采用单像素时间分辨测量的生物组织NIR-II宽场荧光断层成像

生物组织对第二近红外窗口(NIR-II)光的低散射效应，可使光学成像深度和清晰度显著提升。随着NIR-II荧光材料的发展，NIR-II荧光成像受到了极大关注，目前主要使用昂贵的InGaAs相机进行平面摄影，无深度分辨和定量能力。实现高性价比的层析功能，对NIR-II荧光成像技术的发展大有裨益。.本项目以发展高性价比的NIR-II宽场荧光层析成像方法为宗旨，针对其理论、技术和应用上存在的关键科学问题开展深入研究: 1)建立具有高灵敏度大动态性能、能提供时空分辨信息的高性价比NIR-II宽场荧光断层成像系统; 2)发展能有效克服NIR-II区光子输运模型复杂性及其反演不适定性、并充分利用时空分辨信息的高稳健、高定量NIR-II荧光断层图像重建深度学习方法；3)探索计算高效的数据仿真技术和训练方法。该研究将不仅建立先进的活体荧光三维成像方法体系，还为新型NIR-II荧光团的发现提供定量评价工具。

It is acknowledged that in vivo fluorescence-based optical imaging in the second near-infrared window (NIR-II, wavelength range 1000-1700 nm) provides significantly improved image quanlity, owing to reduced photon scattering, deeper tissue penetration and lower autofluorescence interference. In recent years, this modality have been highlighted by the increasingly developing palette of biocompatible NIR-II fluorephores including short-wave carbon nanotubes、quantum dots, organic dyes, and rare-earth-doped nanoparticles etc. The state-of-the-art of NIR-II imaging is majorly implemented by planar photography using an expensive but sensitivity- and resolution-limited deeply cooling InGaAs camera, and also offers neither depth resolution nor quantitative ability that are the important requirements for preclinical studies and clinical applications. Therefore, an achievement of a cost-effecitve in vivo NIR-II fluorescence tomograpgy can be greatly beneficial to both technological promotion and application extension of the modality..The objective of this proposal is to develope a NIR-II wide-field fluorescence tomography methodology for in vivo applications, by studying in depth its relevant key scientific problems in theory, technology and application. Firstly, a cost-effective NIR-II wide-field fluorescence tomography system is established with high sensitivity and large dynamic performance, where a NIR-II single-pixel imaging (SPI) scheme is adopted within the framework of the spatial frequency domain (SFD) measurement to replace the expensive InGaAs camera and to favoriably enable the highly-sensitive, information-riched time-correlated single photon counting (TCSPC) detection. Secondly, a deep-learning algorithm for quantitative image reconstruction of NIR-II fluorescence tomography is explored with the physics-informed U-net structure that can make full use of the space-time resolved measurement providing by the SFD-TCSPC-SPI system to effectively mitigate the physical model complexity and inversion uncertainty incurred by NIR-II photon migration and wide field time-domain measurement. Finally, a data stimulator of high-fedality is adopted based on the computationally more efficient hybrid transport-diffusion model for reduction of the training burden. The success of the project not only provides an advanced methodology of three-dimensional fluorescence imaging modality for in vivo applications, but also a quantitative evaluation tool for discovery of novel NIR-II fluorophores.