Euna Lee, David Hunt, Yavuz Cakir, David Kuo, Ziqi Zhou, Miroslav Pajic, Majda Hadziahmetovic
{"title":"人工智能在老年性黄斑变性中的应用:促进诊断、预后和治疗。","authors":"Euna Lee, David Hunt, Yavuz Cakir, David Kuo, Ziqi Zhou, Miroslav Pajic, Majda Hadziahmetovic","doi":"10.1016/j.survophthal.2025.09.007","DOIUrl":null,"url":null,"abstract":"<p><p>Age-related macular degeneration (AMD) is a leading cause of irreversible vision loss in older adults. While anti-vascular endothelial growth factor (anti-VEGF) therapy and novel treatments for geographic atrophy have improved management, timely diagnosis and personalized intervention remain a challenge. Artificial intelligence (AI), such as machine learning and deep learning models, shows promise in AMD diagnosis, classification, and treatment planning. This review summarizes AI's recent advancements, highlights its clinical utility, and addresses key limitations for wider real-world implementation in AMD. We conducted systematic search of PubMed from its conception up to August 1, 2024. Studies utilizing AI-based algorithms for AMD management were identified and categorized into early detection/classification and prediction of disease progression/treatment response. Data extraction focused on AI model performance, imaging modalities, and clinical applicability. Of 193 records screened, 47 studies were included, in which 19 studies focused on early detection/classification and 28 on prediction of disease progression/treatment response. AI models demonstrated high accuracy in AMD classification and progression prediction, including in real-world settings. Prediction models for treatment response, particularly anti-VEGF therapy, could provide recommendations on optimizing injection timelines. Recent studies have also begun tackling previous challenges, such as algorithmic biases, limited generalizability, and AI's \"black-box\" nature. AI-based models offer significant potential to transform AMD care through timely detection and personalized treatment; however, clinical integration depends on improving model interpretability and validating tools across diverse populations. As AI continues to evolve, ongoing research is needed to refine AI models and support their translation into evidence-based, real-world applicability in AMD.</p>","PeriodicalId":22102,"journal":{"name":"Survey of ophthalmology","volume":" ","pages":""},"PeriodicalIF":5.9000,"publicationDate":"2025-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Artificial intelligence in age-related macular degeneration: Advancing diagnosis, prognosis, and treatment.\",\"authors\":\"Euna Lee, David Hunt, Yavuz Cakir, David Kuo, Ziqi Zhou, Miroslav Pajic, Majda Hadziahmetovic\",\"doi\":\"10.1016/j.survophthal.2025.09.007\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><p>Age-related macular degeneration (AMD) is a leading cause of irreversible vision loss in older adults. While anti-vascular endothelial growth factor (anti-VEGF) therapy and novel treatments for geographic atrophy have improved management, timely diagnosis and personalized intervention remain a challenge. Artificial intelligence (AI), such as machine learning and deep learning models, shows promise in AMD diagnosis, classification, and treatment planning. This review summarizes AI's recent advancements, highlights its clinical utility, and addresses key limitations for wider real-world implementation in AMD. We conducted systematic search of PubMed from its conception up to August 1, 2024. Studies utilizing AI-based algorithms for AMD management were identified and categorized into early detection/classification and prediction of disease progression/treatment response. Data extraction focused on AI model performance, imaging modalities, and clinical applicability. Of 193 records screened, 47 studies were included, in which 19 studies focused on early detection/classification and 28 on prediction of disease progression/treatment response. AI models demonstrated high accuracy in AMD classification and progression prediction, including in real-world settings. Prediction models for treatment response, particularly anti-VEGF therapy, could provide recommendations on optimizing injection timelines. Recent studies have also begun tackling previous challenges, such as algorithmic biases, limited generalizability, and AI's \\\"black-box\\\" nature. AI-based models offer significant potential to transform AMD care through timely detection and personalized treatment; however, clinical integration depends on improving model interpretability and validating tools across diverse populations. As AI continues to evolve, ongoing research is needed to refine AI models and support their translation into evidence-based, real-world applicability in AMD.</p>\",\"PeriodicalId\":22102,\"journal\":{\"name\":\"Survey of ophthalmology\",\"volume\":\" \",\"pages\":\"\"},\"PeriodicalIF\":5.9000,\"publicationDate\":\"2025-09-18\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Survey of ophthalmology\",\"FirstCategoryId\":\"3\",\"ListUrlMain\":\"https://doi.org/10.1016/j.survophthal.2025.09.007\",\"RegionNum\":2,\"RegionCategory\":\"医学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"OPHTHALMOLOGY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Survey of ophthalmology","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1016/j.survophthal.2025.09.007","RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"OPHTHALMOLOGY","Score":null,"Total":0}
Artificial intelligence in age-related macular degeneration: Advancing diagnosis, prognosis, and treatment.
Age-related macular degeneration (AMD) is a leading cause of irreversible vision loss in older adults. While anti-vascular endothelial growth factor (anti-VEGF) therapy and novel treatments for geographic atrophy have improved management, timely diagnosis and personalized intervention remain a challenge. Artificial intelligence (AI), such as machine learning and deep learning models, shows promise in AMD diagnosis, classification, and treatment planning. This review summarizes AI's recent advancements, highlights its clinical utility, and addresses key limitations for wider real-world implementation in AMD. We conducted systematic search of PubMed from its conception up to August 1, 2024. Studies utilizing AI-based algorithms for AMD management were identified and categorized into early detection/classification and prediction of disease progression/treatment response. Data extraction focused on AI model performance, imaging modalities, and clinical applicability. Of 193 records screened, 47 studies were included, in which 19 studies focused on early detection/classification and 28 on prediction of disease progression/treatment response. AI models demonstrated high accuracy in AMD classification and progression prediction, including in real-world settings. Prediction models for treatment response, particularly anti-VEGF therapy, could provide recommendations on optimizing injection timelines. Recent studies have also begun tackling previous challenges, such as algorithmic biases, limited generalizability, and AI's "black-box" nature. AI-based models offer significant potential to transform AMD care through timely detection and personalized treatment; however, clinical integration depends on improving model interpretability and validating tools across diverse populations. As AI continues to evolve, ongoing research is needed to refine AI models and support their translation into evidence-based, real-world applicability in AMD.
期刊介绍:
Survey of Ophthalmology is a clinically oriented review journal designed to keep ophthalmologists up to date. Comprehensive major review articles, written by experts and stringently refereed, integrate the literature on subjects selected for their clinical importance. Survey also includes feature articles, section reviews, book reviews, and abstracts.