To assess whether smaller increment and regionalised subjective grading improves the repeatability of corneal fluorescein staining assessment, and to determine the neurological approach adopted for subjective grading by practitioners. Experienced eye-care practitioners (n = 28, aged 45 ± 12 years), graded 20 full corneal staining images of patients with mild to severe Sjögren's syndrome with the Oxford grading scheme (both in 0.5 and 1.0 increments, globally and in 5 regions), expanded National Eye Institute (NEI) and SICCA Ocular Staining Score (OSS) grading scales in randomised order. This was repeated after 7–10 days. The digital images were also analysed objectively to determine staining dots, area, intensity and location (using ImageJ) for comparison. The Oxford grading scheme was similar with whole and half unit grading (2.77vs2.81,p = 0.145), but the variability was reduced (0.14vs0.12,p < 0.001). Regional grade was lower (p < 0.001) and more variable (p < 0.001) than global image grading (1.86 ± 0.44 for whole increment grading and 1.90 ± 0.39 for half unit increments). The correlation with global grading was high for both whole (r = 0.928,p < 0.001) and half increment (r = 0.934,p < 0.001) grading. Average grading across participants was associated with particle number and vertical position, with 74.4–80.4% of the linear variance accounted for by the digital image analysis. Using half unit increments with the Oxford grading scheme improve its sensitivity and repeatability in recording corneal staining. Regional grading doesn't give a comparable score and increased variability. The key neurally extracted features in assigning a subjective staining grade by clinicians were identified as the number of discrete staining locations (particles) and how close to the vertical centre was their spread, across all three scales.

中文翻译：

---

优化干燥综合征干眼病角膜染色的主观分级

评估较小的增量和区域化主观分级是否可以提高角膜荧光素染色评估的可重复性，并确定从业者采用的神经学方法进行主观分级。经验丰富的眼保健从业者（n = 28，年龄 45 ± 12 岁），根据牛津分级方案对 20 张轻度至重度干燥综合征患者的全角膜染色图像进行分级（增量为 0.5 和 1.0，全球和 5 个地区） ，按随机顺序扩展了国家眼科研究所 (NEI) 和 SICCA 眼部染色评分 (OSS) 分级量表。 7-10 天后重复此操作。还对数字图像进行客观分析，以确定染色点、面积、强度和位置（使用 ImageJ）进行比较。牛津评分方案与整体和半单元评分相似（2.77vs2.81，p = 0.145），但变异性降低（0.14vs0.12，p < 0.001）。与全局图像分级相比，区域分级较低 (p < 0.001)，且变化较大 (p < 0.001)（整个增量分级为 1.86 ± 0.44，半单位增量为 1.90 ± 0.39）。整体评分（r = 0.928，p < 0.001）和半增量评分（r = 0.934，p < 0.001）与全局评分的相关性很高。参与者的平均评分与颗粒数量和垂直位置相关，数字图像分析占线性方差的 74.4-80.4%。在牛津分级方案中使用半单位增量可以提高记录角膜染色的灵敏度和可重复性。区域评分无法提供可比较的分数并且增加了变异性。临床医生分配主观染色等级时的关键神经提取特征被确定为离散染色位置（颗粒）的数量以及它们在所有三个尺度上的分布距离垂直中心的距离。