Skalet Alison H, Quinn Graham E, Ying Gui-Shuang, Gordillo Luz, Dodobara Luz, Cocker Ken, Fielder Alistair R, Ells Anna L, Mills Monte D, Wilson Clare, Gilbert Clare
Division of Pediatric Ophthalmology, The Children's Hospital of Philadelphia and Scheie Eye Institute, University of Pennsylvania Health System, Philadelphia, Pennsylvania 19104, USA.
J AAPOS. 2008 Jun;12(3):252-8. doi: 10.1016/j.jaapos.2007.11.009. Epub 2008 Mar 4.
To determine the feasibility in a middle-level human development country of onsite training, image collection, Internet transfer, and remote grading of digital retinal images from babies screened for retinopathy of prematurity (ROP).
Two experienced nurses in a neonatal nursery in Lima, Peru, were trained to take posterior pole (30 degrees ) digital retinal images. Nurses obtained posterior pole retinal images from babies undergoing routine ROP screening and selected images for uploading via Internet for remote evaluation by five masked ROP experts. Results of gradings were compared with same-day clinical diagnostic examinations by an experienced ophthalmologist. Success rates for image acquisition and transfer for grading by expert readers were calculated.
Serial image sets from 26 of the 28 babies enrolled in this study were obtained; two babies were too unstable for imaging. Fifty-six of 58 (96.6%) imaging sessions were successful in obtaining retinal images. Three hundred thirty of 336 (98.2%) images obtained were successfully uploaded to an interactive database. Remote graders judged 93.6% to 97.3% of image sets suitable for ROP grading. Preliminary results indicate sensitivities for detection of serious ROP from 45.5% to 95.2% among individual readers, with specificities of 61.7% to 96.2% when images were gradable.
A telemedicine approach for ROP screening using digital retinal images obtained by nonophthalmologists is feasible in rapidly developing countries that lack ROP-trained ophthalmologists. If remote grading of digital images is validated as an effective method for identifying referral-warranted ROP (RW-ROP), images obtained by nonphysicians may provide a means of identifying babies who require a diagnostic examination by an ophthalmologist.
确定在一个中等人类发展水平的国家，对接受早产儿视网膜病变（ROP）筛查的婴儿进行数字视网膜图像的现场培训、图像采集、互联网传输及远程分级的可行性。
秘鲁利马一家新生儿重症监护室的两名经验丰富的护士接受了拍摄后极部（30度）数字视网膜图像的培训。护士们从接受常规ROP筛查的婴儿中获取后极部视网膜图像，并选择图像通过互联网上传，由五名蒙面的ROP专家进行远程评估。将分级结果与一位经验丰富的眼科医生当天的临床诊断检查结果进行比较。计算图像采集和专家读者分级传输的成功率。
本研究纳入的28名婴儿中有26名获得了系列图像集；两名婴儿因过于不稳定而无法成像。58次成像过程中有56次（96.6%）成功获得视网膜图像。所获得的336幅图像中有330幅（98.2%）成功上传至交互式数据库。远程分级人员判断93.6%至97.3%的图像集适合进行ROP分级。初步结果表明，个体读者检测严重ROP的敏感度为45.5%至95.2%，图像可分级时的特异度为61.7%至96.2%。
在缺乏经过ROP培训的眼科医生的快速发展国家，采用非眼科医生获取的数字视网膜图像进行ROP筛查的远程医疗方法是可行的。如果数字图像的远程分级被验证为识别需要转诊的ROP（RW-ROP）的有效方法，非医生获取的图像可能提供一种识别需要眼科医生进行诊断检查的婴儿的手段。
