基于光学相干层析的视网膜图像分割

利用光学相干层析（optical coherence tomography,OCT）技术可以得到清晰的视网膜层状结构,实现视网膜层状结构自动分割功能是解决OCT技术应用于视网膜疾病诊断的一项基础问题。本文通过图像平滑、峰值探测、Snake模型、贪婪算法和样条插值等综合技术,对OCT视网膜图像进行分割,自动提取层状结构轮廓并获取视网膜厚度分布图。将以上算法应用于24例正常人眼底图像,并与专家手动标记轮廓提取的厚度相比,结果证实上述视网膜自动测量算法与专家人工标记取得较好一致性。本文提出的测量算法有望应用于视网膜变异性评估。     

基于Contourlet变换和PCNN的CT图像椎体解剖轮廓特征提取方法的研究

提出一种新的基于Contourlet变换和脉冲耦合神经网络(PCNN)的医学图像解剖轮廓特征提取算法。首先对原始椎体CT图像进行Contourlet变换,得到能稀疏表示图像边缘以及方向信息的子带和低频子带;然后结合PCNN对低频子带进行边缘轮廓细节提取,最后利用处理后的所有子带系数,通过Contourlet逆变换,提取出图像的边缘轮廓。实验将本算法提取的结果与Canny算子、区域生长法以及结合小波变换和PCNN的算法提取的图像边缘轮廓进行比较,结果表明新算法能够有效的实现医学图像解剖结构轮廓特征的提取。     

超声图像处理中Snake模型研究

Snake模型是一种基于高层信息的有效目标轮廓提取算法,其优点是作用过程及最后结果的目标轮廓是一条完整的曲线,从而引起广泛的关注.鉴于医学超声图像的信噪比较低,用经典的边缘提取算法无法得到较好的结果,因此人们将Snake模型进行了各种各样的改进,并且越来越多地将它运用到医学超声图像处理中来.本文对乳腺超声图像进行阈值分割、形态滤波等一系列预处理后,将改进的Snake模型对乳腺超声图像进行肿瘤的边缘提取,得到了比较好的结果.     

863计划“活体人眼视网膜细胞自适应光学成像仪研制及临床应用”课题已通过验收

在国家863计划、中科院知识创新工程、国家自然科学基金等支持下,中科院成都光电所将自行建立的世界上第一套基于19单元整体集成式微小变形镜的轻小型“活体人眼视网膜自适应光学成像系统”做了改进,研制出基于37单元微小变形镜的“活体人眼视网膜细胞自适应光学成像仪”。经实验应用,可稳定获取更高分辨率的相对于黄斑中心凹不同区域、不同层面的视网膜细胞图像,以及更高分辨率的眼底视网膜毛细血管图像。     

MRI图像中颅骨不连续外轮廓的多分辨率提取

目前,许多医学图像处理（例如图像三维重构,图像建模,图像配准和图像融合等）都需要进行图像的边缘提取,因此,边缘提取方法在医学图像处理中有极其重要的意义,轮廓闭合图像的边缘比较容易提取,而解剖和外伤等原因造成轮廓不连续时边缘提取比较复杂,美国和欧洲等发达国家从20世纪90年代后期开始着手医学图像不连续边缘提取方法的研究,国内的有关研究仍少见报道,本课题以脑图像为例,试图综合运用多分辨率法与八邻距离转换法来提取脑图像的不连续外轮廓,研究结果表明,此方法在提取不连续脑图像的连缘时快速有效,同时可以去掉高频干扰,有利于图像的配准。     

基于视通路多感受野朝向性关联的轮廓检测方法

基于视通路中各层次的感受野特性以及层次间感受野的关联特性,提出一种图像轮廓检测新方法。根据神经元的放射状树突接收特性,构建检测空间差异性信息的视网膜神经元网络,对轮廓信息进行前级编码;建立LGN细胞的非经典感受野调节机制,结合感受野的层次变换特性,对前级编码结果进行全局调节;提出简单细胞的多感受野朝向性关联模型,模拟初级视皮层简单细胞的方向选择特性;融合多个方向上的轮廓响应,经过非极大值抑制和阈值处理,得到轮廓检测结果。以RuG图库40幅图片为实验对象,检测结果与基准轮廓图的平均P指标为0.43,结果显示该方法能够有效凸显主体轮廓,强化轮廓与纹理边缘的区分度。利用多感受野的层次关联特性,实现图像轮廓信息的编码与检测应用,为研究高级视皮层的图像理解和视觉认知提供新的思路。     

