Zhang Wanjun, Li Huiqi
School of Information and Electronics, Beijing Institute of Technology, Beijing, 100081, China.
Med Biol Eng Comput. 2017 May;55(5):769-779. doi: 10.1007/s11517-016-1554-1. Epub 2016 Aug 4.
Cataract leads to visual impairment. Among different types of cataract, posterior subcapsular cataract (PSC) can develop rapidly and surgery is usually needed. An approach to detect PSC opacities in retro-illumination images is proposed. Watershed and Markov random fields (MRF) method are employed to opacities in anterior retro-illumination images. It results in a mixture of PSC, cortical opacities and noise. Then, information in both anterior and posterior retro-illumination images is utilized. Two features are extracted to identify PSC: mean gradient comparison (MGC) between anterior and posterior retro-illumination images, and spatial location. This is the first time that comparison between anterior and posterior retro-illumination images is proposed and MGC is proposed as the feature of comparison in PSC detection. Experiments show that the sensitivity and specificity of PSC screening is 91.2 and 90.1 %, respectively, based on the 519 pairs of testing images. To the best of our knowledge, it is the best performance reported in automatic detection of PSC. Compared with the methods in the literatures, considerable improvement is achieved when there are large areas of PSC opacities.
白内障会导致视力受损。在不同类型的白内障中，后囊下白内障（PSC）发展迅速，通常需要进行手术。本文提出了一种在逆光图像中检测PSC混浊的方法。分水岭和马尔可夫随机场（MRF）方法被用于处理前向逆光图像中的混浊。这会导致PSC、皮质混浊和噪声的混合。然后，利用前向和后向逆光图像中的信息。提取了两个特征来识别PSC：前向和后向逆光图像之间的平均梯度比较（MGC）以及空间位置。这是首次提出前向和后向逆光图像之间的比较，并将MGC作为PSC检测中的比较特征。实验表明，基于519对测试图像，PSC筛查的灵敏度和特异性分别为91.2%和90.1%。据我们所知，这是PSC自动检测中报告的最佳性能。与文献中的方法相比，当存在大面积的PSC混浊时，有了显著的改进。
