并行多尺度特征融合的热层析图像分割算法

doi:10.3778/j.issn.1002-8331.2103-0079

摘要/Abstract

摘要： 热层析技术应用于医学乳腺肿瘤检测具有重要的应用前景，但是医生在临床应用过程中，通过热层析图像诊断容易出现主观差异性误诊现象，为此，提出了一种智能化分割算法用于辅助诊断。然而，医学热层析图像由于目前数据量匮乏，且病灶区域占比小，依靠经典的分割模型，如FCN、U-Net容易出现分割不连续，边界细节分割不精细等问题。设计了一种基于并行多尺度特征融合的语义分割模型，模型通过并行的多分辨率特征子网之间反复的信息交换，在保证分割结果语义信息准确之外，还能有效地抓取病灶区域的细节特征。该方法在热层析医学图像数据集上取得了0.635?7的均交并比，相较于经典的U-Net分割网络，取得了5.14个百分点的提升，在肿块和血管等小目标区域的细节分割上有着更出色的表现。实验结果表明，该算法对热层析临床用于乳腺癌的辅助诊断具有现实意义。

关键词: 热层析图像, 小目标, 多尺度特征融合, 图像分割, 并行架构

Abstract: The application of thermal tomography in the detection of medical breast tumors has an important application prospect, but in the process of clinical application, doctors are prone to misdiagnosis through thermal tomography images. Therefore, an intelligent segmentation algorithm is proposed to assist diagnosis. However, due to the lack of data and small proportion of lesion area, medical thermal tomography images rely on classical segmentation models, such as FCN and U-Net, which are prone to problems such as discontinuous segmentation and unrefined segmentation of boundary details. Therefore, a semantic segmentation model based on parallel multi-scale feature fusion is designed. Through repeated information exchange among parallel multi-resolution feature subnets, the model can not only ensure the accuracy of semantic information of segmentation results, but also effectively capture the detail features of the focal area. The proposed method achieves the mean intersection over union of 0.635?7 on the dataset of thermal tomography medical images, an improvement of 5.14?percentage points compared with the classic U-NET segmentation network, and has a better performance in the segmentation of small target areas such as lumps and blood vessels. The experimental results show that the proposed algorithm has practical significance for the clinical application of thermal tomography in the auxiliary diagnosis of breast cancer.

Key words: thermal tomography, small targets, multi-scale feature fusion, image segmentation, parallel architecture

胡长康, 李凯扬. 并行多尺度特征融合的热层析图像分割算法[J]. 计算机工程与应用, 2022, 58(19): 233-241.

HU Changkang, LI Kaiyang. Parallel Multi-Scale Feature Fusion Algorithm for Thermal Tomography Image Segmentation[J]. Computer Engineering and Applications, 2022, 58(19): 233-241.

参考文献

[1] 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，2015：234-241.
[2] CHEN L C.Semantic image segmentation with deep convolutional nets and fully connected CRFS[J].arXiv：1412.7062，2014.
[3] CHEN L C.Deeplab：Semantic image segmentation with deep convolutional nets，atrous convolution，and fully connected CRFS[J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2017，40：834-848.
[4] CHEN L C.Rethinking atrous convolution for semantic image segmentation[J].arXiv：1706.05587，2017.
[5] DONG R S，PAN X Q，LI F Y.DenseU-net-based semantic segmentation of small objects in urban remote sensing images[J].IEEE Access，2019，7：65347-65356.
[6] ZHOU Z W.U-Net++：A nested U-Net architecture for medical image segmentation[C]//Proceedings of International Workshop on Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support，2018：3-11.
[7] HUANG H，LIN L，TONG R，et al.U-Net 3+：A full-scale connected unet for medical image segmentation[C]//Proceedings of IEEE International Conference on Acoustics，Speech and Signal Processing（ICASSP），2020：1055-1059.
[8] 曹源，王妍，王文强，等.基于委员会查询和自步多样性学习的医学图像分割[J].西北大学学报（自然科学版），2020，50（2）：285-294.
CAO Y，WANG Y，WANG W Q，et al.Medical image segmentation based on committee query and self-stepping multiplicity learning[J].Journal of Northwest University（Natural Science Edition），2020，50（2）：285-294.
[9] QIAN B X，XIAO Z Y，SONG W.An improved convolutional neural network for lung image segmentation[J].Computer Science and Exploration，2020，14（8）：1358-1367.
[10] WANG X P，WANG Q S，JIAO J J，et al.Fast adaptive non-local spatial weighting and membership connecting fuzzy C-means noise image segmentation algorithm[J].Journal of Electronics and Information Technology，2021，43（1）：171-178.
[11] ZHAO Z Y，ZENG Z，XU K X，et al.DSAL：Deeply supervised active learning from strong and weak labelers for biomedical image segmentation[J].arXiv：2011.09057，2021.
[12] PANG S，PANG C，ZHAO L，et al.SpineParseNet：Spine parsing for volumetric MR image by a two-stage segmentation framework with semantic image representation[J].IEEE Transactions on Medical Imaging，2021，40（1）：262-273.
[13] SUN K.Deep high-resolution representation learning for human pose estimation[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition，2019：5693-5703.
[14] SZEGEDY C.Going deeper with convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2015：1-9.
[15] IOFFE S，SZEGEDY C.Batch normalization：Accelerating deep network training by reducing internal covariate shift[C]//Prceedings of International Conference on Machine Learning，2015：448-456.
[16] HOWARD A G.Mobilenets：Efficient convolutional neural networks for mobile vision applications[J].arXiv：1704. 04861，2017.
[17] HE K M.Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2016：770-778.
[18] KAMPFFMEYER M，SALBERG A，JENSSEN R.Semantic segmentation of small objects and modeling of uncertainty in urban remote sensing images using deep convolutional neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops，2016：1-9.
[19] LI B.Study on the application value of thermal tomography in breast examination[J].Radiology in Practice，2014（2）.
[20] LI J L.Temperature distribution in the diagnosis of breast disease by thermotomography[Z].2014.
[21] KINGMA D，BA J.ADAM：A method for stochastic optimization[J].arXiv：1412.6980，2014.