基于免疫遗传形态学的视网膜光学相干断层图像边缘检测方法

提出了基于免疫遗传算法的形态学自适应结构元素生成算法,并将其用于光学相干断层成像(optical coherence tomography,OCT)图像中视网膜组织边缘检测.首先将图像进行去噪和粗分割的预处理,并将图像划分为若干子图像;其次对每一子图利用免疫遗传算法求取自适应结构元,初始随机生成固定长度的二进制数串作为抗体,并将其转化为结构元素格式,以图像二维熵定义抗体适应度,根据子图像本身结构特征信息,寻找最优抗体结构元素;最后利用寻优得到的各结构元素对子图进行形态学边缘检测,合并各子图的分割结果,实现整体图像目标边界提取.实验结果表明了该方法在图像目标边界提取的有效性.     

基于免疫遗传形态学的视网膜光学相干断层图像边缘

提出了基于免疫遗传算法的形态学自适应结构元素生成算法,并将其用于光学相干断层成像(optical coherence tomography,OCT) 图像中视网膜组织边缘检测. 首先将图像进行去噪和粗分割的预处理,并将图像划分为若干子图像; 其次对每一子图利用免疫遗传算法求取自适应结构元,初始随机生成固定长度的二进制数串作为抗体,并将其转化为结构元素格式,以图像二维熵定义抗体适应度,根据子图像本身结构特征信息,寻找最优抗体结构元素; 最后利用寻优得到的各结构元素对子图进行形态学边缘检测,合并各子图的分割结果,实现整体图像目标边界提取. 实验结果表明了该方法在图像目标边界提取的有效性.      

基于图像重建方法的岩体损伤形态表征研究*

工程岩体内部的损伤形态难以表征。基于图像三维重建,结合数字图像处理技术、数值模拟,重建岩体损伤模型,通过对指定数值模型剖面图的损伤区域进行图像辨识和量化,重建岩体内部损伤的三维形态。结果表明:(1)所重建的三维损伤模型具有良好的可观察性,能够直观地观察到在复杂应力条件下岩体内部模型损伤结构的起裂位置和发展形态;(2)重建模型较好地量化了岩体损伤的程度、预测了损伤区域的发展;(3)重建方法不受尺寸效应约束,具有良好的适用性。研究成果对于认识岩体损伤演化行为、分析工程问题具有一定的参考意义。     

焊管辊成形过程的计算机仿真

利用文献（１）提出的仿真方法，对碳钢和不锈钢直缝焊管的辊成形加工过程进行了系统的描述，分析了辊轮间距离，辊轮数及管壁的厚度与直径比对成形方向膜应变最大值（εｘｏ）ｍａｘ的影响，探讨了在给定的最大膜就变平均值（Ｅｘｏ）ｍａｘ情况下管径与带料厚度在成形中的关系，所得结果与实际生产中的变形特征吻合良好。     

影响碲膜光学性能的因素

用XRD，SEM，FTIR光谱仪、UV／VIS／NIR光谱仪表征了由化学气相沉积制备的碲薄膜的结构和光学性能，结果表明碲膜的光学性能受膜的厚度、基体表面性质和膜的表面形貌的影响。沉积在裸聚乙烯箔基体上的碲膜（膜厚293nm）不仅在8-13μm光谱范围的透过率比沉积在Mn-0覆盖聚乙烯箔基体上的碲膜（膜厚216nm）的透过率高，而且还能有效地阻挡太阳光的透过，这些性能说明具有该厚度的碲膜可以应用于太阳光辐射屏与辐射制冷设备。     

基于图像处理的冷轧薄板板形识别

本文提出一种新的基于图像处理的冷轧薄板板形识别方法。首先对图像进行直方图均衡化，使处理后的图像对比度、图像边界的清晰度提高，然后再进行形态学增强，改善图像效果，对于处理后的图像利用canny算子提取其边缘，利用图像的均值、方差和对比度的统计特征作为人工神经网络分类器的输入进行缺陷特征分类。根据上述方法进行了板形识别系统的硬件、软件设计，实际应用表明，本文提出的方法可以有效地识别出常见的板形缺陷。     

