AR system slashes tumor margin errors by 71% in head and neck surgery

A team led by Yue Yang (Vanderbilt University) has published a marker-free augmented reality system to help surgeons locate positive margins during head and neck cancer surgery. The current standard relies on verbal descriptions from pathologists, leading to average localization errors of 21mm. The new workflow digitizes the resected specimen, deforms it via contour constraints and suture-based correspondences, then fuses the margin label onto the patient's resection bed through a head-mounted display—all without external fiducial markers.

In cadaveric experiments with five surgeons, AR guidance reduced end-to-end margin error to 6.19mm—a 71% improvement over verbal guidance (p < 0.001). Deformation target errors were 7.63mm (cheek) and 3.72mm (scalp), while marker-free fusion matched marker-based accuracy at ~2.15mm. The entire online fusion process took just over 5 seconds. The system's ability to work without external markers makes it practical for real surgical settings where fiducial placement is often infeasible.