GAO Yuman,GAO Lin,HE Jin.Colorectal Polyp Segmentation Algorithm Using Dual-Hard-Net Network[J].Journal of Chengdu University of Information Technology,2023,38(03):291-297.[doi:10.16836/j.cnki.jcuit.2023.03.007]
基于Dual-Hard-Net网络的结直肠息肉分割算法
- Title:
- Colorectal Polyp Segmentation Algorithm Using Dual-Hard-Net Network
- 文章编号:
- 2096-1618(2023)03-0291-07
- Keywords:
- medical image segmentation; colorectal polyps; encoding/decoding structure; multi-scale information; attention mechanism
- 分类号:
- TP391
- 文献标志码:
- A
- 摘要:
- 由于结直肠息肉在形状,色泽和质地等方面各异,且息肉与其周边正常区域的边界分辨模糊等特点,导致息肉医学图像分割存在较大挑战。为提高结直肠息肉的分割准确率,提出一种改进的 Dual-Hard-Net 网络分割算法。算法采用经典的编解码结构,以Hardnet为共享编码器提取必要的多尺度特征,运用两对并行排列的解码器对提取的特征加以充分利用,并引入改进的注意力机制和残差模块减少计算量,解决梯度消失等问题。分别在两个不同的数据集上进行比较和实验,dice及mIoU系数取得了0.895和0.859的准确率。与当前的多种主流算法相比,算法具有更高的分割准确性和精确度。
- Abstract:
- Colorectal polyps are different in size, color and texture, and the boundaries between the polyps and the surrounding mucosa are not clear, leading to significant challenges in polyp segmentation. In order to improve the segmentation accuracy of colorectal polyps, this paper proposes an improved Dual-Hard-Net network segmentation algorithm which applies the classical encoding/decoding structure. In this structure, we use the Hardnet as the shared encoder to extract multi-scale information, and use two Juxtaposed decoders to further process the extracted features. To enhance the semantic information of the feature map, reduce the amount of computation and settle the Vanishing Gradient Problem, the advanced attention mechanism and residual blocks are introduced for original network. The results of experiments on two medical image data sets show that the Dice coefficient and mIoU are increased, which are 0.895 and 0.859 respectively compared with the original network. The experimental results compared with the main classic algorithms verify the feasibility and high accuracy of this method.
参考文献/References:
[1] Long J,Shelhamer E,Darrell T.Fully Convolutional Networks for Semantic Segmentation[J/OL].http://arxiv.org/abs/1411.4038.2015.
[2] RONNEBERGER O,FISCHER P,BROX T.U-Net:Convolutional Networks for Biomedical Image Segmentation[J/OL].http://arxiv.org/abs/1505.04597.2015.
[3] Zhou Zongwei,Md Mahfuzur Rahman Siddiquee,Nima Tajbakhsh,et al.UNet++:A Nested U-Net Architecture for Medical Image Segmentation[J/OL].http://arxiv.org/abs/1807.10165.2018:3-11.
[4] ISENSEE F,PETERSEN J,KLEIN A,et al.2018a.nnU-Net:Self-adapting Framework for U-Net-Based Medical Image Segmentation[J/OL].http://arxiv.org/abs/1809.10486.2018:778-789.
[5] JHA D,RIEGLER M A,JOHANSEN D,et al.DoubleU-Net:A Deep Convolutional Neural Network for Medical Image Segmentation[J/OL].http://arxiv.org/abs/2006.04868.2020.
[6] Guan S,Khan A,Chitnis P V,et al.Fully Dense UNet for 2D Sparse Photoacoustic Tomography Reconstruction[J].arXiv:1808.10848 IEEE.2018:7-23.
[7] IBTEHAZ N,RAHMAN M S.MultiResUNet:Rethinking the U-Net Architecture for Multimodal Biomedical Image Segmentation[J/OL].Neural Networks,2020,121:74-87.
[8] HUANG H,LIN L,TONG R,et al.UNet 3+:A Full-Scale Connected UNet for Medical Image Segmentation[C/OL].ICASSP 2020-2020 IEEE International Conference on Acoustics,Speech and Signal Processing(ICASSP).Barcelona,Spain:IEEE:2021:1055-1059.
[9] BIPARVA M,TSOTSOS J.STNet: Selective Tuning of Convolutional Networks for Object[J/OL].http://arxiv.org/abs/1708.06418.2022-07-25.
[10] OKTAY O,SCHLEMPER J,FOLGOC L L,et al.2018. Attention U-Net: Learning Where to Look for the Pancreas[J/OL].http://arxiv.org/abs/1804.03999.2018.
[11] ROY A G,NAVAB N,WACHINGER C.Concurrent Spatial and Channel Squeeze & Excitation in Fully Convolutional Networks[J/OL].http://arxiv.org/abs/1803.02579.2018.
[12] LI X,WANG W,HU X,et al.Selective Kernel Networks[M/OL].http://arxiv.org/abs/1903.06586,2022-07-25.
[13] FU J,LIU J,TIAN H,et al.Dual Attention Network for Scene Segmentation[J/OL].http://arxiv.org/abs/1809.02983.2019.
[14] TOMAR N K,JHA D,ALI S,et al.DDANet:Dual Decoder Attention Network for Automatic Polyp Segmentation[J/OL].http://arxiv.org/abs/2012.15245.2020.
[15] CHAO P,KAO C Y,RUAN Y S,et al.HarDNet: A Low Memory Traffic Network[M/OL].http://arxiv.org/abs/1909.00948,2019.2022-07-25.
[16] Lu Z,Pu H,Wang F,et al.The expressive power of neural networks:A view from the width[C].Advances in Neural Information Processing Systems.2017:6231-6239.
[17] YU F,KOLTUN V,Multi-Scale Context Aggregation by Dilated Convolutions[M/OL].http://arxiv.org/abs/1511.07122.2022-07-29.
[18] Prashant Brahmbhatt,Siddhi Nath Rajan.Skin Lesion Segmentation using SegNet with Binary Cross-Entropy[J].International Conference on Artificial Intelligence and Speech Technology(AIST2019),2019:223-245.
[19] JHA D,SMEDSRUD P H,RIEGLER M A,et al.2019.ResUNet++:An Advanced Architecture for Medical Image Segmentation[J/OL].http://arxiv.org/abs/1911.07067.2019.
[20] PEREZ L,WANG J.The Effectiveness of Data Augmentation in Image Classification using Deep Learning[M/OL].http://arxiv.org/abs/1712.04621.2023-03-20.2017.
[21] 王亚刚,郗怡媛,潘晓英.改进DeepLabv3+网络的肠道息肉分割方法[J].计算机科学与探索,2020,14(7):1243-1250.
备注/Memo
收稿日期:2022-08-07
基金项目:四川省科技计划资助项目(2020YFS0316)