Depth Maps From Stereo Images of the Optic Disc-Phase II

据估计，美国有300万人患有青光眼，2005年全球将有6700万人患有青光眼。至少有一半的人不知道自己患有青光眼，因为青光眼在早期通常没有症状。目前的研究旨在开发早期诊断方法，以便在早期阶段发现疾病，此时治疗最有效，最大限度地减少不可逆转的视力损失。视神经头(ONH)早期病理改变的识别对青光眼的诊断至关重要。目前测量这些变化的金标准是通过临床立体观察ONH。来自美国眼科学会(AAO)2000年的研究发现，由有经验的临床观察者对立体ONH照片进行定性评估的诊断精度优于任何其他目前可用的视盘评估方法(识别早期青光眼变化，Wollstein，Health，2000 AAOO)。立体评价的主要缺点是主观性，它造成了读者之间和读者内部观察者之间的可变性。需要一种新的方法来评估立体照片，以一致地分析这些变化 在ONH上。这可以通过使用定量分析ONH的变化来实现。第一阶段项目演示了一种基于一对立体图像计算ONH三维拓扑的技术。手动注解的杯和盘与自动生成的等值差之间的高度相关性(&gt；0.9) 轮廓被记录在案。第二阶段的主要目标是通过与多个读取器和 使用现有的临床仪器，如海德堡视网膜断层扫描仪(HRT)和GDX神经纤维分析仪。视差算法的准确性将基于模型视网膜的模型进行评估，该模型已被加工成具有精确的视盘状特征。这种模体在放射学中被用来校准断层成像设备。第二阶段的第二个目标是为算法的使用开发一个接口，并在临床环境中进行测试，以证明该软件是商业级产品。用于测试和验证的数据将包括传统的立体眼底照相、数字刺激照相、GDX神经纤维分析仪、HRT、Goldman和Humphrey视野以及眼压。

It is estimated that 3 million Americans have glaucoma,and 67 million peopleworldwidewill have glaucomaby2005. At least half do not know they have it because glaucoma usually has no symptoms in the early stages. Current research is aimed at developing methods of early diagnosis to detect the disease in the early stages, when treatment is most effective in minimizing irreversible vision loss. The identification of early pathologic changes at the Optic Nerve Head (ONH) is crucial to the diagnosis of glaucoma. The current gold standard for measuring these changes is through clinical stereoscopic observation of the ONH. From the American Academy of Ophthalmology (AAO) 2000, "Previous studies have found that the diagnostic precision of qualitative evaluation of stereoscopic ONHphotographs by experienced clinical observers is superior to any other currently available method of optic disc assessment" (Identifying Early Glaucomatous Changes, Wollstein, Health, 2000 AAOO). The major drawback to the stereoscopic evaluation is the subjectivity that creates inter- and intra-reader observer variability. A new method for evaluating the stereoscopic photographs is needed for consistent analysis of the changes on the ONH. This can be brought about by the use of quantitative analysis of the ONH changes. The Phase I project demonstrated a technique for calculating 3-dimensional topology of the ONH based on a pair of stereoscopic images. High correlations (> 0.9) between the manually annotated cup and disc and the automatically generated iso-disparity contours was documented. The principal aim of Phase II is to validate the technique by performing correlations with multiple readers and with existing clinical instruments such as the Heidelberg Retina Tomographer (HRT), and GDx Nerve Fiber Analyzer. The disparity algorithm's accuracy will be evaluated based on a phantom in the form of a model retina which has been machined with precise optic disc-like features. Such phantoms are used in radiology to calibrate tomography equipment. A second aim of the Phase II is to develop an interface for the use of the algorithm and to be tested in a clinical environment to demonstrate the software as a commercial grade product. Data for testing and validation will consist of traditional stereo fundus photography, digital stimultaneous photography, GDx nerve fiber analyzer, HRT,Goldman and Humphrey visual fields, and Inter Occular Pressure.