目的：获取眼表图像的综合信息，建立眼表疾病综合诊断和评估。方法：将超高分辨率光学相干断层成像仪(ultra-high resolution optical coherence tomography，UHR-OCT)与基于裂隙灯生物显微镜的微血管成像系统相结合，开发了一种多模态、非接触式的眼科光学成像平台。结果：UHR-OCT模块在组织中实现轴向分辨率约为2 μm 。眼表微血管成像模块在最大放大倍率下横向分辨率约为3.5 μm。通过集成在裂隙灯显微镜成像光学路径的不同模块，多模态成像平台能够执行实时前段OCT结构成像、结膜微血管成像和传统裂隙灯成像功能。利用自主开发的软件，进一步分析结膜血管网络图像和血流图像，获取血管分形维数、血流速度、血管直径等定量形态学和血流动力学参数。结论：通过在健康受试者和角膜炎患者的在体成像测试，表明多模态眼前段成像设备可为眼科临床应用及人工智能提供结构和功能信息数据。

Objective: To obtain the comprehensive information of the anterior eye image, establish complementary information for the diagnosis and evaluation of ocular diseases. Methods: We developed a multi-modal, non-invasive optical imaging platform by combining ultra-high resolution optical coherence tomography (UHR-OCT) with a microvascular imaging system based on slit-lamp biomicroscopy. Results: The uHR-OCT module achieved an axial resolution of approximately 2 μm in tissues. The lateral resolution of the ocular surface microvascular imaging module under maximum magnification was approximately 3.5 μm. By combining the imaging optical paths of different modules, the customized multi-modal eye imaging platform was capable of performing real-time cross-sectional UHR-OCT imaging of the anterior eye, conjunctival vessel network imaging, high-resolution conjunctival blood flow videography, and traditional slit-lamp imaging on a single device. With self-developed software, a conjunctival vessel network image and blood flow videography were further analyzed to acquire quantitative morphological and hemodynamics parameters, including vessel fractal dimensions, blood flow velocity and vessel diameters. Conclusion: The ability of the multi-modal anterior eye imager to provide both structural and functional information for ophthalmic clinical applications can be demonstrated in a healthy human subject and a keratitis patient.

目的：分析人眼的睑板腺形态学特征，探索睑板腺分析系统在眼表疾病的应用研究。方法：中山眼科中心入组正常受试者24例(42眼)，进行睑板腺红外摄影。选取受试者中的10例(20眼)在同型号的设备上由二名操作员分别进行睑板腺红外摄影。图像通过自行设计的分析软件对上睑结膜中央5条腺体形态学参数进行定量分析，对数据进行重复性测试。结果：测量的生物参数腺体直径为(0.48±0.09) mm，腺体长度为(5.25±0.68) mm，腺体面积为(2.12±0.53) mm，腺体形变系数为10.01±3.85，显影值为6.32±1.23，中央五条腺体占中央区域面积百分比为(10.94±2.20)%，腺体占上睑结膜面积百分比为(58.07±8.13)%。各指标两次测量值差异无统计学意义(P>0.05)。重复性分析结果显示：腺体各项生物参数的变异系数(coefficients of variation，CV)均小于5%，组内变异系数(intraclass correlation coefficient，ICC)均大于0.95。结论：睑板腺综合分析系统对腺体的形态学分析有良好的可靠性和一致性，有望为临床上对睑板腺腺体功能评估提供新的非侵入性参考指标。

Objective: To analyze the morphological characteristics of meibomian glands in human eyes and to explore the application research of meibomian glands analysis system in ocular surface diseases. Methods: A total of 24 healthy subjects were recruited by Zhongshan Ophthalmic Center to infrared photography of meibomian glands. Ten of healthy subjects were selected by the two operators for infrared photography of meibomian glands on the same model of equipment. The images were repeatedly measured and analyzed by the self-designed analysis software on the morphological measurements of the five glands in the center of the upper eyelid. Results: The measured biological parameters are shown below: the average gland diameter was (0.48±0.09) mm, the average gland length was (5.25±0.68) mm, the average gland area was (2.12±0.53) mm, the gland deformation coefficient was 10.01±3.85, the development value was 6.32±1.23, the percentage of the five central glands in the central area was (10.94±2.20)%, and the glands accounted for (58.07±8.13)% of the upper conjunctiva area. There was no statistical difference between the two measurements of each index (P>0.05). Repeatability analysis results showed that coefficients of variation (CV) of all biological parameters of glands were less than 5% and the intraclass correlation coefficient (ICC) in both groups were greater than 0.95. Conclusion: The Meibomian Gland Bioimage Analyzer provides good reliability and consistency for morphological measurements of the meibomian gland, and it is expected to provide new non-invasive indicators for clinical assessment of the meibomian glands function.

