Xiang Ao, Yanlin Leng, Yunfan Gan, Zhitao Cheng, Haomin Lin, Jiayi Fan QiaoLi, Junfeng Wang, Tingfang Wu, Linru Zhou, Haoxin Li, Liu Zheng, Yong Tang, Song Su, Jiali Wu
{"title":"Gauze detection and segmentation in laparoscopic liver surgery: a multi-center study.","authors":"Xiang Ao, Yanlin Leng, Yunfan Gan, Zhitao Cheng, Haomin Lin, Jiayi Fan QiaoLi, Junfeng Wang, Tingfang Wu, Linru Zhou, Haoxin Li, Liu Zheng, Yong Tang, Song Su, Jiali Wu","doi":"10.1186/s40001-025-03190-2","DOIUrl":null,"url":null,"abstract":"<p><strong>Background: </strong>Surgical gauze is commonly used in laparoscopic procedures. However, due to its small size, limited visibility to the naked eye, and subtle tactile presence, gauze can easily be overlooked during surgery.</p><p><strong>Purpose: </strong>To support surgeons in detecting gauze, reduce the risk of omission, and enhance surgical safety and efficiency, we developed a deep learning framework designed to identify gauze in laparoscopic liver surgeries.</p><p><strong>Methods: </strong>In total, 33 laparoscopic liver surgery videos were collected from 2 hospitals and used as internal and external datasets. Deep learning models were trained on individual video frames to detect and segment gauze, with each frame containing one or more gauze objects. To better evaluate model performance, we introduced a quantitative approach that classified gauze detection difficulty into three categories: easy, moderate, and difficult. Model performance across categories was assessed using data from both centers.</p><p><strong>Results: </strong>Among the tested models, YOLOv8n achieved the highest accuracy in gauze detection. In the internal test set, recall and precision reached 0.8443 and 0.9034, while in the external test set, they were 0.8289 and 0.9103. For gauze segmentation, the FCN-ResNet101 model demonstrated superior performance, achieving Dice scores of 0.9389 in the internal test set and 0.9081 in the external test set.</p><p><strong>Conclusion: </strong>The findings highlight the strong capability of the proposed framework to detect and segment gauze across varying levels of background complexity in laparoscopic liver surgery videos. This approach has the potential to significantly improve gauze management during surgery.</p>","PeriodicalId":11949,"journal":{"name":"European Journal of Medical Research","volume":"30 1","pages":"905"},"PeriodicalIF":3.4000,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12481765/pdf/","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"European Journal of Medical Research","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1186/s40001-025-03190-2","RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"MEDICINE, RESEARCH & EXPERIMENTAL","Score":null,"Total":0}
引用次数: 0
Abstract
Background: Surgical gauze is commonly used in laparoscopic procedures. However, due to its small size, limited visibility to the naked eye, and subtle tactile presence, gauze can easily be overlooked during surgery.
Purpose: To support surgeons in detecting gauze, reduce the risk of omission, and enhance surgical safety and efficiency, we developed a deep learning framework designed to identify gauze in laparoscopic liver surgeries.
Methods: In total, 33 laparoscopic liver surgery videos were collected from 2 hospitals and used as internal and external datasets. Deep learning models were trained on individual video frames to detect and segment gauze, with each frame containing one or more gauze objects. To better evaluate model performance, we introduced a quantitative approach that classified gauze detection difficulty into three categories: easy, moderate, and difficult. Model performance across categories was assessed using data from both centers.
Results: Among the tested models, YOLOv8n achieved the highest accuracy in gauze detection. In the internal test set, recall and precision reached 0.8443 and 0.9034, while in the external test set, they were 0.8289 and 0.9103. For gauze segmentation, the FCN-ResNet101 model demonstrated superior performance, achieving Dice scores of 0.9389 in the internal test set and 0.9081 in the external test set.
Conclusion: The findings highlight the strong capability of the proposed framework to detect and segment gauze across varying levels of background complexity in laparoscopic liver surgery videos. This approach has the potential to significantly improve gauze management during surgery.
期刊介绍:
European Journal of Medical Research publishes translational and clinical research of international interest across all medical disciplines, enabling clinicians and other researchers to learn about developments and innovations within these disciplines and across the boundaries between disciplines. The journal publishes high quality research and reviews and aims to ensure that the results of all well-conducted research are published, regardless of their outcome.