Non-mydriatic panoramic fundus imaging using a non-contact scanning laser-based system.
Non-mydriatic panoramic fundus imaging using a non-contact scanning laser-based system.
复制标题
使用非接触式扫描激光系统进行免散瞳全景眼底成像。
DOI：
10.3928/1542-8877-20031101-15
复制
发表时间：
2003
期刊：
影响因子：
--
通讯作者：
A. Eller
中科院分区：
文献类型：
--
作者：
T. Friberg;Angana Pandya;A. Eller
背景和目标
一种基于非接触式扫描激光的视网膜成像系统通过使用大椭圆镜的新颖用途来产生200度宽视场的眼底全景图像。作者调查了是否仅凭这些图像就能做出准确的诊断。
患者和方法
图像从随机选择的患者的双眼获得，并由视网膜专家进行回顾性审查。每位医生做出诊断，建议随访间隔，并仅根据图像检查确定是否需要任何干预。糖尿病视网膜病变分为单纯背景性视网膜病变、增殖性前期视网膜病变、黄斑病变和增殖性视网膜病变。然后将影像复查结果与临床检查结果进行比较，计算复查的敏感性和特异性。
结果
两位专家的总体诊断正确率分别为87%和77%；特异度和敏感度分别平均为76%。在86%的病例中，影像回顾的随访间隔建议与临床检查的建议一致。糖尿病视网膜病变被正确识别为一般诊断，敏感性为94%，随访建议与临床建议相符的比例为82%。62%的图片被评为质量良好或更好，只有6.4%的图片被认为不可读。虹膜颜色不影响图像质量。
结论
基于激光的非接触式扫描视网膜成像系统提供了足够分辨率和清晰度的图像，以在很高比例的病例中做出准确的一般诊断。它作为特定疾病的筛查工具是可行的，值得进一步研究。
BACKGROUND AND OBJECTIVE
A non-contact scanning laser-based retinal imaging system produces a 200 degrees wide-field panoramic image of the fundus through the novel use of a large elliptical mirror. The authors investigated whether accurate diagnoses could be made from these images alone.
PATIENTS AND METHODS
Images were obtained from both eyes of arbitrarily selected patients and were retrospectively reviewed by retinal specialists. Each physician made a diagnosis, suggested the follow-up interval, and determined whether any intervention was needed based solely on image review. Diabetic retinopathy was categorized as background retinopathy only, preproliferative retinopathy, maculopathy, and proliferative retinopathy. The results of the image review were then compared with the results of clinical examination, and the sensitivity and specificity of the review were calculated.
RESULTS
The correct overall diagnosis was made by the two specialists in 87% and 77% of cases, respectively; the specificity and sensitivity each averaged 76%. Follow-up interval recommendations from image review matched those made on clinical examination in 86% of cases. Diabetic retinopathy was correctly identified as a general diagnosis with a sensitivity of 94% and the follow-up recommendations corresponded to the clinical recommendations in 82%. Sixty-two percent of the images were judged to be of good quality or better, with only 6.4% deemed unreadable. Iris color did not influence image quality.
CONCLUSION
The non-contact scanning laser-based retinal imaging system provides images of sufficient resolution and clarity to make accurate general diagnoses in a high percentage of cases. Its use as a screening device for specific diseases is feasible and merits further study.