目的：针对活体共聚焦显微镜(in vivo confocal microscopy，IVCM)和传统光学相干层析技术(optical coherence tomography，OCT)在人眼角膜成像各自存在成像视野小或无法细胞成像的限制，开发具有高分辨率的非接触全视场光学相干层析系统(full-field optical coherence tomography，FFOCT)，实现活体人眼角膜细胞结构FFOCT成像。方法：FFOCT系统采用高数值孔径干燥显微物镜及高速面阵相机，使用双相位调制图像处理方法，实现系统高速高分辨率非接触成像。利用系统对健康人眼进行角膜各深度层的活体FFOCT成像验证其可行性。结果：本研究团队研发了FFOCT的新型活体人眼角膜高分辨率成像系统，实现理论平面成像分辨率1.7 μm，成像视野1.26 mm×1.26 mm，成像速率达275帧/s。利用该系统对正常活体人眼角膜成像实验，在非接触情况下获取了角膜各主要结构层的高分辨率结构影像。结论：FFOCT高分辨率活体人眼角膜成像系统兼具了传统OCT的非接触、大成像视野及IVCM的细胞级别平面分辨率的优势，将为角膜疾病的研究及临床诊疗提供全新的成像分析技术。

Objective: Due to the limitations of small imaging field of view of in vivo confocal microscopy (IVCM) or the incapability of cellular imaging of traditional optical coherence tomography (OCT) in human corneal imaging, this study was designed to develop a novel high-resolution in vivo human corneal imaging system based on full-field OCT (FFOCT). Methods: The FFOCT system utilized a high numerical aperture air immersion microscope objective and a high-speed area array CMOS camera with two-phase modulation image processing algorithm to achieve high-speed high-resolution non-contact imaging of human cornea. To verify its feasibility, in vivo cornea imaging at different depth was performed on a healthy human subject. Results: The FFOCT system achieved a theoretical lateral imaging resolution of 1.7 μm, an imaging field of view of 1.26 mm×1.26 mm, and an imaging rate of 275 Hz/s. High-resolution FFOCT images of the main structural layers of cornea were achieved by imaging a healthy human cornea in vivo with this system in a non-contact way. Conclusion: The FFOCT human corneal imaging system combines the advantages of the non-contractness and the large imaging field of view of traditional OCT with the cellular lateral resolution of IVCM, potentially providing a new imaging system for the research and clinical diagnosis and treatment of corneal diseases.

目的：对视网膜光学相干断层扫描图像中不同层和积液区域的分割。方法：提出一种基于深度学习的轻量级的神经网络，参考DRUNet体系、膨胀卷积和残差网络的架构，通过连接不同深度网络处得到的上采样输出，进行多尺度特征融合，使网络能够更好地识别出图像中的边界信息。结果：改进型DRUNet显著提升了视网膜分层的效果，准确率较U-Net提高了1.25%，同时能提前1~2次迭代达到传统U-Net的准确度。结论：本文采用的网络结构提高了对视网膜光学相干断层扫描图像的分割性能，同时降低了网络参数，具有强大的应用潜力。

Objective: To achieve the segmentation of different layers and fluid areas on the optical coherence tomography (OCT) image of the retina. Methods: A lightweight neural network based on deep learning was proposed. The network structure adopted in this study was designed based on the architecture of dilated-residual U-Net. By connecting the upsampling output obtained at different depth networks, multi-scale feature fusion was performed to enable the system to accurately identify the boundaries on the OCT image. Results: Compared with U-Net, this algorithm could achieve the same accuracy with 1–2 epochs less, and the accuracy was also improved by 1.25%. Conclusion: The proposed network improves the segmentation performance of retinal OCT images, and reduces the number of parameters, which demonstrates the network has great application potential.