视网膜中央动脉阻塞性缺血对黄斑区视网膜神经上皮的影响

目的:应用光学相干断层扫描(OCT)探讨视网膜中央动脉阻塞(CRAO)性缺血对黄斑区视网膜组织结构的影响。方法:用OCT对单眼CRAO 2-3d,黄斑区视网膜血液循环未完全恢复的患者 14例(14眼 )和 11例(11眼 )对侧健眼进行经中心小凹的水平和垂直扫描,分别测量及计算中心小凹、中心凹及黄斑的平均视网膜神经上皮层(RNL)厚度。结果:正常眼中心小凹、中心凹及黄斑平均RNL厚度分别为(169.91± 10.96) μm、(176.36± 11.74) μm、(256.45± 16.95) μm;CRAO黄斑区RNL明显增厚,中心小凹、中心凹及黄斑平均RNL厚度分别为(235.64± 47.02) μm、(241.84± 49.36) μm、(401.57± 54.53) μm,与正常眼比较有显著性差异(P <0.05,P <0.01)。 结论:CRAO性缺血明显地改变了黄斑区正常视网膜组织结构,使视网膜神经上皮层细胞水肿.     

基于小波变换和拉普拉斯算子的细胞图像边缘检测方法

目的：介绍一种基于小波变换和拉普拉斯算子的血液细胞图像边缘识别方法。材料和方法：取正常人血样5mL。制成血液细胞图片12片，对血图像进行预处理后，利用小波变换的多分辨率特性滤除细胞图像中的干扰成份。根据血液细胞边缘附近的灰度分布梯度较大的特点，采用拉普拉斯算子及双阈值法对其进行边缘检测和识别。结果和结论：实验结果表明，结合小波变换和拉普拉斯算子的边缘提取算法对血液细胞图像边缘提取有良好的效果．为下一步对血液细胞的形态学分析、分类和识别提供了新途径。     

针对颅脑CT的边缘提取

边缘提取是ＣＴ图像三维重建前期工作中的关键步骤。本文根据图像的统计信息，一方面利用类间方差测度准则，求取图像的灰度阈值；另一方面求取图像的梯度阈值，利用这两个阈值将颅脑ＣＴ图像中各器官的边缘提取出来，作为三维重建的轮廓输入。实验结果表明：这种方法实时性好，适应性强，提取的边缘清晰完整，准确性高，证明了该算法的实用性和可行性     

基于抑制性突触多层神经元群放电编码的图像边缘检测

图像边缘检测所获得的目标轮廓细节质量,对于后续图像分析或理解过程具有重要作用。提出一种基于视觉机制的图像边缘检测新方法:构建抑制性突触连接的多层神经元群,在待检测图像的激励下,分析7像素×7像素视觉感受野窗口内互连接神经元的脉冲放电过程,记录发放时刻进行次序编码;同时考虑神经元之间的侧向抑制作用,在选择注意机制作用下获得增强图像;之后利用Log Gabor滤波器模拟视觉系统中的方向选择特性,获取8个方向的滤波结果,经过输出层神经元群的融合处理并经灰度映射到0～255的范围后获得边缘图像。对含有丰富边缘细节特性的24幅菌落图像进行处理,其处理结果的ROC评价指标均值为0.698 4,优于PCNN法的0.659 3;从评价指标的均方差来看,该结果具有更好的一致性。另外,从信息熵评价指标来看,该方法同样具有一定的优势,能够有效提取图像的边缘信息,而且也能反映更多层次的图像细节。所提出的方法为利用视觉生理特性进行图像处理提供了崭新而有效的思路。     

Retinal cone photoreceptor distribution in the American black bear (Ursus americanus)

The distribution of cone photoreceptor subtypes (important for color vision and vision quality) varies widely in different carnivore species, but there have been limited studies on bear (ursid) cone distribution. A previous behavioral study suggests that American black bears (Ursus americanus) are dichromatic, indicating that they possess two cone subtypes, although the retinal distribution of cones is unknown. The purpose of this study was to examine the subtype and topography of cones in American black bear retinas to further predict the nature of their color vision and image resolution. We studied 10 eyes from seven individual legally hunted black bears in northeastern North Carolina. Cryosections and retinal wholemounts were labeled using antibodies targeting two cone opsin subtypes: long/medium (L/M) wavelength sensitive and short (S) wavelength sensitive. Cones in fluorescent microscopy images were counted and density maps were created for retinal wholemounts. The black bear retina contains both cone subtypes and L/M cones outnumber S cones by at least 3:1, a finding confirmed in retinal frozen sections. There are higher concentrations of S cones present than typically seen in other carnivores with some evidence for co-expression of L/M and S cones. A cone-dense area centralis is present dorsotemporal to the optic nerve, similar to other carnivores. These results confirm that American black bears are predicted to have a dichromatic vision with high acuity indicated by the presence of a dorsotemporally located area centralis.     

