Can LLM-Generated Text Empower Surgical Vision-Language Pre-training?

TL;DR

This paper introduces SurgLIME, a novel framework for surgical vision-language pre-training that effectively utilizes noisy LLM-generated narratives to enhance multi-modal understanding without degrading visual priors.

Contribution

It presents LIME, a scalable surgical video dataset with LLM-generated annotations, and SurgLIME, a parameter-efficient VLP method that mitigates noise through confidence estimation and preserves medical priors.

Findings

SurgLIME achieves competitive zero-shot cross-modal alignment.

The framework maintains robust linear probing performance.

Public dataset, code, and models are available at the provided GitHub link.

Abstract

Recent advancements in self-supervised learning have led to powerful surgical vision encoders capable of spatiotemporal understanding. However, extending these visual foundations to multi-modal reasoning tasks is severely bottlenecked by the prohibitive cost of expert textual annotations. To overcome this scalability limitation, we introduce \textbf{LIME}, a large-scale multi-modal dataset derived from open-access surgical videos using human-free, Large Language Model (LLM)-generated narratives. While LIME offers immense scalability, unverified generated texts may contain errors, including hallucinations, that could potentially lead to catastrophically degraded pre-trained medical priors in standard contrastive pipelines. To mitigate this, we propose \textbf{SurgLIME}, a parameter-efficient Vision-Language Pre-training (VLP) framework designed to learn reliable cross-modal alignments using noisy narratives. SurgLIME preserves foundational medical priors using a LoRA-adapted dual-encoder architecture and introduces an automated confidence estimation mechanism that dynamically down-weights uncertain text during contrastive alignment. Evaluations on the AutoLaparo and Cholec80 benchmarks show that SurgLIME achieves competitive zero-shot cross-modal alignment while preserving the robust linear probing performance of the visual foundation model. Dataset, code, and models are publicly available at https://github.com/visurg-ai/SurgLIME.