新加坡国家眼科中心Tien Y. Wong小组取得一项新突破。他们研究了人工智能检测视神经乳头水肿的效果。 该成果发表在2020年4月14日出版的《新英格兰医学杂志》上。

非眼科医生无法自信地直接进行检眼镜检查。利用人工智能是否能从眼底图像中检出视神经乳头水肿和其他视盘异常尚未明确。

研究组对深度学习系统进行了培训、验证和外部测试，从15846例回顾性收集的眼底照片中，将视盘分类为正常或视神经乳头水肿或其他异常。在这些照片中，来自11个国家/地区的19个站点的14341张照片用于培训和验证，来自其他5个站点的1505张照片用于外部测试。通过计算接收工作特性曲线下面积（AUC）、灵敏度和特异性来评估视盘外观分类的性能，并与神经眼科医生的临床诊断参考标准进行比较。

培训和验证数据集包括来自6779名患者的14341张照片：9156张正常视盘，2148张视神经乳头水肿和3037张其他异常视盘。各站点归类为正常视盘的百分比为9.8-100%，归类为视神经乳头水肿的百分比为0-59.5%。

在验证数据集中，系统可将视神经乳头水肿从正常视盘和非乳头水肿异常视盘中区分开来，AUC为0.99；亦可将正常视盘从异常视盘中辨别出来，AUC为0.99。在外部测试数据集的1505张照片中，系统检出视神经乳头水肿的AUC为0.96，灵敏度为96.4%，特异性为84.7%。

总之，一种深度学习系统，使用眼底照片和药理扩张的瞳孔，可准确辨别视神经乳头水肿、正常视盘和非乳头水肿异常视盘。

附：英文原文

Title: Artificial Intelligence to Detect Papilledema from Ocular Fundus Photographs | NEJM

Author: Dan Milea, M.D., Ph.D.,, Raymond P. Najjar, Ph.D.,, Jiang Zhubo, M.Sc.,, Daniel Ting, M.D., Ph.D.,, Caroline Vasseneix, M.D.,, Xinxing Xu, Ph.D.,, Masoud Aghsaei Fard, M.D.,, Pedro Fonseca, M.D.,, Kavin Vanikieti, M.D.,, Wolf A. Lagrèze, M.D.,, Chiara La Morgia, M.D., Ph.D.,, Carol Y. Cheung, Ph.D.,, Steffen Hamann, M.D., Ph.D.,, Christophe Chiquet, M.D., Ph.D.,, Nicolae Sanda, M.D., Ph.D.,, Hui Yang, M.D., Ph.D.,, Luis J. Mejico, M.D.,, Marie-Bénédicte Rougier, M.D.,, Richard Kho, M.D.,, Tran Thi Ha Chau, M.D.,, Shweta Singhal, M.B., B.S., Ph.D.,, Philippe Gohier, M.D.,, Catherine Clermont-Vignal, M.D.,, Ching-Yu Cheng, M.D., Ph.D., M.P.H.,, Jost B. Jonas, M.D.,, Patrick Yu-Wai-Man, M.B., B.S., Ph.D.,, Clare L. Fraser, M.B., B.S., M.Med.,, John J. Chen, M.D., Ph.D.,, Selvakumar Ambika, D.O., D.N.B.,, Neil R. Miller, M.D.,, Yong Liu, Ph.D.,, Nancy J. Newman, M.D.,, Tien Y. Wong, M.D., Ph.D.,, and Valérie Biousse, M.D.

Issue&Volume: 2020-04-14

Abstract: Abstract

Background

Nonophthalmologist physicians do not confidently perform direct ophthalmoscopy. The use of artificial intelligence to detect papilledema and other optic-disk abnormalities from fundus photographs has not been well studied.

Methods

We trained, validated, and externally tested a deep-learning system to classify optic disks as being normal or having papilledema or other abnormalities from 15,846 retrospectively collected ocular fundus photographs that had been obtained with pharmacologic pupillary dilation and various digital cameras in persons from multiple ethnic populations. Of these photographs, 14,341 from 19 sites in 11 countries were used for training and validation, and 1505 photographs from 5 other sites were used for external testing. Performance at classifying the optic-disk appearance was evaluated by calculating the area under the receiver-operating-characteristic curve (AUC), sensitivity, and specificity, as compared with a reference standard of clinical diagnoses by neuro-ophthalmologists.

Results

The training and validation data sets from 6779 patients included 14,341 photographs: 9156 of normal disks, 2148 of disks with papilledema, and 3037 of disks with other abnormalities. The percentage classified as being normal ranged across sites from 9.8 to 100%; the percentage classified as having papilledema ranged across sites from zero to 59.5%. In the validation set, the system discriminated disks with papilledema from normal disks and disks with nonpapilledema abnormalities with an AUC of 0.99 (95% confidence interval [CI], 0.98 to 0.99) and normal from abnormal disks with an AUC of 0.99 (95% CI, 0.99 to 0.99). In the external-testing data set of 1505 photographs, the system had an AUC for the detection of papilledema of 0.96 (95% CI, 0.95 to 0.97), a sensitivity of 96.4% (95% CI, 93.9 to 98.3), and a specificity of 84.7% (95% CI, 82.3 to 87.1).

Conclusions

A deep-learning system using fundus photographs with pharmacologically dilated pupils differentiated among optic disks with papilledema, normal disks, and disks with nonpapilledema abnormalities.

DOI: 10.1056/NEJMoa1917130

Source: https://www.nejm.org/doi/full/10.1056/NEJMoa1917130