Quantitative morphology of the central fovea in the primate retina

Electron micrograph composites of tangential sections of the fovea centralis of three cynomolgus monkeys (Macaco, irus) and one baboon (Papio anubis) were used to determine the spatial density of the principal retinal cells. In the center of the foveola, the density of cones ranged from 113,000 to 230,000/mm², and pigment epithelial cells from 4,900 to 7,000/mm². At a distance of 500μm from the foveolar center the density of the cone cell pedicles ranged from 29,000 to 36,300/mm², and the density of horizontal cells ranged from 19,000 to 25,100/mm². Densities of bipolar, Muller, and amacrine cells were determined in only two monkeys and in the baboon. The fact that the cone cell pedicles have a larger diameter than the foveolar cones explains the geometry of the fovea. The morphology of the junction between foveolar cone outer segments and the pigment epithelium reflects the complex metabolism of this functional unit. The comparison with the peripheral primate retina suggests that the densities of horizontal and bipolar cells, but not of amacrine and Muller cells, are correlated with the density of cone cell pedicles.     

The simple perfection of quantum correlation in human vision

A theory is presented that specifies the amount of light that is needed for the perception of any stimulus that is defined in space, time and color. For detection and discrimination mechanistic neural elements with deterministic procedures exist. Twin pairs of red and green cones are ordered in three sets along clockwise and counter clockwise revolving spirals and along circles around the center of the fovea. In the rod-free fovea the red pairs are ordered along the spirals and the green along the circles. Each cone is accompanied by--dependent on retinal eccentricity--up to 100 satellite rods. For the retinal signal processing such a receptor group constitutes a space-quantum in analogy with time-quanta of about 0.04 s. In the peripheral retina the red and green twin pairs of space-quanta are roughly ordered along and at random distributed over the spirals and circles. Over each time-quantum, the cone and rods of a space-quantum sum their responses in a common nerve circuit of the luminosity channel. The summation's results from twin pairs of the same set of space-quanta are correlated by two-fold spatio-temporal coincidence mechanisms in the retina. Their outcome signals the perception of light, movement and edge. In the fused binocular visual field the movement and edge signals of the three sets from both eyes perfectly join vectorially together, provided the responding pairs of space-quanta are binocularly in perfect register as they normally are. The receptor's Weber gain control makes the receptor an all-or-none-system. The space-quantum's De Vries gain control makes its sensitivity equal to the average of the poisson fluctuations in quantum absorption per time-quantum. The controls are based on, respectively, arithmetically feed forward and backward inhibitive nerve mechanisms. The thermal noise of the photo-pigment resets the controls. The response to the second quantum absorption in a time-quantum in the individual rod, red or green cone has accession to the white, red or green nerve color circuit, respectively, and produces there a corresponding color signal. Already a single absorption in a blue cone is for a blue signal. In the retina, for the generation of yellow signals, the color circuits of individual red and green cones of each mixed entwined triple of red and green twin pairs of space-quanta are cross-connected through a nerve opponent color circuit. In the lateral geniculate nucleus in groups of seven neighboring triples, through two nerve opponent color circuits that are common for the two eyes together, the red and green signals as well as the yellow and blue mutually annihilate each other's color. White signals remain. In anomalous trichromacy, the space-quanta of some pairs have different cones or in one of them the cone is missing. In dichromacy, all pairs have different cones or one type of cones is missing. For perceptive resolution the periodic scanning of the retinal image by the eye tremor in synchrony with the time-quanta, overrules the limit of optical resolution as set by diffraction in the eye optics. Dependent on pupil diameter the scanning contributes up to a factor of about 30 to resolution. The action potentials of the Purkinje cells in the myocardium generate the time-quanta of the central nervous system as well as the mechanical scanning of the retinal image through the synchronic periodic variation of the tonus in the eye muscles.     

