Gregory Schuit

Generative image models have achieved remarkable progress in both natural and medical imaging. In the medical context, these techniques offer a potential solution to data scarcity-especially for low-prevalence anomalies that impair the performance of AI-driven diagnostic and segmentation tools. However, questions remain regarding the fidelity and clinical utility of synthetic images, since poor generation quality can undermine model generalizability and trust. In this study, we evaluate the effectiveness of state-of-the-art generative models-Generative Adversarial Networks (GANs) and Diffusion Models (DMs)-for synthesizing chest X-rays conditioned on four abnormalities: Atelectasis (AT), Lung Opacity (LO), Pleural Effusion (PE), and Enlarged Cardiac Silhouette (ECS). Using a benchmark composed of real images from the MIMIC-CXR dataset and synthetic images from both GANs and DMs, we conducted a reader study with three radiologists of varied experience. Participants were asked to distinguish real from synthetic images and assess the consistency between visual features and the target abnormality. Our results show that while DMs generate more visually realistic images overall, GANs can report better accuracy for specific conditions, such as absence of ECS. We further identify visual cues radiologists use to detect synthetic images, offering insights into the perceptual gaps in current models. These findings underscore the complementary strengths of GANs and DMs and point to the need for further refinement to ensure generative models can reliably augment training datasets for AI diagnostic systems.

This article presents a controlled case study focused on implementing and using Generative Artificial Intelligence (AI), specifically Large Language Models (LLM) in physiotherapy education to assist instructors with formulating effective technology-mediated feedback for students. It outlines how these advanced technologies have been integrated into an existing feedback-oriented platform to guide instructors in providing feedback inputs and establish a reference framework for future innovations in practical skills training for health professions education. Specifically, the proposed solution uses LLM to automatically evaluate feedback inputs made by instructors based on pre-defined and literature-based quality criteria and generates actionable textual explanations for reformulation. In addition, if the instructor requires, the tool supports summary generation for large sets of text inputs to achieve better student reception and understanding. The case study describes how these features were integrated into the feedback-oriented platform, how their effectiveness was evaluated in a controlled setting with documented feedback inputs, and the results of its implementation with real users through cognitive walkthroughs. Initial results indicate that this innovative implementation holds great potential to enhance learning and performance in physiotherapy education and has the potential to expand to other health disciplines where the development of procedural skills is critical, offering a valuable tool to assess and improve feedback based on quality standards for effective feedback processes. The cognitive walkthroughs allowed us to determine participants' usability decisions in the face of these new features and to evaluate the perceived usefulness, how this would integrate into their workload, and their opinion regarding the potential for the future within this teaching strategy. The article concludes with a discussion of the implications of these findings for practice and future research directions in this developing field.

Remote technology has been widely incorporated into health professions education. For procedural skills training, effective feedback and reflection processes are required. Consequently, supporting a self-regulated learning (SRL) approach with learning analytics dashboards (LADs) has proven beneficial in online environments. Despite the potential of LADs, understanding their design to enhance SRL and provide useful feedback remains a significant challenge. Focusing on LAD design, implementation, and evaluation, the study followed a mixed-methods two-phase design-based research approach. The study used a triangulation methodology of qualitative interviews and SRL and sensemaking questionnaires to comprehensively understand the LAD’s effectiveness and student SRL and feedback uptake strategies during remote procedural skills training. Initial findings revealed the value students placed on performance visualization and peer comparison despite some challenges in LAD design and usability. The study also identified the prominent adoption of SRL strategies such as help-seeking, elaboration, and strategic planning. Sensemaking results showed the value of personalized performance metrics and planning resources in the LAD and recommendations to improve reflection and feedback uptake. Subsequent findings suggested that SRL levels significantly predicted the levels of sensemaking. The students valued the LAD as a tool for supporting feedback uptake and strategic planning, demonstrating the potential for enhancing procedural skills learning.

RESUMO O cenário do treinamento cirúrgico está evoluindo rapidamente com o surgimento da inteligência artificial (IA) e sua integração na educação e simulação. Este artigo explora as aplicações e benefícios potenciais do treinamento cirúrgico assistido por IA, em particular o uso de modelos de linguagem avançados (MLAs), para aprimorar a comunicação, personalizar o feedback e promover o desenvolvimento de habilidades. Discutimos os avanços no treinamento baseado em simulação, ferramentas de avaliação impulsionadas por IA, sistemas de avaliação baseados em vídeo, plataformas de realidade virtual (RV) e realidade aumentada (RA), e o papel potencial dos MLAs na transcrição, tradução e resumo do feedback. Apesar das oportunidades promissoras apresentadas pela integração da IA, vários desafios devem ser abordados, incluindo precisão e confiabilidade, preocupações éticas e de privacidade, viés nos modelos de IA, integração com os sistemas de treinamento existentes, e treinamento e adoção de ferramentas assistidas por IA. Ao abordar proativamente esses desafios e aproveitar o potencial da IA, o futuro do treinamento cirúrgico pode ser remodelado para proporcionar uma experiência de aprendizado mais abrangente, segura e eficaz para os aprendizes, resultando em melhores resultados para os pacientes.

The landscape of surgical training is rapidly evolving with the advent of artificial intelligence (AI) and its integration into education and simulation. This manuscript aims to explore the potential applications and benefits of AI-assisted surgical training, particularly the use of large language models (LLMs), in enhancing communication, personalizing feedback, and promoting skill development. We discuss the advancements in simulation-based training, AI-driven assessment tools, video-based assessment systems, virtual reality (VR) and augmented reality (AR) platforms, and the potential role of LLMs in the transcription, translation, and summarization of feedback. Despite the promising opportunities presented by AI integration, several challenges must be addressed, including accuracy and reliability, ethical and privacy concerns, bias in AI models, integration with existing training systems, and training and adoption of AI-assisted tools. By proactively addressing these challenges and harnessing the potential of AI, the future of surgical training may be reshaped to provide a more comprehensive, safe, and effective learning experience for trainees, ultimately leading to better patient outcomes. .