基于视网膜神经纤维层测量的高度近视合并青光眼早期识别的模式研究

High myopia is an independent risk factor for glaucoma. Due to the enlongation and anatomical changes of the globle, traditional diagnositic appoaches, like optic disc morphology, visual field, and HRT, are no longer able to recognize glaucoma effectively. According to our previous research, RNFL were probably the most stable parameter for glaucoma detection in high myopia, whose distributional deviations were basically spacial but not structural. The variation of RNFL was associated with myopic refraction, the optic disc-macular position, and the spacial location of the optic disc. We suppose that the deviation of RNFL has some certain rules and could be corrected, standardized, and then be comparable. Besides our group found that the assymmetry between upper and lower RNFL exist in eyes including emmetropia, myopia and even high myopia, however was broken in early glaucoma. We aim to investigate the characteristics and related physiological and anatomical factors of RNFL distribution of glaucomatous and non-glaucomatous eyes in varied refractive gradients in a large sample size. With the aid of mathematic methods, a recognition mode of glaucoma detection in high myopia will be achieved after procedures like traits extraction and assymmetry analysis based on the corrected and standardized RNFL curves. The database of RNFL in varied refraction status will also be established. The problem of early diagnosis of glaucoma in high myopia patients is supposed to be partially solved.

高度近视是青光眼的独立危险因素。由于眼轴拉长、眼内组织结构发生变化，视盘形态、视野、HRT等传统手段无法对合并青光眼进行早期识别，贻误治疗。前期研究发现在高度近视中，RNFL可能是相对最稳定的青光眼参数，随眼球扩张可能仅发生空间性改变，其变异与近视屈光度、视盘黄斑相对位置及视盘空间结构密切相关，有规律可循，在控制相关参数后有可能对RNFL曲线个体变异进行校正而形成标准化曲线；此外还发现近视及高度近视眼同正视眼一样，都存在RNFL分布的规律性及上下半侧对称性，而此对称性在青光眼中被打破。本项目将通过对大样本青光眼及非青光眼的RNFL曲线及相关生理、解剖参数进行研究，以此对曲线进行校正及标准化处理，利用模式识别技术对标化RNFL曲线进行特征向量提取、上下方相似性分析等，构建适于高度近视的青光眼判别模型。此外建立不同屈光度的RNFL数据库，综合解决高度近视合并青光眼诊断的难题。

高度近视是青光眼的独立危险因素。由于高度近视眼轴拉长、眼内组织结构发生变化，视盘形态视野、HRT等传统手段无法对合并青光眼进行早期识别，贻误治疗，高度近视合并青光眼诊断成为困扰临床医生的难点问题。为了提高高度近视合并青光眼的自动诊断能力，本研究基于大样本视盘旁视网膜纤维层（RNFL）数据，通过对RNFL曲线进行分解，研究高度近视RNFL分布是否存在规律性，其变异是否能够被校正；研究高度近视合并青光眼是否具有特征性，并能够进行识别；基于神经网络方法，研究并建立基于RNFL厚度测量的高度近视近视合并青光眼的早期自动诊断模式。利用已知全身及眼部参数，对近视及高度近视的RNFL曲线进行补偿（校正），补偿包含位置补偿和厚度补偿。经过补偿后，RNFL进行高度近视合并青光眼诊断效力明显提升，对眼轴最长前10%眼（平均眼轴26.0±0.9mm）、对眼轴最长前20%眼（平均眼轴25.3±1.0mm）、对眼轴最长前30%眼（平均眼轴24.9±1.0mm）、及所有眼（平均眼轴23.5±1.2mm）的青光眼诊断准确率分别提高19.8%、18.9%、16.2% 及11.3% 。该补偿方法大大提升了青光眼自动诊断能力，尤其对于高度近视合并青光眼，提升高达20%，该方法为首创，显著优于目前已知的高度近视合并青光眼诊断方法，具有重要社会价值及应用潜力。