CoLM: Contrastive learning and multiple instance learning network for lung cancer classification of surgical options based on frozen pathological images
{"title":"CoLM: Contrastive learning and multiple instance learning network for lung cancer classification of surgical options based on frozen pathological images","authors":"","doi":"10.1016/j.bspc.2024.107097","DOIUrl":null,"url":null,"abstract":"<div><div>Histopathological images are regarded as the gold standard in cancer diagnosis. Formalin-fixed paraffin-embedded (FFPE) tissues are routinely collected and archived for pathological examination. However, the time-consuming procedures of tissue fixation and embedding render FFPE tissues unsuitable for intraoperative diagnosis, where immediate results are crucial during surgical procedures. In contrast, obtaining a fresh frozen section (FS) takes a very short time. FS samples are widely utilized for intraoperative diagnosis, whereas the diagnostic accuracy of FS is currently limited by the presence of potential histological artifacts. In this paper, we propose a contrastive learning image translation and multiple instance learning network (CoLM) for lung cancer classification. CoLM efficiently translates FS images into FFPE-style images and facilitates whole slide image classification. The entire framework encompasses two crucial stages. In the first stage, we employ a contrastive learning translation network with a dual-attention module (CL-DAM) for image translation. In the second stage, we utilize a hybrid transformer multi-instance learning-based network (HTM) to address the challenge posed by weak labels. We conduct experiments on lung cancer datasets to validate the performance of our proposed approach. The results demonstrate that our method achieve superior classification performance over other state-of-the-art methods, effectively mitigating the impact of blurred FS images. The proposed framework not only elevates the precision of intraoperative diagnosis when employing FS but also provides valuable reference for pathologists through the application of synthetic images.</div></div>","PeriodicalId":55362,"journal":{"name":"Biomedical Signal Processing and Control","volume":null,"pages":null},"PeriodicalIF":4.9000,"publicationDate":"2024-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Biomedical Signal Processing and Control","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1746809424011558","RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENGINEERING, BIOMEDICAL","Score":null,"Total":0}
引用次数: 0
Abstract
Histopathological images are regarded as the gold standard in cancer diagnosis. Formalin-fixed paraffin-embedded (FFPE) tissues are routinely collected and archived for pathological examination. However, the time-consuming procedures of tissue fixation and embedding render FFPE tissues unsuitable for intraoperative diagnosis, where immediate results are crucial during surgical procedures. In contrast, obtaining a fresh frozen section (FS) takes a very short time. FS samples are widely utilized for intraoperative diagnosis, whereas the diagnostic accuracy of FS is currently limited by the presence of potential histological artifacts. In this paper, we propose a contrastive learning image translation and multiple instance learning network (CoLM) for lung cancer classification. CoLM efficiently translates FS images into FFPE-style images and facilitates whole slide image classification. The entire framework encompasses two crucial stages. In the first stage, we employ a contrastive learning translation network with a dual-attention module (CL-DAM) for image translation. In the second stage, we utilize a hybrid transformer multi-instance learning-based network (HTM) to address the challenge posed by weak labels. We conduct experiments on lung cancer datasets to validate the performance of our proposed approach. The results demonstrate that our method achieve superior classification performance over other state-of-the-art methods, effectively mitigating the impact of blurred FS images. The proposed framework not only elevates the precision of intraoperative diagnosis when employing FS but also provides valuable reference for pathologists through the application of synthetic images.
期刊介绍:
Biomedical Signal Processing and Control aims to provide a cross-disciplinary international forum for the interchange of information on research in the measurement and analysis of signals and images in clinical medicine and the biological sciences. Emphasis is placed on contributions dealing with the practical, applications-led research on the use of methods and devices in clinical diagnosis, patient monitoring and management.
Biomedical Signal Processing and Control reflects the main areas in which these methods are being used and developed at the interface of both engineering and clinical science. The scope of the journal is defined to include relevant review papers, technical notes, short communications and letters. Tutorial papers and special issues will also be published.