Nikon disqualifies microscopy contest winner after AI use revealed

The first-place video in Nikon's Small World in Motion competition has been disqualified after the company determined it violated rules against generative AI. The entry, submitted by Dr. Ning Xu, appeared to show the movement of cilia — tiny hair-like structures — in the airways of a child with primary ciliary dyskinesia, a rare genetic disorder that impairs ciliary motion. After doubts about the footage's authenticity surfaced online, Nikon opened an investigation and announced the disqualification last week.

In a LinkedIn post, Dr. Xu acknowledged using an unsupervised neural-network method for AI-assisted post-processing. The aim, he said, was to distinguish and illustrate features in grayscale images produced by super-resolution optical imaging. Nikon stressed that the decision "should not be interpreted as a judgment of the professional reputation, scientific contributions, or intent of the submitter."

Following the disqualification, a video by Nguyen Nam Nhat moves into first place in the updated rankings. Dr. Xu's entry has been removed from Nikon's website. The company said it intends to "re-examine the rules and evaluation procedures for future submissions," a step that reflects the growing difficulty of separating legitimate scientific image processing from synthetic content that alters the observed reality.

The case illustrates the central problem facing scientific imaging competitions today: the line between enhancing noisy signals — a long-accepted practice in microscopy — and adding details the sensor never captured is blurring. Neural-network-based methods can "fill in" missing information in ways that look entirely convincing to the human eye, yet may depict structures that do not exist in the original sample. Nikon declined to specify which parts of the video were model-generated, stating only that the rules had been violated.

Small World in Motion is regarded as one of the most prestigious venues for scientific motion imaging, and the public disqualification is likely to pressure organizers of similar competitions, as well as scientific publishers, to define clearer policies on the use of AI tools in the imaging pipeline. Until now, most rulebooks addressed manual editing or classical filters; they now need explicit provisions for generative models capable of inventing detail at the single-pixel level.