The increasing integration of Artificial Intelligence (AI) into dermatology holds significant potential to enhance the diagnosis and prognosis of skin diseases, which represent a considerable public health burden. While numerous AI models for skin conditions have been developed, their successful translation into real-world clinical practice critically depends on rigorous external validation to ensure generalizability and reliability beyond their development environment. Despite the growing recognition of external validation’s importance in medical AI, a comprehensive synthesis specifically addressing this process for AI models in dermatology has been lacking. This systematic literature review aimed to identify and synthesize studies focused exclusively on the external validation of AI models for skin diseases, thereby assessing their clinical applicability, methodological robustness, and identifying key challenges. Following the PRISMA framework, a systematic search was conducted across five major digital libraries (ACM Digital Library, IEEE Xplore, Web of Science, PubMed, and Scopus). Study quality was assessed using a customized checklist, and data extraction focused on validation practices, model characteristics, targeted diseases, performance assessment, and identified limitations. The systematic study selection led to a detailed analysis of 41 studies. We observed a clear predominance of retrospective external validation studies focused on diagnostic models, while prognostic models and prospective validation designs were notably scarce. Although some models demonstrated promising performance in external validation, several critical limitations persist. These include the narrow representation of skin disease categories, lack of subgroup fairness analyses, small and imbalanced sample sizes, and insufficient standardization of data collection and evaluation metrics. These findings underscore the need for more rigorous and transparent validation frameworks to support the reliable integration of AI into dermatological practice.

This research letter presents the co-design process for RG4Face, a mime therapy–based serious game that uses computer vision for human facial movement recognition and estimation to help health care professionals and patients in the facial rehabilitation process.

This study aimed to compare the performance of various Large Language Models (LLMs) in answering multiple-choice questions from the last six editions (2017 to 2024) of Revalida exam. The evaluation focused on models capable of processing content in Brazilian Portuguese (PT-BR), including open-source models, namely LLaMA 3.1 (8B parameters), Qwen 2.5 (7B parameters), and their reasoning-oriented distilled variants based on the DeepSeek-R1 architecture, as well as open-access commercial models such as GPT-3.5, GPT-4o, and Gemini. After evaluating the models’ accuracy against the official answer keys, GPT-4o emerged as the top-performing model, achieving an average accuracy of 63.85%. Next, GPT-4o was prompted to justify its answers to the 2024 exam, and its explanations were independently reviewed by three licensed physicians. The evaluators reported full agreement with the clinical reasoning presented, indicating the model’s ability to produce coherent and medically relevant justifications. Lastly, justifications generated by GPT-4o for correctly answered questions from previous exams (2017–2023) were compiled into a knowledge base, which was then used to enhance GPT-4o through retrieval-augmented generation and to fine-tune LLaMA 3.1, leading to measurable performance improvements on the 2024 exam. Despite promising performance, these models still demonstrate variability in responses, hallucinations, and limited reliability in high-stakes contexts. As such, their outputs should always be reviewed by qualified professionals, and human expertise remains essential in clinical decision-making and medical education scenarios, considering the PT-BR language. However, the observed gains from integrating prior exam content indicate that domain-specific adaptation strategies may help mitigate some of these limitations and enhance model alignment.

Ensinar programação em contextos educacionais ainda representa um desafio, marcado pela abstração conceitual, pela dificuldade de aplicar teoria na prática e pelo desinteresse dos estudantes. Habilidades como raciocínio lógico, resolução de problemas e pensamento computacional, muitas vezes pouco desenvolvidas nos anos iniciais, intensificam essas dificuldades. Esta revisão integrativa buscou identificar ferramentas e metodologias que favoreçam o ensino e a aprendizagem de algoritmos e programação, oferecendo subsídios para práticas pedagógicas mais eficazes. Foram analisados 1.024 estudos publicados entre 2019 e 2024, dos quais 20 atenderam aos critérios de seleção e apresentaram evidências relevantes. Os resultados indicam que estratégias como gamificação, plataformas interativas e jogos educacionais aumentam o engajamento e facilitam a aprendizagem. A robótica educacional também se destaca por associar programação a atividades práticas, promovendo desenvolvimento técnico, cognitivo e socioemocional. Além disso, linguagens em blocos, como Scratch e Blockly, favorecem a compreensão inicial dos conceitos ao reduzir a complexidade sintática e apoiar o foco na lógica algorítmica. Apesar dos avanços, limitações estruturais persistem, sobretudo em escolas com baixo acesso à tecnologia. Assim, políticas públicas, formação docente contínua e investimentos em infraestrutura são essenciais para ampliar a adoção dessas práticas e garantir um ensino inclusivo e de qualidade em algoritmos e programação.