Topography of the retina and striate cortex and its relationship to visual acuity in rhesus monkeys and squirrel monkeys

Summary Density of cones and ganglion cells was studied in horizontal sections of retina in the rhesus monkey (Macaca, mulatta) and the squirrel monkey (Saimiri sciureus). The lower angular density of cones in the fovea of Saimiri correlates with its visual acuity, which is poorer than that of Macaca (0.74 and 0.65 min of arc respectively). Cone density falls more steeply with angular eccentricity from the fovea in Saimiri, in accordance with its relatively poorer peripheral acuity. Comparable results were obtained with retinal ganglion cells, but the comparisons at the fovea itself are more difficult because of the lateral displacement of these elements in the foveal region.The cortical magnification of the visual field (that is, the number of mm of cortex per degree of visual field) is lower for both the foveal and parafoveal representations at the striate cortex in Saimiri. This was correlated with its poorer foveal and parafoveal acuity. It was shown that with increasing eccentricity from the fovea, the fall in the magnification of the visual field at the striate cortex is approximately proportional to the decrease in ganglion cell density at the retina. The results of this study, in which acuity and topography of the visual system are compared in two species of monkey, are consistent with the view that both retinal topography, and the cortical magnification of the visual field, are closely related to visual acuity.This work was supported by Medical Research Council Grant G.967/2/B. We wish to thank Mrs L. Bowman and Mr D. Canwell for their help in preparing histological material.     

Death by color: differential cone loss in the aging mouse retina

Differential cell death is a common feature of aging and age-related disease. In the retina, 30% of rod photoreceptors are lost over life in humans and rodents. However, studies have failed to show age-related cell death in mouse cone photoreceptors, which is surprising because cone physiological function declines with age. Moreover in human, differential loss of short wavelength cone function is an aspect of age-related retinal disease. Here, cones are examined in young (3-month-old) and aged (12-month-old) C57 mice and also in complement factor H knock out mice (CFH−/−) that have been proposed as a murine model of age-related macular degeneration. In vivo imaging showed significant age-related reductions in outer retinal thickness in both groups over this period. Immunostaining for opsins revealed a specific significant decline of >20% for the medium/long (M/L)-wavelength cones but only in the periphery. S cones numbers were not significantly affected by age. This differential cell loss was backed up with quantitative real-time polymerase chain reaction for the 2 opsins, again showing S opsin was unaffected, but that M/L opsin was reduced particularly in CFH−/− mice. These results demonstrate aged cone loss, but surprisingly, in both genotypes, it is only significant in the peripheral ventral retina and focused on the M/L population and not S cones. We speculate that there may be fundamental differences in differential cone loss between human and mouse that may question the validity of mouse models of human outer retinal aging and pathology.     

Is abnormal retinal development in albinism only a mammalian problem? Normality of a hypopigmented avian retina

The central retina in hypopigmented mammals is underdeveloped. In the outer retina this deficit is confined to rods. Also, many ganglion cells in temporal regions project inappropriately to the contralateral hemisphere. This study addresses the question of whether pigment-related abnormalities occur in the central retina of a non-mammal, the bird. Birds have a highly developed central retina, but unlike most mammals they do not have a significant uncrossed retinal projection. Consequently, examination of the retinae of hypopigmented birds will reveal whether there is a relationship between the two abnormalities. Also if one of the primary effects of albinism is centred on rods, then albino birds may not show a deficit, because their retinae are cone dominated. Retinae from normally pigmented and two forms of hypopigmented budgerigars (Melopsittacus undulatus) were studied. Measurements of layer thickness, cell density and cell size were made at a range of locations in the ganglion cell layer and in the inner and the outer nuclear layers. Estimates of cone numbers were also made. Each strain of bird had an area of increased retinal layer thickness in dorso-temporal regions, but not a fovea. Although there were variations in the measurements undertaken between the strains, none were pigment related or consistent with the abnormality found in the central retina in albino mammals. Consequently, the underdevelopment of the central retina seen in hypopigmented mammals does not occur in this bird. There are two possible explanations for this result. First, normal mammalian retinal development may depend partly on time-dependent interactions in the maturation of the retinal pigment epithelium and the neural retina. Although there is a common time table for the development of the mammalian visual system when expressed in terms of the caecal period, which is between conception and eye opening, the pace of retinal development in birds is accelerated, which may alter interactions between these regions. Second, as the bird retina is cone dominated, any deficits in albino strains may be relatively minor.     

