Title : Artificial intelligence use in acne diagnosis and management - A scoping review
Abstract:
Acne vulgaris is one of the most prevalent skin conditions in dermatology, affecting 9.38% of the global population. Acne holds the potential to significantly impact both the physical and psychological health of patients. Conditions such as permanent, irreversible scarring can negatively influence self-image. Artificial intelligence (AI) techniques can allow for early diagnosis and treatment of acne, which may help mitigate these potential adverse consequences.
Objective: To determine: 1) the types of AI based tools developed for acne 2) the various applications of AI in acne diagnosis and management and 3) the performance of these tools.
Methods: We queried PubMed, Cochrane and Scopus databases using the following terms: “acne”, “artificial intelligence”, “machine learning”, “deep learning”, “large language model”, and “chatgpt”. We discovered 292 articles, with 131 articles that met the eligibility criteria. After reviewing the manuscript, 105 relevant articles were included for analysis.
Results: Of the 105 research articles, (96.2%, N=101) were focused on acne diagnosis only, 9.5% (N=10) on acne management only, and (5.7%, N=6) on both. Most manuscripts used image-based models, including deep learning (76.2%, N=80), classical machine learning techniques (9.5%, N=10), and hybrid models, which use multiple models to form an ensemble (11.4%, N=12). In contrast, only (2.9% N=3) papers used language-based models. The ensemble models had the highest mean accuracy (89.7%), followed by deep learning (88.5%), large language models (87.5%) and machine learning models (86.9%). All models are further evaluated individually by each respective task.
Conclusions And Relevance: Given the visual nature of dermatology, the vast majority of manuscripts are focused on image-based AI models. Reflecting previous literature, ensemble models demonstrated superior performance followed by deep learning models. This scoping review identified several limitation themes across multiple manuscripts, including small dataset size, variation in image quality, skewed Fitzpatrick representation, proprietary datasets, and limited representation of alternate anatomic locations other than the face. Future work can enhance model performance and equality in the diagnostics and management of acne.