面向视网膜脱离手术的硅油填充模拟

针对传统医疗手段无法有效量化评估手术中不同硅油加注量对于视网膜裂孔贴附效果的问题，本文提出一种面向视网膜脱离手术的硅油填充模拟方法，基于物理建模与计算机数值离散化技术对眼内受力、硅油填充状态进行分析，并对填充模拟过程进行三维模型构建与可视化，实现医疗过程决策辅助目的。首先对人类眼球与手术器具进行基础建模与模型采样，模拟手术流程中眼球状态；然后，根据水与硅油的密度、黏滞系数、表面张力等不同物理性质，对水?硅油两相流动及交互进行模拟；最后，构建固液交互模型，实现多相液体在眼球中的运动与填充。实验结果表明，本文方法能够较好地呈现眼球内多相流体运动交互效果，实现了诸如表面张力、固液耦合、液体分层、连通器效应等效果，实现了对眼内腔中通过导管注入硅油与排出水分流程的模拟，为预测硅油填充后的眼内状态提供了一种有效的方式，辅助医生进行手术流程规划与效果预测。      

基于监督学习的前列腺MR/TRUS图像分割和配准方法

前列腺核磁超声图像配准融合有助于实现前列腺肿瘤的靶向穿刺。传统的配准方法主要是针对手动分割的前列腺核磁（Magnetic resonance, MR）和经直肠超声（Trans-rectal ultrasound, TRUS）图像上对应的生理特征点作为参考点,进行刚体或非刚体配准。针对超声图像因成像质量低导致手动分割配准效率低下的问题,提出一种基于监督学习的前列腺MR/TRUS图像自动分割方法,与术前核磁图像进行非刚体配准。首先,针对图像分割任务训练前列腺超声图像的活动表观模型（Active appearance model, AAM）,并基于随机森林建立边界驱动的数学模型,实现超声图像自动分割。接着,提取术前分割的核磁图像与自动分割的超声图像建立轮廓的形状特征矢量,进行特征匹配与图像配准。实验结果表明,本文方法能准确实现前列腺超声图像自动分割与配准融合,9组配准结果的戴斯相似性系数（Dice similarity coefficient, DSC）均大于0.98,同时尿道口处特征点的平均定位精度达1.64 mm,相比传统方法具有更高的配准精度。      

采动影响型地下煤火诱发地表裂隙率的时空分布模型

基于随机介质移动理论,构建了采动影响型地下煤火诱发地表裂隙率的时空统一分布模型,并实例分析了矩形火区引发地表线（张）裂隙率、面裂隙率以及剪裂隙率的分布及动态变化规律.地表线（张）裂隙率、面裂隙率和剪裂隙率的极大值分别分布在采（燃）空区边界内侧约20 m（约为煤层厚度的3~4倍）的位置、四周边界线的四个中点位置和四个边角端点所对应的地表区域.随着煤层燃烧,垂直于煤火发展方向上的线裂隙率以及空区边界处对应的地表面裂隙率均呈半正态曲线形式变化并最终稳定于最大值;而剪裂隙率、煤火发展方向上的线裂隙率及空区内部对应的地表面裂隙率均呈正态曲线形式变化.      

基于因果推断的冷轧带钢厚度异常根因分析

厚度精度是衡量冷轧带钢质量的重要指标之一，快速诊断带钢厚度异常并定位异常发生的根本原因，对提升带钢质量及生产稳定性具有非常重要的意义。为此，通过轧制机理模型确定厚度影响因素，采用多元线性回归方法构建厚度增量残差模型，对残差进行核密度估计来检测厚度异常；针对厚度异常，基于因果推断计算影响因素的因果效应，辨识出厚度异常的根本原因。实际应用结果表明，该方法能准确地诊断厚度异常和辨识异常的根本原因，与常规方法相比，其对冷连轧过程变量之间具有高度相关性的工况更为有效。     

A comparison of retinal morphology viewed by optical coherence tomography and by light microscopy

OBJECTIVE: To compare the cross-sectional images of primate retinal morphology obtained by optical coherence tomography (OCT) with light microscopy to determine the retinal components represented in OCT images. METHODS: Laser pulses were delivered to the retina to create small marker lesions in a Macaca mulatta. These lesions were used to align in vivo OCT scans and ex vivum histologic cross sections for image comparison. RESULTS: The OCT images demonstrated reproducible patterns of retinal morphology that corresponded to the location of retinal layers seen on light microscopic overlays. Layers of relative high reflectivity corresponded to horizontally aligned retinal components such as the nerve fiber layer and plexiform layers, as well as to the retinal pigment epithelium and choroid. In contrast, the nuclear layers and the photoreceptor inner and outer segments demonstrated relative low reflectivity by OCT. CONCLUSIONS: Retinal morphology and macular OCT imaging correlate well, with alignment of areas of high and low reflectivity to specific retinal and choroidal elements. Resolution of retinal structures by OCT depends on the contrast in relative reflectivity of adjacent structures. Use of this tool will enable expanded study of retinal morphology, both normal and pathologic, as it evolves in vivo.     

