A generalizable foundation model for intraoperative understanding across surgical procedures

TL;DR

This paper introduces ZEN, a large, diverse foundation model trained on over 4 million surgical video frames, which generalizes across procedures and enhances intraoperative understanding for AI-assisted surgical decision-making.

Contribution

The authors present ZEN, a novel self-supervised foundation model trained on extensive surgical data, demonstrating superior generalization across procedures and tasks compared to existing models.

Findings

ZEN outperforms existing models on multiple tasks

ZEN shows robust cross-procedure generalization

ZEN improves intraoperative scene understanding

Abstract

In minimally invasive surgery, clinical decisions depend on real-time visual interpretation, yet intraoperative perception varies substantially across surgeons and procedures. This variability limits consistent assessment, training, and the development of reliable artificial intelligence systems, as most surgical AI models are designed for narrowly defined tasks and do not generalize across procedures or institutions. Here we introduce ZEN, a generalizable foundation model for intraoperative surgical video understanding trained on more than 4 million frames from over 21 procedures using a self-supervised multi-teacher distillation framework. We curated a large and diverse dataset and systematically evaluated multiple representation learning strategies within a unified benchmark. Across 20 downstream tasks and full fine-tuning, frozen-backbone, few-shot and zero-shot settings, ZEN consistently outperforms existing surgical foundation models and demonstrates robust cross-procedure generalization. These results suggest a step toward unified representations for surgical scene understanding and support future applications in intraoperative assistance and surgical training assessment.