增强上，项目组开展了各类基于深度学习和传统方法的图像增强，提出使用基于图神经网络的深度学习算法对视盘为中心的眼科OCT图像进行分割，并将分割精度提高了1.5个百分点，项目组同时对眼科OCT图像修补进行了研究，相关工作获得SPIE Medical Imaging 2023图像处理最佳学生论文奖；在系统应用上，课题组搭建扫频光源OCT显微和眼科系统各一套，为后续进一步开展研究工作打下了坚实的基础。通过本次项目，课题组通过本项目掌握了FD-OCT在超越轴向分辨率极限条件下的关键物理机制，构建了相关的数理模型，设计了相位泄漏问题的消除算法，建立样机试验系统，为将来进一步研究高性能FD-OCT系统提供了理论知识和技术储备。