《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (8): 2712-2719.DOI: 10.11772/j.issn.1001-9081.2024071019

• 多媒体计算与计算机仿真 • 上一篇    

基于跨尺度注意力网络的胸部疾病分类方法

林进浩1, 罗川1(), 李天瑞2, 陈红梅2   

  1. 1.四川大学 计算机学院,成都 610065
    2.西南交通大学 计算机与人工智能学院,成都 611756
  • 收稿日期:2024-07-19 修回日期:2024-11-05 接受日期:2024-11-05 发布日期:2024-12-03 出版日期:2025-08-10
  • 通讯作者: 罗川
  • 作者简介:林进浩(1999—),男,广东阳江人,硕士研究生,主要研究方向:深度学习、医学图像处理
    李天瑞(1969—),男,福建莆田人,教授,博士,主要研究方向:人工智能、数据挖掘、知识发现
    陈红梅(1971—),女,四川成都人,教授,博士,主要研究方向:数据挖掘、粒计算。
  • 基金资助:
    国家自然科学基金资助项目(62476182);国家自然科学基金资助项目(62076171);国家自然科学基金资助项目(62376230);四川省自然科学基金资助项目(2022NSFSC0898)

Thoracic disease classification method based on cross-scale attention network

Jinhao LIN1, Chuan LUO1(), Tianrui LI2, Hongmei CHEN2   

  1. 1.College of Computer Science,Sichuan University,Chengdu Sichuan 610065,China
    2.School of Computing and Artificial Intelligence,Southwest Jiaotong University,Chengdu Sichuan 611756,China
  • Received:2024-07-19 Revised:2024-11-05 Accepted:2024-11-05 Online:2024-12-03 Published:2025-08-10
  • Contact: Chuan LUO
  • About author:LIN Jinhao, born in 1999, M. S. candidate. His research interests include deep learning, medical image processing.
    LI Tianrui, born in 1969, Ph. D., professor. His research interests include artificial intelligence, data mining, knowledge discovery.
    CHEN Hongmei, born in 1971, Ph. D., professor. Her research interests include data mining, granular computing.
  • Supported by:
    National Natural Science Foundation of China(62476182);Natural Science Foundation of Sichuan Province(2022NSFSC0898)

摘要:

从胸部X光片中自动识别胸部疾病是计算机辅助诊断的重要研究领域。然而,现有的许多胸部疾病分类方法在处理病变区域大小差异方面存在困难,并且无法准确识别和定位不同疾病的病变区域。针对上述问题,提出一种基于跨尺度注意力网络(CANet)的胸部疾病分类方法。该方法使用DenseNet-121作为特征提取网络,并集成自感知注意力(SAA)、向上聚焦注意力(UFA)和向下引导注意力(DGA)3个主要模块。SAA模块通过提取与胸部疾病相关的通道和异常区域信息,细化空间位置上的病理特征,并减少不相关区域的干扰。为了实现不同尺度空间上下文信息的跨尺度交互,使用UFA和DGA模块进行图像特征校准。此外,提出空间注意力金字塔池化(SAPP)模块用于融合不同特征图的多尺度特征,从而提高胸部疾病的检测性能。在ChestX-ray14和DR-Pneumonia数据集上的实验结果表明,所提方法的平均曲线下面积(AUC)值分别达到了83.4%和82.6%,优于DualCheXNet、A3Net和CheXGAT等方法。具体地,与CheXGAT方法相比,所提方法的平均AUC值分别提高了0.7和0.1个百分点。可见,所提方法可以识别胸部X光片中的重要信息,有效提高了胸部疾病分类的性能。

关键词: 胸部疾病分类, 胸部X光片, 注意力机制, 卷积神经网络, 特征融合

Abstract:

Automatic identification of thoracic diseases from chest X-rays is a significant area of research in computer-aided diagnosis. However, many existing methods for thoracic disease classification struggle to handle differences in lesion area sizes and often fail to identify and localize the lesion areas of different diseases accurately. To address the above problems, a thoracic disease classification method based on Cross-scale Attention Network (CANet) was proposed. In this method, DenseNet-121 was employed as the feature extraction network, and three main modules were integrated: Self Aware Attention (SAA), Upward Focus Attention (UFA), and Downward Guidance Attention (DGA) modules. In the SAA module, the spatial pathological features were refined and the interference from irrelevant areas was reduced by extracting channel and abnormal area information relevant to thoracic diseases. In order to achieve cross-scale interaction of spatial context information of different scales, image feature calibration was performed using the UFA and DGA modules. Additionally, the Spatial Attention Pyramid Pooling (SAPP) module was proposed to fuse multi-scale features from different feature maps, thereby enhancing the detection performance for thoracic diseases. Experimental results on ChestX-ray14 and DR-Pneumonia datasets show that the proposed method has the average Area Under Curve (AUC) values of 83.4% and 82.6%, respectively, outperforming DualCheXNet, A3Net, and CheXGAT methods. Specifically, compared with CheXGAT method, the proposed method improves the average AUC values by 0.7 and 0.1 percentage points, respectively. It can be seen that the proposed method identifies critical information in chest X-rays effectively, improving the performance of thoracic disease classification significantly.

Key words: thoracic disease classification, chest X-ray, attention mechanism, Convolutional Neural Network (CNN), feature fusion

中图分类号: