Traini Daniele Omar, Palmisano Gerardo, Guerriero Cristina, Peris Ketty
Dermatologia, Dipartimento Universitario di Medicina e Chirurgia Traslazionale, Università Cattolica del Sacro Cuore, 00168 Rome, Italy.
Dermatologia, Dipartimento Scienze Mediche e Chirurgiche, Fondazione Policlinico Universitario A. Gemelli IRCCS, 00168 Rome, Italy.
J Pers Med. 2025 Jun 6;15(6):238. doi: 10.3390/jpm15060238.
Acne vulgaris is a common dermatological condition, particularly affecting adolescents during critical developmental stages, which may have lasting psychosocial impacts. Traditional assessments, including global severity grading and lesion counting, are limited by subjectivity and time constraints. : This review aims to systematically assess the recent advancements in artificial intelligence (AI) applications for acne diagnosis, lesion segmentation/counting, and severity grading, highlighting the potential of AI-driven methods to improve objectivity, reproducibility, and clinical efficiency. : A comprehensive literature search was conducted across PubMed, Scopus, arXiv, Embase, and Web of Science for studies published between 1 January 2017 and 1 March 2025. The search strategy incorporated terms related to "acne" and various AI methodologies (e.g., "neural network", "deep learning", "convolutional neural network"). Two independent reviewers screened 345 articles, with 29 studies ultimately meeting inclusion criteria. Data were extracted on study design, dataset characteristics (including internal and publicly available resources such as ACNE04 and AcneSCU), AI architectures (predominantly CNN-based models), and performance metrics. : While AI-driven models demonstrated promising accuracy, as high as 97.6% in controlled settings, the limited availability of large public datasets, the predominance of data from specific ethnic groups, and the lack of extensive external validation underscore critical barriers to clinical implementation. : The findings indicate that although AI has the potential to standardize acne assessments, reduce observer variability, and enable self-monitoring via mobile platforms, significant challenges remain in achieving robust, real-world applicability. Future research should prioritize the development of large, diverse, and publicly accessible datasets and undertake prospective clinical validations to ensure equitable and effective dermatological care.
寻常痤疮是一种常见的皮肤病，尤其在关键发育阶段影响青少年，可能产生持久的心理社会影响。传统评估方法，包括整体严重程度分级和皮损计数，受主观性和时间限制。本综述旨在系统评估人工智能（AI）在痤疮诊断、皮损分割/计数及严重程度分级方面的最新进展，突出人工智能驱动方法在提高客观性、可重复性和临床效率方面的潜力。通过在PubMed、Scopus、arXiv、Embase和Web of Science上进行全面文献检索，查找2017年1月1日至2025年3月1日期间发表的研究。检索策略纳入了与“痤疮”及各种人工智能方法相关的术语（如“神经网络”“深度学习”“卷积神经网络”）。两名独立评审员筛选了345篇文章，最终29项研究符合纳入标准。提取了有关研究设计、数据集特征（包括内部和公开可用资源，如ACNE04和AcneSCU）、人工智能架构（主要是基于卷积神经网络的模型）及性能指标的数据。虽然人工智能驱动的模型显示出有前景的准确性，在受控环境中高达97.6%，但大型公共数据集的可用性有限、特定种族群体数据占主导以及缺乏广泛的外部验证凸显了临床应用的关键障碍。研究结果表明，尽管人工智能有潜力使痤疮评估标准化、减少观察者差异并通过移动平台实现自我监测，但在实现强大的现实世界适用性方面仍存在重大挑战。未来研究应优先开发大型、多样且可公开获取的数据集，并进行前瞻性临床验证，以确保公平有效的皮肤科护理。
