新兴人工智能:美容整形领域的新数字时代
本综述梳理了26项将人工智能或计算方法应用于社交媒体、在线评论等数字内容、并报告患者或公众感知相关结局的整形外科原创研究,发现研究数量自2024年快速增长,但多为描述性、单平台且缺乏验证。
要点
- 美容整形领域AI研究集中于面部美容手术,情感分析是最常用方法。
- 多数研究仅分析单一平台文本数据,使用未针对研究数据训练的预训练模型。
- 仅少数研究报告人工对照或外部验证,无研究链接患者生成数据与经验证的患者报告结局。
核心结论
- 该交叉领域研究快速扩展,但仍以描述性、单平台和 largely unvalidated 为特征。
研究设计与样本
- 综述纳入26项2019年至2026年间发表的原创实证研究,这些研究将AI或计算方法应用于社交媒体、在线评论、公共数字图像、调查或众包数据,并报告患者或公众感知相关结局。
主要发现
- 26项研究中19项(73%)发表于2024年及以后。
- 面部美容手术是代表性最高的领域(n=9;35%)。
- RealSelf是最常见的主要数据来源(n=8;31%),18项研究(69%)分析单一平台,18项(69%)仅分析文本。
- 情感或情绪分析是最常用的计算方法(n=13;50%),16项研究(62%)直接使用发布的预训练模型或词典而未在研究数据上训练。
- 10项研究(38%)报告了人工对照,1项(4%)进行了外部验证;无研究将患者生成数字数据与经验证的患者报告结局指标关联,也无研究评估这些数据用于外科教育。
局限性
- 现有研究多为描述性、单平台且 largely unvalidated,缺乏外部验证和与患者报告结局的链接。
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
Patients increasingly form expectations about plastic surgery from social media and other online sources, and artificial intelligence (AI) now functions both as a tool for analyzing that content and as a source of the content itself. This review maps original empirical studies that applied an AI or computational method beyond conventional statistical analysis to social media, online reviews, public digital images, surveys, or crowdsourced data in plastic surgery while reporting an outcome related to the perceptions, questions, experiences, or information exposure of patients or the public. Twenty-six studies published between 2019 and 2026 met inclusion criteria, 19 (73%) of them since 2024. Facial aesthetic surgery was the most represented domain (n = 9; 35%). RealSelf was the most common primary data source (n = 8; 31%), 18 studies (69%) analyzed a single platform, and 18 (69%) analyzed text alone. Sentiment or emotion analysis was the most common computational method (n = 13; 50%), and 16 studies (62%) applied pretrained models or lexicons as released, without training on study data. A human comparator was reported in 10 studies (38%) and external validation in 1 (4%). No study linked patient-generated digital data to a validated patient-reported outcome measure or evaluated these data for surgical education. Work at this intersection is expanding rapidly but remains descriptive, single-platform, and largely unvalidated.
引用
- Gould M, Dhawan R, Zucker E, et al. Emerging Artificial Intelligence: A New Digital Age Across The Aesthetic Landscape. Aesthet Surg J. 2026. PMID: 42925558. doi:10.1093/asj/sjag241
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。