TLDR: The ASNR MICCAI BraTS 2025 Lighthouse Challenge developed an educational platform to train medical students and radiology trainees in AI-assisted neuroradiology. Through guided image annotation, lectures, and workshops, participants significantly improved their familiarity with image segmentation software and brain tumor features. The initiative highlights the critical need for integrating AI education into medical curricula to prepare future physicians for an AI-driven clinical environment, while also generating high-quality annotated datasets for AI algorithm development.
The integration of Artificial Intelligence (AI) into clinical practice, particularly in fields like neuroradiology, is rapidly transforming healthcare. This evolution necessitates a new approach to medical education, ensuring that future physicians are well-equipped to understand, utilize, and critically evaluate AI-assisted tools. A recent initiative, the ASNR MICCAI Brain Tumor Segmentation (BraTS) 2025 Lighthouse Challenge, has developed an innovative educational platform to address this crucial need.
The core of this initiative is a multimodal educational approach designed for medical students and radiology trainees. It combines traditional didactic lectures with interactive, hands-on learning experiences, primarily through guided data annotation. The goal is to enhance trainees’ understanding of AI algorithm development, reinforce the concept of data reference standards, and diversify opportunities for AI-driven image analysis among future physicians.
The project built upon experiences from the BraTS Challenges of 2023 and 2024, where 56 volunteer medical students and radiology trainees participated in preliminary data annotation. These “annotation coordinators” were guided by faculty-led discussions on neuropathology MRI. A survey of 54 medical students revealed a strong belief (93%) that AI would influence their careers, yet 87% reported no AI-focused education at their institutions, highlighting a significant gap that this program aims to fill.
For the BraTS Lighthouse Challenge 2025, a unique one-on-one mentorship model was implemented. Fourteen select volunteers were paired with board-certified neuroradiology faculty for guided annotation sessions. These sessions focused on annotating brain metastases, meningiomas, and glioblastomas on MR images. The annotation process involved four stages, including manual annotation from scratch and refinement of automated pre-segmented images, designed to ensure high-quality data and deep learning for the trainees.
The results were compelling. Annotation coordinators reported a significant increase in their familiarity with image segmentation software, moving from an average score of 6 to 8.9 on a 1-10 scale. Similarly, their familiarity with brain tumor features improved from 6.2 to 8.1. This demonstrates the effectiveness of interactive learning and one-on-one mentorship in developing both technical skills and clinical knowledge.
Beyond the annotation sessions, the educational platform expanded its reach through an online presence. Lectures on neuroanatomy, pathology, and AI were organized, along with journal clubs and workshops led by data scientists. These resources, some of which are available on YouTube, provided asynchronous learning opportunities and further enriched the educational experience. For instance, workshops introduced beginners to algorithm development and strategies for improving accuracy in AI training, even leading to two algorithm submissions to the BraTS 2025 Lighthouse Challenge by trainees.
The lectures covered a range of topics, from clinically relevant MRI neuroanatomy to the fundamentals of AI in brain tumor imaging. Surveys from 97 lecture attendees showed that 95% felt their knowledge of the topics improved after the sessions, with a majority reporting a “moderate” to “very good” understanding post-lecture.
This initiative underscores the critical need for medical education reform to incorporate AI as a core component of curricula. It provides a blueprint for integrating AI-focused skills into radiology training by coupling the generation of high-quality annotated imaging datasets with comprehensive education. The project not only prepares future radiologists for an AI-integrated clinical environment but also actively contributes valuable reference standard data to the scientific community for further AI algorithm development. For more details, you can refer to the full research paper.
Also Read:
- Multi-Agent AI Framework Enhances Radiology Report Generation and Evaluation
- Generative AI’s Impact on Medical Education: A Comprehensive Review of Efficacy and Future Prospects
The authors propose strategies for medical educators to develop formal AI-focused education in medical school and residency, emphasizing interactive lectures, workshops, and real-world application scenarios. This approach fosters technical expertise in AI tools alongside a strong understanding of radiological principles, ensuring that the next generation of physicians can effectively navigate the evolving landscape of AI in healthcare.
