基于对比学习的腹腔镜手术图像实时去雾研究  

Real-time Dehazing of Laparoscopic Surgery Images Based on Contrastive Learning

作　　者：张蓝亓 李智[1] 杨帆[1] Zhang Lan-qi;Li Zhi;Yang Fan(Sichuan University,Chengdu 610065,Sichuan Province,China)

机构地区：[1]四川大学,四川成都610065

出　　处：《科学与信息化》2025年第8期160-162,共3页Technology and Information

摘　　要：随着手术微创化发展的大势所趋,腹腔镜成为必不可少的手术器械。目前的腹腔镜手术中,多种原因引起雾气,严重降低实时手术图像传递质量,从而影响医生通过精细解剖实施重要手术步骤,降低手术质量,带来医疗安全隐患。针对此问题,本研究通过对比学习,构建一种实时去雾图像算法。该算法包含多层次去雾模块,可有效避免无烟区域颜色失真,并引入特征注意力模块,有针对性地捕捉图像中的有用信息。通过将模糊图像和清晰图像分别作为负样本和正样本,共同指导去雾网络的训练,以实现优异的去雾效果。同时引入动态特征增强模块,以确保医学上的原始特征不受损。With the increasing trend towards minimally invasive surgery,laparoscopy has become an essential surgical instrument.However,fogging in laparoscopic surgeries caused by various factors severely degrades the quality of real-time surgical images,thereby affecting the surgeon’s ability to perform critical surgical steps through precise dissection,reducing surgical quality,and posing medical safety hazards.To address this issue,this study constructs a real-time dehazing image algorithm based on contrastive learning.The algorithm includes a multi-level dehazing module that effectively avoids color distortion in smoke-free areas and introduces a feature attention module to selectively capture useful information in the images.By using foggy images as negative samples and clear images as positive samples to jointly guide the training of the dehazing network,the algorithm achieves excellent dehazing effects.Additionally,a dynamic feature enhancement module is introduced to ensure that the original medical features are not compromised.

关 键 词：对比学习 图像去雾 腹腔镜图像 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]