差分干涉光场数字全息生物组织三维成像研究

Research on biological composition and activity is an important way for human understanding of the nature of life and, dealing with diseases. Besides, the cognitive degree of nerves and brain structures bring great driving force for the development of artificial intelligence. In order to meet the requirements of larger application range, faster imaging speed and higher imaging precision proposed in the field of biomedicine, a technology combining DMD device and Eccentric beam expanding system based on micro-lens array is proposed to realize two-dimensional angular modulation and coding. By combining this technique with digital holography, the measurement of the two-dimensional projection of the three-dimensional phase distribution of the biological tissue in each light propagation direction can be realized. By using the oblique illumination sample beams corresponding to the adjacent micro-lenses as the measurement light and the reference light of the digital holography, respectively, the angle between the measuring beam and the reference beam is reduced, so that the phase information carried by the light passing through the sample at a large angle can be accurately recorded to provide more adequate sampling data for high-precision three-dimensional phase distribution reconstruction of the sample. In view of the existence of scattering and diffractive effects in biological tissues, a coarse three-dimensional phase distribution structure is obtained by using Fourier projection slice method. By dividing the tissue into multiple slices, the real and accurate three-dimensional phase distribution can be obtained with an iterative way while the coarse 3D phase map and the two-dimensional phase maps from digital holography serve as the initial values and constrain conditions, respectively. Finally, three-dimensional phase distribution of biological tissues with a sample size larger than 5 mm × 5 mm × 3 mm is achieved at a high speed (less than 0.5 seconds for single three-dimensional data acquisition) and high-resolution (better than 1 μm) measurement and reconstruction.

研究生物组成及活动是人类认识生命本质、寻求疾病治疗的重要手段，对神经及脑结构的深层认知也是人工智能领域发展的极大驱动。针对生物医学提出的更大范围、更快速度、更高精度的成像要求，本申请提出将DMD与透镜阵列偏心扩束系统结合，实现光束传播方向的快速二维调制与编码，此技术与数字全息相结合能够实现生物组织三维相位分布沿各传播方向积分投影的测量；分别将经过相邻微透镜对应的光束作为数字全息的测量光和参考光，减小了测量光与参考光间夹角，使得大角度光束所携带样本相位信息可被准确记录，为三维相位分布重构提供更充足数据；提出利用傅里叶切片法得到相位三维分布初始结构，并以各投影相位测量结果作为约束条件，将样本平行分割为多个薄切片，通过重复迭代更新的方式获得三维相位分布。最终实现样本尺寸大于5mm×5mm×3mm的生物组织的三维相位分布高速(单幅三维数据采集时间小于0.5s)、高分辨率(优于1μm)测量与重构。

光学显微成像是探索生物样本微观结构的主要手段，实现对透明/半透明生物样本的三维成像是进一步了解微观生命体组成结构的主要途径。然而，由于携带更多样本信息的相位信息无法被光电探测器直接探测，而能够实现相位测量的经典的干涉法存在着系统庞杂、鲁棒性差等问题，限制着低成本、高效率、高性能生物样本的三维成像技术发展和设备研制。本项目旨在探索计算显微成像的物理模型、算法框架以及系统构建和参数校正方法，为下一步高性能计算显微成像理论及平台的构建提供了理论基础和技术支撑。项目的具体研究内容包括：.一、研究光与透明/半透明样本相互作用机理的基础上，设计构建了基于阵列光源照明的生物样本三维层析及大视场、高分辨率傅里叶叠层显微成像系统；.二、结合物理先验知识与数学逆问题优化算法，实现了傅里叶叠层显微系统各项参数误差的高精度校正，并利用离焦样本倾斜照明图像偏移的特性，实现快速地数字冲聚焦，将显微物镜的景深提升10倍以上，大大扩展了其三维层析成像的能力；.三、本项目还将深度学习用于计算显微成像中，构建全参数化物理网络模型，一次性对基于阵列光源照明的傅里叶叠层显微成像系统的多种参数误差进行高精度地校正。