Diagnostic performance of deep learning for infectious keratitis: a systematic review and meta-analysis.

IF 9.6 1区医学 Q1 MEDICINE, GENERAL & INTERNAL EClinicalMedicine Pub Date : 2024-10-18 eCollection Date: 2024-11-01 DOI:10.1016/j.eclinm.2024.102887

Zun Zheng Ong, Youssef Sadek, Riaz Qureshi, Su-Hsun Liu, Tianjing Li, Xiaoxuan Liu, Yemisi Takwoingi, Viknesh Sounderajah, Hutan Ashrafian, Daniel S W Ting, Jodhbir S Mehta, Saaeha Rauz, Dalia G Said, Harminder S Dua, Matthew J Burton, Darren S J Ting

{"title":"Diagnostic performance of deep learning for infectious keratitis: a systematic review and meta-analysis.","authors":"Zun Zheng Ong, Youssef Sadek, Riaz Qureshi, Su-Hsun Liu, Tianjing Li, Xiaoxuan Liu, Yemisi Takwoingi, Viknesh Sounderajah, Hutan Ashrafian, Daniel S W Ting, Jodhbir S Mehta, Saaeha Rauz, Dalia G Said, Harminder S Dua, Matthew J Burton, Darren S J Ting","doi":"10.1016/j.eclinm.2024.102887","DOIUrl":null,"url":null,"abstract":"Background: Infectious keratitis (IK) is the leading cause of corneal blindness globally. Deep learning (DL) is an emerging tool for medical diagnosis, though its value in IK is unclear. We aimed to assess the diagnostic accuracy of DL for IK and its comparative accuracy with ophthalmologists.Methods: In this systematic review and meta-analysis, we searched EMBASE, MEDLINE, and clinical registries for studies related to DL for IK published between 1974 and July 16, 2024. We performed meta-analyses using bivariate models to estimate summary sensitivities and specificities. This systematic review was registered with PROSPERO (CRD42022348596).Findings: Of 963 studies identified, 35 studies (136,401 corneal images from >56,011 patients) were included. Most studies had low risk of bias (68.6%) and low applicability concern (91.4%) in all domains of QUADAS-2, except the index test domain. Against the reference standard of expert consensus and/or microbiological results (seven external validation studies; 10,675 images), the summary estimates (95% CI) for sensitivity and specificity of DL for IK were 86.2% (71.6-93.9) and 96.3% (91.5-98.5). From 28 internal validation studies (16,059 images), summary estimates for sensitivity and specificity were 91.6% (86.8-94.8) and 90.7% (84.8-94.5). Based on seven studies (4007 images), DL and ophthalmologists had comparable summary sensitivity [89.2% (82.2-93.6) versus 82.2% (71.5-89.5); P = 0.20] and specificity [(93.2% (85.5-97.0) versus 89.6% (78.8-95.2); P = 0.45].Interpretation: DL models may have good diagnostic accuracy for IK and comparable performance to ophthalmologists. These findings should be interpreted with caution due to the image-based analysis that did not account for potential correlation within individuals, relatively homogeneous population studies, lack of pre-specification of DL thresholds, and limited external validation. Future studies should improve their reporting, data diversity, external validation, transparency, and explainability to increase the reliability and generalisability of DL models for clinical deployment.Funding: NIH, Wellcome Trust, MRC, Fight for Sight, BHP, and ESCRS.","PeriodicalId":11393,"journal":{"name":"EClinicalMedicine","volume":null,"pages":null},"PeriodicalIF":9.6000,"publicationDate":"2024-10-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11513659/pdf/","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"EClinicalMedicine","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1016/j.eclinm.2024.102887","RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2024/11/1 0:00:00","PubModel":"eCollection","JCR":"Q1","JCRName":"MEDICINE, GENERAL & INTERNAL","Score":null,"Total":0}

引用次数: 0

Abstract

Background: Infectious keratitis (IK) is the leading cause of corneal blindness globally. Deep learning (DL) is an emerging tool for medical diagnosis, though its value in IK is unclear. We aimed to assess the diagnostic accuracy of DL for IK and its comparative accuracy with ophthalmologists.

Methods: In this systematic review and meta-analysis, we searched EMBASE, MEDLINE, and clinical registries for studies related to DL for IK published between 1974 and July 16, 2024. We performed meta-analyses using bivariate models to estimate summary sensitivities and specificities. This systematic review was registered with PROSPERO (CRD42022348596).

