Answers are grounded in this paper and cite their sources.
Artificial intelligence (AI) and digital imaging technologies are increasingly used in clinical skin assessment, enabling automated and standardized evaluation of visible skin conditions. Although advances have been reported in wound care and dermatology, automated tuberculin skin test (TST) reading remains incipient, despite its importance for diagnosing latent tuberculosis infection (LTBI) and achieving the End TB Strategy targets. This scoping review aimed to map digital imaging technologies for automated assessment of externally visible skin conditions, emphasizing methods potentially applicable to TST reading. The review followed JBI methodology and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews (PRISMA-ScR). Searches covered bibliographic databases and grey literature sources, including studies published between 2021 and 2025 in English, Portuguese, and Spanish. Eligible studies employed digital imaging for automated assessment of visible skin conditions. Data were descriptively synthesized according to technology type and clinical application. Of 1,225 records identified, five studies met the inclusion criteria. Convolutional neural network-based approaches predominated in detecting and measuring skin abnormalities, showing promising performance in controlled settings but limited external validation. Only one study specifically investigated automated TST reading. Mobile applications showed potential for remote assessment, although evidence of their performance and feasibility in primary health care remains limited. Considerable heterogeneity was observed in image acquisition, dataset characteristics, and evaluation methods. Digital imaging technologies may support automated skin assessment; however, evidence remains limited and heterogeneous. Further multicenter studies are needed to assess external validity, performance across diverse populations and devices, real-world feasibility, and ethical data governance. Specifically, evidence remains insufficient to support routine implementation of automated TST reading.
No takes yet. Share an insight, caveat, or question.
Stolarz et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: