Distilling Expert Surgical Knowledge: How to train local surgical VLMs for anatomy explanation in Complete Mesocolic Excision

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Maack, Lennart, Graß, Julia-Kristin, Toscha, Lisa-Marie, Melling, Nathaniel, Schlaefer, Alexander
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909946467057664
author Maack, Lennart
Graß, Julia-Kristin
Toscha, Lisa-Marie
Melling, Nathaniel
Schlaefer, Alexander
author_facet Maack, Lennart
Graß, Julia-Kristin
Toscha, Lisa-Marie
Melling, Nathaniel
Schlaefer, Alexander
contents Recently, Vision Large Language Models (VLMs) have demonstrated high potential in computer-aided diagnosis and decision-support. However, current VLMs show deficits in domain specific surgical scene understanding, such as identifying and explaining anatomical landmarks during Complete Mesocolic Excision. Additionally, there is a need for locally deployable models to avoid patient data leakage to large VLMs, hosted outside the clinic. We propose a privacy-preserving framework to distill knowledge from large, general-purpose LLMs into an efficient, local VLM. We generate an expert-supervised dataset by prompting a teacher LLM without sensitive images, using only textual context and binary segmentation masks for spatial information. This dataset is used for Supervised Fine-Tuning (SFT) and subsequent Direct Preference Optimization (DPO) of the locally deployable VLM. Our evaluation confirms that finetuning VLMs with our generated datasets increases surgical domain knowledge compared to its base VLM by a large margin. Overall, this work validates a data-efficient and privacy-conforming way to train a surgical domain optimized, locally deployable VLM for surgical scene understanding.
format Preprint
id arxiv_https___arxiv_org_abs_2512_05740
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Distilling Expert Surgical Knowledge: How to train local surgical VLMs for anatomy explanation in Complete Mesocolic Excision
Maack, Lennart
Graß, Julia-Kristin
Toscha, Lisa-Marie
Melling, Nathaniel
Schlaefer, Alexander
Computer Vision and Pattern Recognition
Recently, Vision Large Language Models (VLMs) have demonstrated high potential in computer-aided diagnosis and decision-support. However, current VLMs show deficits in domain specific surgical scene understanding, such as identifying and explaining anatomical landmarks during Complete Mesocolic Excision. Additionally, there is a need for locally deployable models to avoid patient data leakage to large VLMs, hosted outside the clinic. We propose a privacy-preserving framework to distill knowledge from large, general-purpose LLMs into an efficient, local VLM. We generate an expert-supervised dataset by prompting a teacher LLM without sensitive images, using only textual context and binary segmentation masks for spatial information. This dataset is used for Supervised Fine-Tuning (SFT) and subsequent Direct Preference Optimization (DPO) of the locally deployable VLM. Our evaluation confirms that finetuning VLMs with our generated datasets increases surgical domain knowledge compared to its base VLM by a large margin. Overall, this work validates a data-efficient and privacy-conforming way to train a surgical domain optimized, locally deployable VLM for surgical scene understanding.
title Distilling Expert Surgical Knowledge: How to train local surgical VLMs for anatomy explanation in Complete Mesocolic Excision
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.05740