Zihan Li, Yunxiang Li, Qingde Li, You Zhang, Puyang Wang, Dazhou Guo, Le Lu, D. Jin, Qingqi Hong
{"title":"LViT: Language meets Vision Transformer in Medical Image Segmentation","authors":"Zihan Li, Yunxiang Li, Qingde Li, You Zhang, Puyang Wang, Dazhou Guo, Le Lu, D. Jin, Qingqi Hong","doi":"10.48550/arXiv.2206.14718","DOIUrl":null,"url":null,"abstract":"Deep learning has been widely used in medical image segmentation and other aspects. However, the performance of existing medical image segmentation models has been limited by the challenge of obtaining sufficient high-quality labeled data due to the prohibitive data annotation cost. To alleviate this limitation, we propose a new text-augmented medical image segmentation model LViT (Language meets Vision Transformer). In our LViT model, medical text annotation is incorporated to compensate for the quality deficiency in image data. In addition, the text information can guide to generate pseudo labels of improved quality in the semi-supervised learning. We also propose an Exponential Pseudo label Iteration mechanism (EPI) to help the Pixel-Level Attention Module (PLAM) preserve local image features in semi-supervised LViT setting. In our model, LV (Language-Vision) loss is designed to supervise the training of unlabeled images using text information directly. For evaluation, we construct three multimodal medical segmentation datasets (image + text) containing X-rays and CT images. Experimental results show that our proposed LViT has superior segmentation performance in both fully-supervised and semi-supervised setting. The code and datasets are available at https://github.com/HUANGLIZI/LViT.","PeriodicalId":13418,"journal":{"name":"IEEE Transactions on Medical Imaging","volume":null,"pages":null},"PeriodicalIF":8.9000,"publicationDate":"2022-06-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"22","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Transactions on Medical Imaging","FirstCategoryId":"5","ListUrlMain":"https://doi.org/10.48550/arXiv.2206.14718","RegionNum":1,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS","Score":null,"Total":0}
引用次数: 22
Abstract
Deep learning has been widely used in medical image segmentation and other aspects. However, the performance of existing medical image segmentation models has been limited by the challenge of obtaining sufficient high-quality labeled data due to the prohibitive data annotation cost. To alleviate this limitation, we propose a new text-augmented medical image segmentation model LViT (Language meets Vision Transformer). In our LViT model, medical text annotation is incorporated to compensate for the quality deficiency in image data. In addition, the text information can guide to generate pseudo labels of improved quality in the semi-supervised learning. We also propose an Exponential Pseudo label Iteration mechanism (EPI) to help the Pixel-Level Attention Module (PLAM) preserve local image features in semi-supervised LViT setting. In our model, LV (Language-Vision) loss is designed to supervise the training of unlabeled images using text information directly. For evaluation, we construct three multimodal medical segmentation datasets (image + text) containing X-rays and CT images. Experimental results show that our proposed LViT has superior segmentation performance in both fully-supervised and semi-supervised setting. The code and datasets are available at https://github.com/HUANGLIZI/LViT.
期刊介绍:
The IEEE Transactions on Medical Imaging (T-MI) is a journal that welcomes the submission of manuscripts focusing on various aspects of medical imaging. The journal encourages the exploration of body structure, morphology, and function through different imaging techniques, including ultrasound, X-rays, magnetic resonance, radionuclides, microwaves, and optical methods. It also promotes contributions related to cell and molecular imaging, as well as all forms of microscopy.
T-MI publishes original research papers that cover a wide range of topics, including but not limited to novel acquisition techniques, medical image processing and analysis, visualization and performance, pattern recognition, machine learning, and other related methods. The journal particularly encourages highly technical studies that offer new perspectives. By emphasizing the unification of medicine, biology, and imaging, T-MI seeks to bridge the gap between instrumentation, hardware, software, mathematics, physics, biology, and medicine by introducing new analysis methods.
While the journal welcomes strong application papers that describe novel methods, it directs papers that focus solely on important applications using medically adopted or well-established methods without significant innovation in methodology to other journals. T-MI is indexed in Pubmed® and Medline®, which are products of the United States National Library of Medicine.