基于改进U-Net的人脑黑质致密部分割

doi:10.19678/j.issn.1000-3428.0063273

摘要/Abstract

摘要： 人脑黑质致密部分割能够为帕金森病的诊断提供一定依据。黑质致密部在人脑核磁共振成像中像素占比低、类间差异小，为提高计算机辅助诊断系统对人脑黑质致密部的分割精度，提出一种基于改进U形神经网络（U-Net）的人脑黑质致密部分割方法。为了提取更多有效的多尺度图像语义特征，结合U-Net的跨连接结构并采用多头注意力机制，同时融合基于Transformer编码器的高维语义编码模块以提取高维语义特征，避免浅层噪声对特征造成的影响。建立多任务模型并设计基于二维高斯核权重掩膜的损失函数，解决神经网络分割模型因多次下采样造成的不连续分割误差问题。构建包括140个帕金森病患者以及48个健康对照者的高精度核磁共振脑成像数据集进行实验，结果表明，相较常用的医疗影像分割方法R2U-Net、HANet等，该方法的多任务分割效果取得明显提升，戴斯相关系数和AUC指标分别达到0.869 1和0.943 9，消融实验结果也验证了改进编码器和改进损失这2个模块的有效性。

关键词: 图像分割, 帕金森病, 黑质致密部, U形神经网络, Transformer模块, 多任务学习

Abstract: The segmentation of brain Substantia Nigra pars compacta(SNpc) can provide the basis for Parkinson's disease diagnosis.SNpc is a low-pixel-ratio object in Magnetic Resonance Imaging(MRI) of the human brain, and the pixels with different labels are similar.This study proposes a method based on an improved U-Shape Neural Network(U-Net) to improve the segmentation accuracy of a computer-aided diagnosis system for brain SNpc.The cross-connection structure of U-Net is combined with the multihead attention mechanism to extract more effective multiscale image semantic features.The high-dimensional semantic encoding module based on a Transformer encoder is fused to extract high-dimensional semantic features and eliminate the effect of shallow noise on features.The multitask model is established, and the loss function based on the two-dimensional Gaussian kernel weight mask is designed to solve the discontinuous segmentation error caused by multiple down-samples of the neural network segmentation model.A high-precision MRI brain imaging dataset comprising 140 Parkinson's patients and 48 healthy controls was constructed for experiments.The results show that compared with the widely used medical image segmentation methods, such as R2U-Net and HANet, the multitask segmentation effect of this method improved significantly.The Dais phase relationship number and the Area Under the Curve(AUC) index reached 0.869 1 and 0.943 9, respectively.Furthermore, the results of the ablation experiment verify the effectiveness of the improved encoder and the improved loss modules.

Key words: image segmentation, Parkinson's disease, Substantia Nigra pars compacta(SNpc), U-Shape Neural Network(U-Net), Transformer module, multi-task learning

中图分类号:

TP391

曹加旺, 田维维, 刘学玲, 李郁欣, 冯瑞. 基于改进U-Net的人脑黑质致密部分割[J]. 计算机工程, 2022, 48(11): 14-21,29.

CAO Jiawang, TIAN Weiwei, LIU Xueling, LI Yuxin, FENG Rui. Segmentation of Brain Substantia Nigra Pars Compacta Based on Improved U-Net[J]. Computer Engineering, 2022, 48(11): 14-21,29.

http://www.ecice06.com/CN/Y2022/V48/I11/14

图/表 13

20230203202055

20230203202058

20230203202102

20230203202105

20230203202109

20230203202112

20230203202115

20230203202119

20230203202122

20230203202126

20230203202129

20230203202133

20230203202136

参考文献

[1] WANG N, LIU X L, LI L, et al.Screening for early-stage Parkinson's disease:swallow tail sign on MRI susceptibility map-weighted images compared with PET[J].Journal of Magnetic Resonance Imaging:JMRI, 2021, 53(3):722-730.
[2] 成涛, 黄小华, 匡静, 等.SBM在中枢神经系统中的研究进展[J].国际医学放射学杂志, 2020, 43(1):35-40. CHENG T, HUANG X H, KUANG J, et al.Research progress of surface-based morphometry in the central nervous system[J].International Journal of Medical Radiology, 2020, 43(1):35-40.(in Chinese)
[3] LIU X, WANG N, CHEN C, et al.Swallow tail sign on Susceptibility Map-Weighted Imaging(SMWI) for disease diagnosing and severity evaluating in Parkinsonism[J].Acta Radiologica, 2021, 62(2):234-242.
[4] LONG J, SHELHAMER E, DARRELL T.Fully convolutional networks for semantic segmentation[C]//Proceedings of 2015 IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2015:3431-3440.
[5] RONNEBERGER O, FISCHER P, BROX T.U-Net:convolutional networks for biomedical image segmentation[C]//Proceedings of International Conference on Medical Image Computing and Computer-Assisted Intervention.Berlin, Germany:Springer, 2015:234-241.
[6] LANGLEY J, HUDDLESTON D E, LIU C J, et al.Reproducibility of locus coeruleus and substantia nigra imaging with neuromelanin sensitive MRI[J].Magnetic Resonance Materials in Physics, Biology and Medicine, 2017, 30(2):121-125.
[7] TAKAHASHI H, WATANABE Y, TANAKA H, et al.Comprehensive MRI quantification of the substantia nigra pars compacta in Parkinson's disease[J].European Journal of Radiology, 2018, 109:48-56.
[8] TAKAHASHI H, WATANABE Y, TANAKA H, et al.Quantifying the severity of Parkinson disease by use of dopaminergic neuroimaging[J].AJR American Journal of Roentgenology, 2019, 213(1):163-168.
[9] ALOM M Z, HASAN M, YAKOPCIC C, et al.Recurrent residual convolutional neural network based on U-Net (R2U-Net) for medical image segmentation[EB/OL].[2021-10-05].https://arxiv.org/abs/1802.06955.
[10] OKTAY O, SCHLEMPER J, FOLGOC L L, et al.Attention U-Net:learning where to look for the pancreas[EB/OL].[2021-10-05].https://arxiv.org/abs/1804.03999.
[11] MATEEN M, WEN J H, NASRULLAH N, et al.Exudate detection for diabetic retinopathy using pretrained convolutional neural networks[J].Complexity, 2020, 15:1-11.
[12] 亢洁, 丁菊敏, 万永, 等.基于分水岭修正与U-Net的肝脏图像分割算法[J].计算机工程, 2020, 46(1):255-261, 270. KANG J, DING J M, WAN Y, et al.Liver image segmentation algorithm based on watershed correction and U-Net[J].Computer Engineering, 2020, 46(1):255-261, 270.(in Chinese)
[13] TAKIKAWA T, ACUNA D, JAMPANI V, et al.Gated-SCNN:gated shape CNNs for semantic segmentation[C]//Proceedings of IEEE/CVF International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2020:5228-5237.
[14] 梅旭璋, 江红, 孙军.基于密集注意力网络的视网膜血管图像分割[J].计算机工程, 2020, 46(3):267-272, 279. MEI X Z, JIANG H, SUN J.Retinal vessel image segmentation based on dense attention network[J].Computer Engineering, 2020, 46(3):267-272, 279.(in Chinese)
[15] 郝华颖, 赵昆, 苏攀, 等.一种基于改进ResU-Net的角膜神经分割算法[J].计算机工程, 2021, 47(1):217-223. HAO H Y, ZHAO K, SU P, et al.A corneal nerve segmentation algorithm based on improved ResU-Net[J].Computer Engineering, 2021, 47(1):217-223.(in Chinese)
[16] 王磐, 强彦, 杨晓棠, 等.基于双注意力3D-UNet的肺结节分割网络模型[J].计算机工程, 2021, 47(2):307-313. WANG P, QIANG Y, YANG X T, et al.Network model for lung nodule segmentation based on double attention 3D-UNet[J].Computer Engineering, 2021, 47(2):307-313.(in Chinese)
[17] DING F, YANG G, WU J, et al.High-order attention networks for medical image segmentation[C]//Proceedings of International Conference on Medical Image Computing and Computer-Assisted Intervention.Berlin, Germany:Springer, 2020:253-262.
[18] VASWANI A, SHAZEER N, PARMAR N, et al.Attention is all you need[EB/OL].[2021-10-05].https://arxiv.org/abs/1706.03762.
[19] DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al.An image is worth 16×16 words:transformers for image recognition at scale[EB/OL].[2021-10-05].https://arxiv.org/pdf/2010.11929.pdf.
[20] LIU Z, LIN Y T, CAO Y, et al.Swin Transformer:hierarchical vision Transformer using shifted windows[C]//Proceedings of 2021 IEEE/CVF International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2021:9992-10002.
[21] DAI X Y, CHEN Y P, YANG J W, et al.Dynamic DETR:end-to-end object detection with dynamic attention[C]//Proceedings of IEEE/CVF International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2021:2968-2977.
[22] SUN Z Q, CAO S C, YANG Y M, et al.Rethinking Transformer-based set prediction for object detection[C]//Proceedings of IEEE/CVF International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2021:3591-3600.
[23] WANG W, CHEN C, DING M, et al.TransBTS:multimodal brain tumor segmentation using Transformer[C]//Proceedings of International Conference on Medical Image Computing and Computer-Assisted Intervention.Berlin, Germany:Springer, 2021:109-119.
[24] ZHENG S X, LU J C, ZHAO H S, et al.Rethinking semantic segmentation from a sequence-to-sequence perspective with transformers[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2021:6877-6886.
[25] CAO J W, JIANG L L, HOU J L, et al.Exploiting deep cross-slice features from CT images for multi-class pneumonia classification[C]//Proceedings of IEEE International Conference on Image Processing.Washington D.C., USA:IEEE Press, 2021:205-209.
[26] YANG S Y, JIANG L Q, CAO Z Q, et al.Deep learning for detecting corona virus disease 2019(COVID-19) on high-resolution computed tomography:a pilot study[J].Annals of Translational Medicine, 2020, 8(7):450.
[27] LIU X L, YANG L Q, LIU F T, et al.Short-echo-time magnitude image derived from quantitative susceptibility mapping could resemble neuromelanin-sensitive MRI image in substantia nigra[J].BMC Neurology, 2020, 20(1):262.

选择文件类型/文献管理软件名称

选择包含的内容