基于计算自适应光学的活体人眼视网膜高分辨成像技术研究

High-resolution in vivo human retina imaging is important for early diagnosis of fundus diseases, pathological study and evaluation of drug efficacy, while the presence of ocular aberrations severely reduces the resolution ability of ordinary retinal imaging techniques. Although hardware-based adaptive optics retinal imaging technology can partly eliminate the impact of ocular aberrations on retinal imaging and obtain high-resolution images, but due to the high economic cost and complexity, small imaging field, the existence of residual aberrations and other factors, this technique has not been widely used in clinical trials. Based on our previous research, the computational adaptive optics technology will be introduced to in vivo retinal imaging in this project. A computational adaptive optics algorithm based on deep learning will be innovatively proposed, so that the high-accuracy detection and correction of human eye aberrations in large field of view can be realized by digital computation. Based on this, the performance of the algorithm and its application boundary will be further analyzed, and the imaging scheme will be designed and optimized. At last, a new in vivo retinal imaging system based on computational adaptive optics with low cost, high resolution and large field of view will be developed and imaging experiments will be carried out. These will lay the foundation for the technology to enter clinical application and promote the development of the scientific instruments of ophthalmology in China.

活体人眼视网膜视细胞级高分辨率成像对眼底疾病的早期诊断、发病机理研究及药物疗效评价等具有重要作用，而人眼像差的存在严重降低了普通视网膜成像技术的分辨率。尽管基于硬件自适应光学的视网膜成像技术能够部分消除人眼像差对成像的影响而获得高分辨率图像，但因经济成本及复杂度过高、成像视场过小、存在残余像差等因素的限制而未能在临床上大规模推广和使用。本项目在课题组前期研究的基础上，将计算自适应光学技术引入活体人眼视网膜成像中，创新性地提出基于深度学习的计算自适应光学算法，从而以数字计算的方式实现大视场下人眼像差的高精度探测及矫正。在此基础上，项目进一步分析算法性能及适用边界，优化设计成像方案，研制建立一套具有低成本、高分辨率和大视场特性的新型计算自适应光学活体人眼视网膜成像系统，并开展眼底高分辨成像实验，为该技术进入临床应用奠定基础，推动我国眼科科学仪器的发展。

活体人眼视网膜视细胞级高分辨率成像对眼底疾病的早期诊断、发病机理研究及药物疗效评价等具有重要作用，而人眼像差的存在严重降低了普通视网膜成像技术的分辨率。尽管基于硬件自适应光学的视网膜成像技术能够部分消除人眼像差对成像的影响而获得高分辨率图像，但因经济成本及复杂度过高、成像视场过小、存在残余像差等因素的限制而未能在临床上大规模推广和使用。本项目在课题组前期研究的基础上，将计算自适应光学技术引入活体人眼视网膜成像中，创新性地提出基于深度学习的计算自适应光学算法，从而以数字计算的方式实现大视场下人眼像差的高精度探测及矫正。在此基础上，项目进一步分析算法性能及适用边界，优化设计成像方案，研制建立一套具有低成本、高分辨率的新型计算自适应光学活体人眼视网膜成像系统，并开展眼底高分辨成像实验，为该技术进入临床应用奠定基础，推动我国眼科科学仪器的发展。