基于Dense-UNet++的关节滑膜磁共振图像分割
收稿日期: 2021-04-06
网络出版日期: 2021-07-01
Magnetic Resonance Images Segmentation of Synovium Based on Dense-UNet++
Received date: 2021-04-06
Online published: 2021-07-01
为解决以往基于深度学习的滑膜磁共振图像分割模型存在的分割精度较低、鲁棒性较差、训练耗时等问题，本文提出了一种基于Dense-UNet++网络的新模型，将DenseNet模块插入UNet++网络中，并使用Swish激活函数进行训练.利用1 036张滑膜磁共振图像数据增广后的14 512张滑膜图像对模型进行训练，并利用68张图像进行测试.结果显示，模型的平均DSC系数为0.819 9，交叉联合度量（IOU）为0.927 9.相较于UNet、ResUNet和VGG-UNet++网络结构，DSC系数和IOU均有提升，DSC振荡系数降低.另外在应用于相同滑膜图像数据集和使用相同的网络结构时，Swish函数相比ReLu函数有助于提升分割精度.实验结果表明，本文提出的算法对于滑膜磁共振图像的病灶区域的分割有较好的效果，能够辅助医生对病情做出判断.
关键词： 磁共振成像（MRI）; 深度学习; 医学图像分割; 滑膜炎
王振宇 , 王颖珊 , 毛瑾玲 , 马伟伟 , 路青 , 石洁 , 汪红志 . 基于Dense-UNet++的关节滑膜磁共振图像分割[J]. 波谱学杂志, 2022 , 39(2) : 208 -219 . DOI: 10.11938/cjmr20212905
To further improve the segmentation accuracy, robustness, and training efficiency of existing articular synovium segmentation algorithms, a new deep learning network based on Dense-UNet++ was proposed. First, we inserted the DenseNet module into the UNet++ network, then applied the Swish activation function to train the model. The network was trained through 14 512 synovial images augmented from 1 036 synovial images, and tested through 68 images. The average accuracy of the model reached 0.819 9 for dice similarity coefficient (DSC), and 0.927 9 for intersection over union (IOU) index. Compared with UNet, ResUNet and VGG-UNet++, DSC coefficient and IOU index were improved, and DSC oscillation coefficient reduced. In addition, when applied in the same synovial image set and using the same network structure, the Swish function can help improve the accuracy of segmentation compared with the ReLu function. The experimental results show that the proposed algorithm performs better in segmenting articular synovium and may assist doctors in disease diagnosis.
| 1 | KONG X Y , WANG R Z . Artificial intelligence and its application in medical field[J]. Journal of Medical Intelligence, 2016, 37 (11): 2- 5. |
| 1 | 孔祥溢, 王任直. 人工智能及在医疗领域的应用[J]. 医学信息学杂志, 2016, 37 (11): 2- 5. |
| 2 | 张明月. 基于深度学习的图像分割研究[D]. 长春: 吉林大学, 2017. |
| 3 | ZHOU H Y , LIU P C . Clinical value and research progress of MR in wrist facet joint lesions[J]. China Medical Engineering, 2010, 18 (3): 172- 175. |
| 3 | 周海燕, 刘鹏程. MR在手腕部小关节病变的临床价值及研究进展[J]. 中国医学工程, 2010, 18 (03): 172- 175. |
| 4 | MA J , CHEN M , LI H , et al. Clinical observation of traditional Chinese medicine combined with antirheumatic drug treatment of rheumatoid arthritis in active stage[J]. Chinese Journal of Experimental Traditional Medical Formulae, 2014, 20 (5): 192- 196. |
| 4 | 马进, 陈岷, 李获, 等. 中药联合抗风湿药治疗类风湿性关节炎活动期的临床观察[J]. 中国实验方剂学杂志, 2014, 20 (05): 192- 196. |
| 5 | 马强. 类风湿关节炎的腕关节MRI及临床应用研究[D]. 太原: 山西医科大学, 2003. |
| 6 | WEI X N , XING J Q , WANG Z Y , et al. Magnetic resonance image segmentation of articular synovium based on improved U-Net[J]. Journal of Computer Applications, 2020, 40 (11): 3340- 3345. |
| 6 | 魏小娜, 邢嘉祺, 王振宇, 等. 基于改进U-Net的关节滑膜磁共振图像的分割[J]. 计算机应用, 2020, 40 (11): 3340- 3345. |
| 7 | LONG Z , ZHANG X , LI C , et al. Segmentation and classification of knee joint ultrasonic image via deep learning[J]. Applied Soft Computing, 2020, 97 (9): 106765. |
| 8 | LING Z, SONKA M, LE L, et al. Combining fully convolutional networks and graph-based approach for automated segmentation of cervical cell nuclei[C]. IEEE ISBI (Oral). IEEE, 2017. |
| 9 | PHAM D L , XU C Y , PRINCE J L . Current methods in medical image segmentation[J]. Annu Rev Biomed Eng, 2000, 2, 315- 337. |
| 10 | XIAO L , LOU Y K , ZHOU H Y . A U-net network-based rapid construction of knee models for specific absorption rate estimation[J]. Chinese J Magn Reson, 2020, 37 (2): 144- 151. |
| 10 | 肖亮, 娄煜堃, 周航宇. 用于SAR估计的基于U-Net网络的快速膝关节模型重建[J]. 波谱学杂志, 2020, 37 (2): 144- 151. |
| 11 | FARAHANI A, MOHSENI H. Medical image segmentation using customized U-Net with adaptive activation functions[J]. Neural Computing & Applications, doi: 10.1007/s00521-020-05396-317. |
| 12 | ZHOU Z, SIDDIQUEE M M R, TAJBAKHSH N, et al. UNet++: A nested U-Net architecture for medical image segmentation[C]. Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support: MICCAI 2018, Granada, Spain, 2018, 110453-11. |
| 13 | ZHAO S Y , WANG Y J . Classification of Alzheimer's disease patients based on magnetic resonance images and an improved UNet++ model[J]. Chinese J Magn Reson, 2020, 37 (3): 321- 331. |
| 13 | 赵尚义, 王远军. 基于磁共振图像和改进的UNet++模型区分阿尔茨海默症患者和健康人群[J]. 波谱学杂志, 2020, 37 (3): 321- 331. |
| 14 | ZHANG K , GUO Y R , WANG X S , et al. Multiple feature reweight densenet for image classification[J]. IEEE Access, 2019, 7, 9872- 9880. |
| 15 | ZENG M J , XIAO N F . Effective combination of densenet and bilSTM for keyword spotting[J]. IEEE Access, 2019, 7, 10767- 10775. |
| 16 | TAO Y , XU M , LU Z , et al. DenseNet-based depth-width double reinforced deep learning neural network for high-resolution remote sensing image per-pixel classification[J]. Remote Sensing, 2018, 10 (5) |
| 17 | CHU X X, ZHANG B, XU R J. Searching beyond mobileNetV3[C]. ICASSP, 2020. https://arxiv.org/pdf/1908.01314.pdf. |
| 18 | SOOMRO T A, AFIFI A J, JUNBIN G, et al. Strided U-net model: retinal vessels segmentation using dice loss[C]//2018 Digital Image Computing: Techniques and Applications (DICTA), 2018. |
| 19 | LEVINE R A . The art of data augmentation - Discussion[J]. Journal of Computational and Graphical Statistics, 2001, 10 (1): 51- 58. |
/
| 〈 |
|
〉 |