Macular imaging with optical coherence tomography

PURPOSE: Optical Coherence Tomography (OCT) is a novel noninvasive and noncontact imaging technique providing cross-sectional representations of the eye structures. OCT is analogous to Ultrasound B-scan, except that it analyzes the reflection of a 850 nm light wave. The aim of this study was to assess the potential of ocular coherence tomography for diagnosing and monitoring macular diseases. METHODS: Cross-sectional images were performed with the Zeiss-Humphrey OCT. Over one year period, we examined approximately 300 patients with idiopathic full thickness macular hole, lamellar hole, cystoid macular edema, choroidal new vessels, epiretinal membrane, diabetic maculopathy, and central serous chorioretinopathy. RESULTS: OCT can provide new information concerning the posterior pole diseases mentioned above. OCT can also be useful in thickness measurements. CONCLUSION: OCT allows tomographic analysis of macular diseases. The information obtained is different from that obtained by histologic study which is sometimes hard to interprete. OCT is mostly useful in studying internal layers of the retina. Further applications may be developed.     

A new method for noninvasive optical sectioning of the chorioretinal vasculature

PURPOSE: To report a new method for optical sectioning of the chorioretinal vasculature to improve the visualization of vascular abnormalities due to chorioretinal eye diseases. METHODS: An imaging system was developed for optical sectioning of the vasculature called chorioretinal optical sectioning (CROS). CROS consists of projecting a laser beam at an angle on the retina after injection of a fluorescent dye and viewing the fluorescence. On the fluorescence optical section (FOS) image, the vasculature of the retina and choroid appear laterally displaced according to their depth location. The laser beam is scanned over a 2 X 2-mm area to generate 40 FOS images, each spatially separated by 50 microm on the retina. Optical section images of the vascular layers are constructed from the series of FOS images. RESULTS: CROS permitted optical separation of vascular layers in living eyes. Optical section images of normal and laser-photocoagulated retinas had higher contrast than conventional angiography because of the separation of the fluorescence from the overlapping layers and allowed enhanced visualization of vascular abnormalities. CONCLUSIONS: CROS enhances the visualization of the retinal and subretinal vasculature and promises to be a beneficial tool for evaluation of chorioretinal diseases.     

Use of the astigmatism correction device on the Zeiss fundus camera for peripheral retinal photography

High quality photographs of the peripheral retina are made possible by the proper positioning of the astigmatism correction device on the Zeiss fundus camera. Using a plus cylinder, axis perpendicular or a minus cylinder, axis parallel to that around which the eye is rotated will eliminate the astigmatism induced by viewing the retina obliquely through the optical surfaces of the eye, resulting in a sharply focused image at the film plane. Proper use of this instrument can be helpful in evaluating and diagnosing retinal lesions that occur in the peripheral fundus.     

Evaluation of the retinal nerve fiber layer

In normal eyes, the retinal nerve fiber layer (RNFL) is usually best visible in the inferior temporal part of the fundus, followed by the superior temporal region, the nasal superior region and the nasal inferior region. This distribution correlates with the configuration of the neuroretinal rim, the diameter of the retinal arterioles, the location of the foveola, and the lamina cribrosa morphology. With increasing age, the RNFL visibility decreases diffusely without preferring special fundus regions and without the development of localized defects. With all optic nerve diseases, the visibility of the RNFL is decreased in addition to the age-related loss, in a diffuse and/or a localized manner. The localized defects are wedge-shaped and not spindle-like defects, running toward or touching the optic disk border. Typically occurring in about 20% of all glaucoma eyes, they can be found also in other ocular diseases, such as optic disk drusen, toxoplasmotic retinochoroidal scars, longstanding papilledema or optic neuritis due to multiple sclerosis. Since they are not present in normal eyes, they almost always signify an abnormality. RNFL evaluation is especially helpful for early glaucoma diagnosis and in glaucoma eyes with small optic disks. In advanced optic nerve atrophy, other examination techniques, such as perimetry, may be more helpful for following optic nerve damage. Considering its great importance in the assessment of optic nerve anomalies and diseases and taking into account the feasibility of its ophthalmoscopic evaluation using green light, the retinal nerve fiber layer should be examined during any routine ophthalmoscopy.     

