{"title":"融合残差注意密集双u网络视网膜血管分割算法","authors":"Chunhui Zhu, Xiaowei Niu, Lu Zuo, Ziwei Liu","doi":"10.1109/ISBP57705.2023.10061315","DOIUrl":null,"url":null,"abstract":"In view of the slightly low accuracy of the existing overall segmentation algorithm of retinal vessels, a fused residual attention dense convolution double-U network, LSCD-UNet (Laddernet network based on scSE-Residual and CBAM and dense cavity convolution) was proposed.. Laddernet, a form of UNet, was introduced into the network. On this basis, the residual module with shared weights was upgraded and replaced with the scSE-Residual module of shared weights to facilitate feature enhancement extraction. A multi-module was introduced at the bottom of this network, consisting of the Convolutional Attention Mechanism Module (CBAM) and Dense Cavity Convolution Module (DAC) in series to expand the field of view and capture more subtle vascular features. Hybrid loss function was used to accelerate the network convergence. The LSCD-UNet algorithm was validated on the public datasets, DRIVE and STARE,. The results showed that the LSCD-UNet algorithm had an accuracy of 97.35% and 97.28%, a sensitivity of 81.80% and 86.23%, an AUC of 98.82% and 99.02%, and an F1 value of 84.89% and 84.97%, respectively, outperforming UNet and Laddernet and other retinal vessel segmentation algorithms.","PeriodicalId":309634,"journal":{"name":"2023 International Conference on Intelligent Supercomputing and BioPharma (ISBP)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2023-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Fused Residual Attention Dense Double-U Network Retinal Vessel Segmentation Algorithm\",\"authors\":\"Chunhui Zhu, Xiaowei Niu, Lu Zuo, Ziwei Liu\",\"doi\":\"10.1109/ISBP57705.2023.10061315\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"In view of the slightly low accuracy of the existing overall segmentation algorithm of retinal vessels, a fused residual attention dense convolution double-U network, LSCD-UNet (Laddernet network based on scSE-Residual and CBAM and dense cavity convolution) was proposed.. Laddernet, a form of UNet, was introduced into the network. On this basis, the residual module with shared weights was upgraded and replaced with the scSE-Residual module of shared weights to facilitate feature enhancement extraction. A multi-module was introduced at the bottom of this network, consisting of the Convolutional Attention Mechanism Module (CBAM) and Dense Cavity Convolution Module (DAC) in series to expand the field of view and capture more subtle vascular features. Hybrid loss function was used to accelerate the network convergence. The LSCD-UNet algorithm was validated on the public datasets, DRIVE and STARE,. The results showed that the LSCD-UNet algorithm had an accuracy of 97.35% and 97.28%, a sensitivity of 81.80% and 86.23%, an AUC of 98.82% and 99.02%, and an F1 value of 84.89% and 84.97%, respectively, outperforming UNet and Laddernet and other retinal vessel segmentation algorithms.\",\"PeriodicalId\":309634,\"journal\":{\"name\":\"2023 International Conference on Intelligent Supercomputing and BioPharma (ISBP)\",\"volume\":\"57 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2023-01-06\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2023 International Conference on Intelligent Supercomputing and BioPharma (ISBP)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ISBP57705.2023.10061315\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2023 International Conference on Intelligent Supercomputing and BioPharma (ISBP)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ISBP57705.2023.10061315","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
In view of the slightly low accuracy of the existing overall segmentation algorithm of retinal vessels, a fused residual attention dense convolution double-U network, LSCD-UNet (Laddernet network based on scSE-Residual and CBAM and dense cavity convolution) was proposed.. Laddernet, a form of UNet, was introduced into the network. On this basis, the residual module with shared weights was upgraded and replaced with the scSE-Residual module of shared weights to facilitate feature enhancement extraction. A multi-module was introduced at the bottom of this network, consisting of the Convolutional Attention Mechanism Module (CBAM) and Dense Cavity Convolution Module (DAC) in series to expand the field of view and capture more subtle vascular features. Hybrid loss function was used to accelerate the network convergence. The LSCD-UNet algorithm was validated on the public datasets, DRIVE and STARE,. The results showed that the LSCD-UNet algorithm had an accuracy of 97.35% and 97.28%, a sensitivity of 81.80% and 86.23%, an AUC of 98.82% and 99.02%, and an F1 value of 84.89% and 84.97%, respectively, outperforming UNet and Laddernet and other retinal vessel segmentation algorithms.