From the Retina Service, Angiogenesis Lab, Massachusetts Eye and Ear Infirmary, Harvard Medical School, Boston, MA, USA.
From the Department of Ophthalmology and Visual Sciences, University of Illinois at Chicago, Chicago, IL, USA.
Sci Rep. 2020 Oct 28;10(1):18459. doi: 10.1038/s41598-020-75501-y.
To develop an automated retina layer thickness measurement tool for the ImageJ platform, to quantitate nuclear layers following the retina contour. We developed the ThicknessTool (TT), an automated thickness measurement plugin for the ImageJ platform. To calibrate TT, we created a calibration dataset of mock binary skeletonized mask images with increasing thickness masks and different rotations. Following, we created a training dataset and performed an agreement analysis of thickness measurements between TT and two masked manual observers. Finally, we tested the performance of TT measurements in a validation dataset of retinal detachment images. In the calibration dataset, there were no differences in layer thickness between measured and known thickness masks, with an overall coefficient of variation of 0.00%. Training dataset measurements of immunofluorescence retina nuclear layers disclosed no significant differences between TT and any observer's average outer nuclear layer (ONL) (p = 0.998), inner nuclear layer (INL) (p = 0.807), and ONL/INL ratio (p = 0.944) measurements. Agreement analysis showed that bias between TT vs. observers' mean was lower than between any observers' mean against each other in the ONL (0.77 ± 0.34 µm vs 3.25 ± 0.33 µm) and INL (1.59 ± 0.28 µm vs 2.82 ± 0.36 µm). Validation dataset showed that TT can detect significant and true ONL thinning (p = 0.006), more sensitive than manual measurement capabilities (p = 0.069). ThicknessTool can measure retina nuclear layers thickness in a fast, accurate, and precise manner with multi-platform capabilities. In addition, the TT can be customized to user preferences and is freely available to download.
为了在 ImageJ 平台上开发一种自动视网膜层厚度测量工具，以定量测量视网膜轮廓内的核层。我们开发了 ThicknessTool（TT），这是一个用于 ImageJ 平台的自动厚度测量插件。为了校准 TT，我们创建了一个带有递增厚度掩模和不同旋转的模拟二进制骨架掩模图像的校准数据集。随后，我们创建了一个训练数据集，并对 TT 和两个掩蔽手动观察者的厚度测量值进行了一致性分析。最后，我们在视网膜脱离图像的验证数据集中测试了 TT 测量的性能。在校准数据集中，测量的和已知厚度掩模之间的层厚度没有差异，总体变异系数为 0.00%。免疫荧光视网膜核层的训练数据集测量结果显示，TT 与任何观察者的平均外核层（ONL）（p = 0.998）、内核层（INL）（p = 0.807）和 ONL/INL 比值（p = 0.944）测量值之间没有显著差异。一致性分析表明，TT 与观察者平均值之间的偏差低于任何观察者平均值之间的偏差，在外核层（0.77 ± 0.34 µm 与 3.25 ± 0.33 µm）和内核层（1.59 ± 0.28 µm 与 2.82 ± 0.36 µm）中都是如此。验证数据集表明，TT 可以检测到显著且真实的 ONL 变薄（p = 0.006），比手动测量能力更敏感（p = 0.069）。ThicknessTool 可以快速、准确和精确地测量视网膜核层厚度，并具有多平台功能。此外，TT 可以根据用户偏好进行定制，并且可以免费下载。