Findings: Of 963 studies identified, 35 studies (136,401 corneal images from >56,011 patients) were included. Most studies had low risk of bias (68.6%) and low applicability concern (91.4%) in all domains of QUADAS-2, except the index test domain. Against the reference standard of expert consensus and/or microbiological results (seven external validation studies; 10,675 images), the summary estimates (95% CI) for sensitivity and specificity of DL for IK were 86.2% (71.6-93.9) and 96.3% (91.5-98.5). From 28 internal validation studies (16,059 images), summary estimates for sensitivity and specificity were 91.6% (86.8-94.8) and 90.7% (84.8-94.5). Based on seven studies (4007 images), DL and ophthalmologists had comparable summary sensitivity [89.2% (82.2-93.6) versus 82.2% (71.5-89.5); P = 0.20] and specificity [(93.2% (85.5-97.0) versus 89.6% (78.8-95.2); P = 0.45].

Interpretation: DL models may have good diagnostic accuracy for IK and comparable performance to ophthalmologists. These findings should be interpreted with caution due to the image-based analysis that did not account for potential correlation within individuals, relatively homogeneous population studies, lack of pre-specification of DL thresholds, and limited external validation. Future studies should improve their reporting, data diversity, external validation, transparency, and explainability to increase the reliability and generalisability of DL models for clinical deployment.

Funding: NIH, Wellcome Trust, MRC, Fight for Sight, BHP, and ESCRS.

查看原文

微信好友朋友圈 QQ好友复制链接

本刊更多论文

深度学习对传染性角膜炎的诊断性能：系统回顾和荟萃分析。

背景：感染性角膜炎（IK）是全球角膜失明的主要原因。深度学习（DL）是一种新兴的医学诊断工具，但其在IK中的价值尚不明确。我们旨在评估深度学习对 IK 的诊断准确性以及与眼科医生的比较准确性：在这项系统性综述和荟萃分析中，我们检索了 EMBASE、MEDLINE 和临床登记处在 1974 年至 2024 年 7 月 16 日期间发表的与 DL 诊断 IK 相关的研究。我们使用双变量模型进行了荟萃分析，以估算灵敏度和特异度。本系统综述已在 PROSPERO 注册（CRD42022348596）：在确定的 963 项研究中，有 35 项研究（来自超过 56 011 名患者的 136 401 张角膜图像）被纳入其中。大多数研究的偏倚风险较低（68.6%），QUADAS-2 所有领域的适用性关注度较低（91.4%），指数测试领域除外。对照专家共识和/或微生物学结果的参考标准（7 项外部验证研究；10,675 幅图像），DL 对 IK 的敏感性和特异性的汇总估计值（95% CI）分别为 86.2% (71.6-93.9) 和 96.3% (91.5-98.5)。在 28 项内部验证研究（16059 张图像）中，灵敏度和特异性的汇总估计值分别为 91.6% (86.8-94.8) 和 90.7% (84.8-94.5)。根据七项研究（4007 张图像），DL 和眼科医生的灵敏度[89.2%（82.2-93.6）对 82.2%（71.5-89.5）；P = 0.20]和特异性[93.2%（85.5-97.0）对 89.6%（78.8-95.2）；P = 0.45]具有可比性：DL模型对IK可能具有良好的诊断准确性，其性能与眼科医生相当。由于基于图像的分析没有考虑个体内部的潜在相关性、研究人群相对单一、缺乏对 DL 阈值的预先指定以及外部验证有限，因此在解释这些研究结果时应谨慎。未来的研究应改进其报告、数据多样性、外部验证、透明度和可解释性，以提高DL模型在临床应用中的可靠性和通用性：美国国立卫生研究院（NIH）、惠康基金会（Wellcome Trust）、英国医学研究中心（MRC）、为视力而战组织（Fight for Sight）、必和必拓公司（BHP）和ESCRS。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文去求助

来源期刊

EClinicalMedicine Medicine-Medicine (all)

CiteScore

18.90

自引率

1.30%

发文量

506

审稿时长

22 days

期刊介绍： eClinicalMedicine is a gold open-access clinical journal designed to support frontline health professionals in addressing the complex and rapid health transitions affecting societies globally. The journal aims to assist practitioners in overcoming healthcare challenges across diverse communities, spanning diagnosis, treatment, prevention, and health promotion. Integrating disciplines from various specialties and life stages, it seeks to enhance health systems as fundamental institutions within societies. With a forward-thinking approach, eClinicalMedicine aims to redefine the future of healthcare.