Peli E, Hedges Iii T R, Schwartz B
Appl Opt. 1989 Mar 15;28(6):1128-34. doi: 10.1364/AO.28.001128.
An image analysis method to measure retinal nerve fiber layer (RNFL) striations from digitized fundus photographs was developed to improve detection and monitoring of progressive diffuse RNFL loss. Striations were measured by comparing the high spatial frequency variability across with the variability along the RNFL. This locally normalized measure of striations compensates for the wide density variations both within individuals and between individuals and RNFL photographs. Five repeated measurements were taken at each of three locations from each retinal image. Measurements from five patients with recorded visual field loss due to optic nerve diseases were compared with five normal subjects and five suspect eyes. Measurements clearly distinguished the three groups when taken at the temporal arcades. Measurements above and below the arcades were also consistent, but did not distinguish normals from suspects. The measure was correlated with graded estimates of RNFL integrity of two trained observers (rho = -0.57, p < 0.001 and rho = -0.61, p < 0.001).
为了改进对进行性弥漫性视网膜神经纤维层(RNFL)丢失的检测和监测,开发了一种从数字化眼底照片测量RNFL条纹的图像分析方法。通过比较RNFL上的高空间频率变异性与沿RNFL的变异性来测量条纹。这种局部归一化的条纹测量方法补偿了个体内部以及个体之间和RNFL照片之间广泛的密度变化。从每个视网膜图像的三个位置中的每个位置进行五次重复测量。将五名因视神经疾病导致视野缺损的患者的测量结果与五名正常受试者和五只可疑眼睛进行比较。当在颞侧弓形区进行测量时,测量结果能清楚地区分这三组。弓形区上方和下方的测量结果也一致,但无法区分正常人和可疑者。该测量方法与两名训练有素的观察者对RNFL完整性的分级估计相关(rho = -0.57,p < 0.001和rho = -0.61,p < 0.001)。