The RSNA 2026 Knee MRI Challenge is gaining remarkable momentum, and I am deeply grateful to everyone contributing. It’s been both a fulfilling and a humbling experience. As of August 29, more than 16,700 people have joined, with nearly 2,900 competitors forming over 2,600 teams and submitting more than 24,700 models.
Almost 500 public notebooks and 85 forum topics reflect a community that is not just competing, but actively teaching and learning. Participation and iterative activity continue to rise, while both leading and median public leaderboard performance have strengthened impressively.
What excites me most is the conversation. Data scientists, trainees, and licensed radiologists from around the world have posted on the discussion boards about the challenge design, the dataset, and the realities of clinical data. Discussions about why there is no beautifully crafted “ground truth,” and what to do with nuanced, free-text radiology reports, go directly to the educational mission of this challenge. These are not side questions; they are central to building useful, clinically meaningful AI.
I am extremely impressed by the competitors’ performance so far, and thankful for the rigor, curiosity, and generosity participants are bringing to the challenge.