Electroretinogram: technique and clinical applications

INTRODUCTION: The retina is a neuronal membrane that covers the posterior part of the eye chamber and is formed by several layers. The retinal function can be synthesize as the light translation into nervous impulses in the optic nerve. MATERIAL AND METHODS: The ERG is the expression of the changes of the electrical potential in the retina after a luminous stimulus. There are different types of retinal stimulators specially flash, ganzfeld and pattern which are the most used in the clinical neurophysiological explorations. RESULTS: We expose the obtained responses, the morphology, amplitude and latencies of the (a) and (b) waves in scotopic and photopic conditions with ganzfeld or flash stimulator. In the ERG pattern we describe the waves for the transitory ERGP as well as for the Steady-State ERGP. We refer to the protocols proposed by the International Society of Electrophysiology of Vision (ISCEV). We mention the clinical applications of these techniques and we give our personal experience in diseases such retinitis pigmentosa, maculopathies, retinal degenerations, vascular diseases, cataracts and traumatic lesions of the retina.     

Infrared imaging of sub-retinal structures in the human ocular fundus

The interaction of infrared light with the human ocular fundus, particularly sub-retinal structures, was studied in vivo. Visible and infra-red wavelengths and a scanning laser ophthalmoscope were used to acquire digital images of the human fundus. The contrast and reflectance of selected retinal and sub-retinal features were computed for a series of wavelengths or modes of imaging. Near infrared light provides better visibility than visible light for sub-retinal features. Sub-retinal deposits appear light and thickened; the optic nerve head, retinal vessels, and choroidal vessels appear dark. Contrast and visibility of features increases with increasing wavelength from 795 to 895 nm. Optimizing the mode of imaging improves the visibility of some structures. This new quantitative basis for near infrared imaging techniques can be applied to a wide range of imaging modalities for the study of pathophysiology and treatment in diseases affecting the retinal pigment epithelium and Bruch's membrane, such as age-related macular degeneration.     

Influence of intercellular tissue connections on airway muscle mechanics

PURPOSE: To examine immunohistochemical markers in straight, well-laminated retinal transplants with special attention paid to the interphotoreceptor matrix, the Müller cells and the ganglion cells as these three retinal components have been abnormal in transplants produced by previous methods. METHODS: Nine rabbits underwent subretinal transplantation of a complete full-thickness embryonic neuroretina. After 31 or 49 days, the transplants were stained for light microscopy and processed for immunohistochemistry. RESULTS: Six of 9 eyes contained transplants with straight, well-laminated regions with all light-microscopic characteristics of a normal retina. In the outer segment region, the expression of peanut agglutinin showed segmental labeling of cone domains in the interphotoreceptor matrix, and interphotoreceptor retinoid binding protein immunoreactivity was found. Glial fibrillary acidic protein and vimentin immunoreactivity revealed normal Müller cell morphology. In 3 transplants the AB5-antibody-labeled ganglion cells in the ganglion cell layer and all transplants contained nerve fibers in the nerve fiber layer labeled by an antibody against neurofilament of 160 kD. The latter also labeled fibers connecting the transplant with the host. CONCLUSIONS: Full-thickness embryonic retinal transplants develop the normal retinal appearance and display several of the retinal components necessary for normal function which are not found in transplants produced by previous methods.     

Quantitative image sequence analysis of fundus fluorescein angiography

Interpretation of retinal angiographic studies has heretofore been largely qualitative. The temporal properties associated with fluorescein fluorescence of normal and pathologic fundus features are quantified to potentiate pixel assignment and fundus feature quantitation for clinical studies where precise image metrology is vital. Fluorescein angiography studies were digitized, and temporally sequential images were spatially registered with polynomial warping algorithms, allowing for the construction of a three-dimensional angiogram vector. Temporal profiles through spatially registered, temporally sequential pixels were computed. Fundus feature fluorescence behavior was quantified. Fundus features may be discriminated on the basis of spatio-temporal fluorescein fluorescence properties. Pixel assignment based on spatio-temporal relationships will facilitate fundus feature quantitation.