文献类型：期刊文献

英文题名：Lesion detection of chest X-Ray based on scalable attention residual CNN

作者：Lin, Cong[1,3];Huang, Yiquan[3];Wang, Wenling[1];Feng, Siling[1];Huang, Mengxing[1,2]

机构：[1]Hainan Univ, Coll Informat & Commun Engn, Haikou 570228, Peoples R China;[2]Hainan Univ, State Key Lab Marine Resource Utilizat South China, Haikou 570228, Peoples R China;[3]Guangdong Ocean Univ, Coll Elect & Informat Engn, Zhanjiang 524088, Peoples R China

年份：2023

卷号：20

期号：2

起止页码：1730

外文期刊名：MATHEMATICAL BIOSCIENCES AND ENGINEERING

收录：SCI-EXPANDED(收录号：WOS:000911277500003)、、EI(收录号：20224713151973)、Scopus(收录号：2-s2.0-85142155738)、WOS

基金：This work is supported by the National Key R&D Program of China (No. 2021ZD0111000) , Hainan Provincial Natural Science Foundation of China (621MS019) , Major Science and Technology Project of Haikou (Grant: 2020-009) , Innovative Research Project of Postgraduates in Hainan Province (Qhyb2021-10) , Guangdong University Student Science and Technology Innovation Cultivation Special Fund Support Project (pdjh2023 a0243) and Key R&D Project of Hainan province (Grant: ZDYF2021SHFZ243) .

语种：英文

外文关键词：chest X-ray; object detection; deep learning; attention mechanism; disease recognition

外文摘要：Most of the research on disease recognition in chest X-rays is limited to segmentation and classification, but the problem of inaccurate recognition in edges and small parts makes doctors spend more time making judgments. In this paper, we propose a lesion detection method based on a scalable attention residual CNN (SAR-CNN), which uses target detection to identify and locate diseases in chest X-rays and greatly improves work efficiency. We designed a multi-convolution feature fusion block (MFFB), tree-structured aggregation module (TSAM), and scalable channel and spatial attention (SCSA), which can effectively alleviate the difficulties in chest X-ray recognition caused by single resolution, weak communication of features of different layers, and lack of attention fusion, respectively. These three modules are embeddable and can be easily combined with other networks. Through a large number of experiments on the largest public lung chest radiograph detection dataset, VinDr-CXR, the mean average precision (mAP) of the proposed method was improved from 12.83% to 15.75% in the case of the PASCAL VOC 2010 standard, with IoU >0.4, which exceeds the existing mainstream deep learning model. In addition, the proposed model has a lower complexity and faster reasoning speed, which is conducive to the implementation of computer-aided systems and provides referential solutions for relevant communities.