压缩三维与二维数字图像对比标准三维幻灯片胶片评估青光眼性视神经特征
这项前瞻性多阅片者验证研究对192只青光眼、可疑或正常视神经眼进行16:1压缩的三维数字(3DD)和二维数字(2DD)成像,并与立体幻灯片胶片比较。结果显示3DD和2DD在评估垂直杯盘比、切迹和盘出血方面均与胶片有极好可重复性,3DD在检测盘出血、切迹和倾斜方面略优于2DD。
要点
- 3DD与胶片评估垂直杯盘比的κ=0.781,2DD与胶片为κ=0.69。
- 3DD检测切迹的敏感度95.2%、特异度95.2%、AUC=0.953,优于2DD(p=0.03)。
- 3DD检测盘出血的敏感度77.8%、特异度98.9%、AUC=0.883,优于2DD(p=0.049)。
核心结论
- 3DD和2DD成像与立体幻灯片胶片相比,在专家评估垂直杯盘比、切迹和盘出血时均表现出极好的可重复性。
- 本研究中3DD在评估盘出血、切迹和倾斜方面比2DD略更准确。
研究设计与样本
- 前瞻性多阅片者验证研究,对青光眼、可疑或正常视神经眼进行16:1压缩的3DD和2DD(1024×1280像素)成像,并与立体幻灯片胶片比较。
- 共192只眼接受每种格式成像,由四名青光眼专家随机分级,并按需进行共识解读以协调数据。
主要发现
- 幻灯片、3DD和2DD的平均垂直杯盘比分别为0.59±0.20、0.60±0.18和0.62±0.17。
- 3DD与胶片垂直杯盘比的一致性κ=0.781,2DD与胶片为κ=0.69。
- 3DD检测切迹的敏感度95.2%、特异度95.2%、AUC=0.953,优于2DD的90.5%、88.6%、AUC=0.895(p=0.03)。
- 3DD检测盘出血的敏感度77.8%、特异度98.9%、AUC=0.883,优于2DD的44.4%、99.5%、AUC=0.72(p=0.049)。
- 3DD与2DD在检测盘倾斜(p=0.7)、视盘周围萎缩(p=0.16)、灰新月(p=0.1)或苍白(p=0.43)方面无差异,但3DD检测倾斜更好(p=0.013)。
关键数值与推荐
- 3DD检测切迹的敏感度95.2%、特异度95.2%、AUC=0.953,优于2DD(p=0.03)。
- 3DD检测盘出血的敏感度77.8%、特异度98.9%、AUC=0.883,优于2DD(p=0.049)。
- 3DD与胶片垂直杯盘比一致性κ=0.781,2DD与胶片为κ=0.69。
临床怎么用
- 临床医生可对压缩3DD和2DD成像评估青光眼性视盘损害重要特征与既往金标准立体幻灯片胶片摄影相当抱有信心,支持将数字成像用于远程青光眼应用。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
SYNOPSIS: Clinicians can feel confident compressed three-dimensional digital (3DD) and two-dimensional digital (2DD) imaging evaluating important features of glaucomatous disc damage is comparable to the previous gold standard of stereoscopic slide film photography, supporting the use of digital imaging for teleglaucoma applications. BACKGROUND/AIMS: To compare the sensitivity and specificity of 3DD and 2DD photography with stereo slide film in detecting glaucomatous optic nerve head features. METHODS: This prospective, multireader validation study imaged and compressed glaucomatous, suspicious or normal optic nerves using a ratio of 16:1 into 3DD and 2DD (1024×1280 pixels) and compared both to stereo slide film. The primary outcome was vertical cup-to-disc ratio (VCDR) and secondary outcomes, including disc haemorrhage and notching, were also evaluated. Each format was graded randomly by four glaucoma specialists. A protocol was implemented for harmonising data including consensus-based interpretation as needed. RESULTS: There were 192 eyes imaged with each format. The mean VCDR for slide, 3DD and 2DD was 0.59±0.20, 0.60±0.18 and 0.62±0.17, respectively. The agreement of VCDR for 3DD versus film was κ=0.781 and for 2DD versus film was κ=0.69. Sensitivity (95.2%), specificity (95.2%) and area under the curve (AUC; 0.953) of 3DD imaging to detect notching were better (p=0.03) than for 2DD (90.5%; 88.6%; AUC=0.895). Similarly, sensitivity (77.8%), specificity (98.9%) and AUC (0.883) of 3DD to detect disc haemorrhage were better (p=0.049) than for 2DD (44.4%; 99.5%; AUC=0.72). There was no difference between 3DD and 2DD imaging in detecting disc tilt (p=0.7), peripapillary atrophy (p=0.16), grey crescent (p=0.1) or pallor (p=0.43), although 3D detected sloping better (p=0.013). CONCLUSIONS: Both 3DD and 2DD imaging demonstrates excellent reproducibility in comparison to stereo slide film with experts evaluating VCDR, notching and disc haemorrhage. 3DD in this study was slightly more accurate than 2DD for evaluating disc haemorrhage, notching and sloping.
引用
- Sandhu S, Rudnisky C, Arora S, et al. Compressed 3D and 2D digital images versus standard 3D slide film for the evaluation of glaucomatous optic nerve features. Br J Ophthalmol. 2017. PMID: 28835424. doi:10.1136/bjophthalmol-2017-310447
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。