Autonomous markerless augmented reality for perforator visualization in DIEP surgery

Augmented reality (AR) navigation is increasingly used in deep inferior epigastric perforator (DIEP) surgery to visualize obscured anatomy, but current methods rely on manual steps and auxiliary devices, raising surgical risks. We propose an autonomous, markerless AR framework that eliminates manual intervention and external localization. Biometric features of the abdominal surface are extracted and matched with an enhanced segment anything model, while markerless tracking is achieved via simultaneous localization and mapping. Validation was performed with phantom and clinical evaluation. In phantom experiment, mean target registration error was 2.24 mm with an average registration time of 58.7 s. In clinical evaluation, virtual models were accurately displayed from arbitrary viewpoints, and five DIEP-AR surgeries were successfully completed with an average error of 2.42 mm. The framework achieves precise automatic registration and robust markerless tracking, supporting efficient intraoperative visualization and demonstrating promise for clinical application.