Tissue Engineering and Stem Cell Group, Singapore Eye Research Institute, Singapore; Singapore National Eye Centre, Singapore; Ophthalmology Academic Clinical Program, Duke-NUS Graduate Medical School, Singapore.
Tissue Engineering and Stem Cell Group, Singapore Eye Research Institute, Singapore; Ophthalmology Academic Clinical Program, Duke-NUS Graduate Medical School, Singapore.
Am J Ophthalmol. 2021 Jan;221:260-272. doi: 10.1016/j.ajo.2020.07.029. Epub 2020 Jul 28.
To describe the validation and implementation of an automated system for the detection and quantification of guttae in Fuchs endothelial corneal dystrophy (FECD).
Observational reliability study.
Patients with FECD underwent retroillumination corneal photography, followed by determination of the distributions and sizes of corneal guttae by an automated image analysis algorithm. Performance of the automated system was assessed via (1) validation against manual guttae segmentation, (2) reproducibility studies to ensure consistency, and (3) evaluation for agreement with the Krachmer scale. It was then deployed to perform large-scale guttae assessment with anatomic subregion analysis in a batch of 40 eyes.
Compared to manual segmentation, the automated system was reasonably accurate in identifying the correct number of guttae (mean count of 78 guttae per 1 × 1 mm test frame, overestimation: +10 per frame), but had a tendency to significantly overestimate guttae size (mean guttae size 1073 μm, overestimation: +255 μm). Automated measurements of guttae counts and sizes were reproducible within a 1% discrepancy range across repeat intra-eye assessments. Automated guttae counts, interguttae distances, and density of interguttae gaps lesser than 40 μm (ie, D40 density) were highly correlated with the Krachmer scale (P < .001 for all). Large-scale guttae assessment demonstrated the automated system's potential to selectively identify a region of the corneal endothelium most affected by densely packed guttae.
Automated guttae assessment facilitates the precise identification and quantification of guttae characteristics in FECD patients. This can be used clinically as a personalized descemetorrhexis zone for Descemet stripping only and/or Descemet membrane transplantation.
描述用于检测和量化 Fuchs 内皮角膜营养不良（FECD）中角膜小滴的自动系统的验证和实施。
观察性可靠性研究。
FECD 患者接受后照角膜摄影，然后通过自动图像分析算法确定角膜小滴的分布和大小。通过（1）与手动小滴分割的验证，（2）确保一致性的重现性研究，以及（3）与 Krachmer 量表的评估来评估自动系统的性能。然后将其部署用于在 40 只眼睛的一批中进行大规模的小滴评估和解剖亚区分析。
与手动分割相比，自动系统在识别正确的小滴数量方面具有相当的准确性（每个 1 × 1mm 测试框的平均小滴数为 78 个，高估：每个框+10 个），但存在明显高估小滴大小的趋势（平均小滴大小 1073μm，高估：+255μm）。自动测量的小滴计数和大小在重复眼内评估的 1%差异范围内具有可重复性。自动小滴计数、小滴之间的距离和小于 40μm 的小滴间间隙密度（即 D40 密度）与 Krachmer 量表高度相关（所有 P<0.001）。大规模的小滴评估表明，自动系统有可能选择性地识别角膜内皮受密集小滴影响最大的区域。
自动小滴评估有助于在 FECD 患者中精确识别和量化小滴特征。这可在临床上用作个性化的 Descemet 撕裂区，仅用于 Descemet 剥离术和/或 Descemet 膜移植。