The morphology and distribution of photoreceptors in the retina of Bufo marinus

Summary The number, cell morphology and retinal distribution of rods and cones were determined in the retina of the toad, Bufo marinus. Adult animals were sacrificed, both eyes were removed and prepared for either tangential section across the outer segments of the photoreceptor layer, or transverse section across the whole retina.Cone densities increased from an average of 7000/mm² in the peripheral to a maximum of 25000/mm² in the central retina. The high cone densities extended across the naso-temporal axis of the retina corresponding to the position of the visual streak in the ganglion cell layer. The total number of cones in the retina was estimated to be 1.1 million. Rod density of 21000/mm² in the central retina decreased to 17000/mm² at 1.5–4 mm eccentricity, and then increased to 29000/mm² in the peripheral retina. The total number of rods amounted to about 2 million. The mean of the crosssectional area of rod outer segments was 11.2 ± 1.5 m² (mean ± SD) in the highest and 17.9 ± 4.7 m² in the lowest density areas of the retina. The length of the rod outer segments extended from 28 m in the ventral peripheral retina to a maximum of 89 m in the dorsal retina, dorsal to the visual streak of the ganglion cell layer.The results of the present study showed a differential retinal distribution of photoreceptors, with a peak density in the retinal centre and a higher density along the naso-temporal axis of the eye. We conclude that the area of high photoreceptor density, matched by high neuron densities of the INL and GCL, corresponds to the site of acute vision of the Bufo retina.On leave from Department of Biology, Fujian Teachers University, Fuzhou, Fujian, People's Republic of China.     

Dependence of the magnitude of the Stiles—Crawford effect on retinal location

1. The directional sensitivity (Stiles—Crawford effect) of retinal cones is supposed to be associated with their shape, but only extrafoveal cones have a cone-like shape; cones in the central fovea are elongated and look like rods.

2. To determine whether the directional sensitivity of cones depends on their shape, the Stiles—Crawford effect was measured both in the central fovea and in the parafovea of the human eye.

2. To ensure that the cone population tested was homogeneous, a small brief test flash, brought into the eye through the centre of the pupil, was placed at threshold by varying the intensity of a large adapting field. The directional sensitivity of the cones was determined by finding the efficiency of light to act as an adapting background as a function of position of entry in the pupil.

4. Central foveal cones have a less pronounced directional sensitivity than parafoveal cones and this lends support to the conclusion that the Stiles—Crawford effect is connected with the shape of the retinal receptors.

     

National Society for Histotechnology 31ST Annual Symposium/Convention Program Preview

Abstract

Microscopic evaluation of the macular region of the retina is essential in preclinical drug safety studies in cynomolgus monkeys. A number of therapeutic agents are suspected of causing damage to the macula in humans; therefore, the macular region of the retina should be evaluated in preclinical drug safety studies in monkeys. Histopathologic examination of the fovea may be required for preclinical drug safety evaluation of certain ocular drugs for human risk assessment. Using the inferior oblique muscle as a landmark, the macula including the fovea was obtained in 100% of 20 eye sections. It is easier to obtain just the macular region of the retina than the fovea in the section. Measurements for the inferior oblique muscle (point of insertion), fovea, macula, and the optic disc for 20 eyes of untreated cynomolgus monkeys were taken. In the cynomolgus monkey, the macula was 2.01 mm in diameter and the fovea was 0.53 mm in diameter. The distance from the center of the optic disc to the center of the fovea was 3.23 mm. In the fixed eyes, the point of insertion of the inferior oblique muscle to the sclera was 5.96 mm. (The J Histotechnol 31:113, 2008)

Submitted January 7, 2008; accepted with revisions June 22, 2008     

Neuronal density in the human retinal ganglion cell layer from 16–77 years

Literature assessing whether or not neurons (retinal ganglion cells and displaced amacrine cells) are lost from the retinal ganglion cell layer in mammals with age is still controversial, some studies finding a decrease in cell density and others not. To date there have been no studies estimating the total number of neurons in the retinal ganglion cell layer of humans throughout life. Recent studies have concentrated on the macular region and examined cell densities, which are reported to decrease during aging. In a study of the human retinal pigment epithelium (RPE), we showed that, while RPE cell number does not change, cell density increases significantly in central temporal retina (macular region) as the retina ages. We speculated that the increase in density represents a “drawing together” of the retinal sheet to maintain high cell densities, in this region of the neural retina, in the face of presumed cell loss from the ganglion cell layer due to aging. Here, therefore, we have sampled the entire ganglion cell layer of the human retina and estimated total neuron numbers in 12 retinae aged from 16 to 77 years. Human retinae, fixed in formalin, were obtained from the Queensland Eye Bank and whole‐mounted, ganglion cell layer uppermost. The total number of neurons was lower in the older than younger retinae and neuronal density was lower in most retinal regions in older retinae. Retinal area increased with age and neuronal density fell throughout the retina with a mean reduction of 0.53% per year. However, the percentage reduction in density was much lower for the macular region, with a value of 0.29% per year. It is possible that this lesser reduction in cell density in the macula is a result of the drawing together of the retinal sheet in this region as we speculated from RPE data. Anat Rec 260:124–131, 2000. © 2000 Wiley‐Liss, Inc.