Segmentation of Breast Tumors Based on Fully Convolutional Network and Dynamic Contrast Enhanced Magnetic Resonance Image

doi:10.11938/cjmr20212921

Abstract

Abstract:

Accurate and reliable breast tumor segmentation is essential for the diagnosis, treatment and prognosis of breast cancer. To address the shortcomings of existing dynamic contrast enhanced-magnetic resonance imaging (DCE-MRI)-based breast tumor segmentation methods, which tend to miss small tumors, we proposed a more reliable and efficient segmentation method for breast tumors in DCE-MRI based on a fully convolutional network (FCN). Firstly, the breast DCE-MRI data was preprocessed, followed by intercepting the image blocks of 128*128, and dividing the dataset into two sub-datasets according to the number of pixels in the tumor region. Secondly, the whole set was used to train CBP5-Net to obtain a classification model. Then, two sub-datasets were used to train RAU-Net to get two segmentation models. Finally, the test set was entered into the network, and the network outputs were post-processed to obtain the final segmentation results. The Dice coefficient, sensitivity, specificity and intersection over union (IoU) index of the method proposed in this paper reached 0.938 8, 0.952 3, 0.998 5 and 0.876 8, respectively. It proves that the proposed method can be used to segment DCE-MRI breast tumors effectively and accurately.

Key words: fully convolutional network (FCN), image segmentation, breast tumor, dynamic contrast enhanced-magnetic resonance imaging (DCE-MRI)

CLC Number:

Yue QIU, Sheng-dong NIE, Long WEI. Segmentation of Breast Tumors Based on Fully Convolutional Network and Dynamic Contrast Enhanced Magnetic Resonance Image[J]. Chinese Journal of Magnetic Resonance, 2022, 39(2): 196-207.

Figures/Tables 11

Fig.1

Fig.2

Table 1

Fig.3

Table 2

Table 3

Fig.4

Table 4

Fig.5

Fig.6

Fig.7

References 36

1	SIEGEL R L , MILLE R K D , JEMAL A . Cancer statistics, 2018[J]. CA: A Cancer Journal for Clinicians, 2018, 68 (1): 7- 30. doi: 10.3322/caac.21442
2	FOROUZANFAR M H , FOREMAN K J , DELOSSANTOS A M , et al. Breast and cervical cancer in 187 countries betweem 1980 and 2010: a systematic analysis[J]. Lancet, 2011, 378 (9801): 1461- 1484. doi: 10.1016/S0140-6736(11)61351-2
3	WILD C P, WEIDERPASS E, STEWART B W. World cancer report: cancer research for cancer prevention[M]//Lyon: International Agency for Research on Cancer, 2020: 23-33.
4	FANG Z Y , QIU X G . Distribution and risk factors of breast cancer[J]. Chinese General Practice, 2002, 5 (6): 429- 430. doi: 10.3969/j.issn.1007-9572.2002.06.004
	方志沂, 邱晓光. 乳腺癌的分布规律及其危险因素[J]. 中国全科医学, 2002, 5 (6): 429- 430. doi: 10.3969/j.issn.1007-9572.2002.06.004
5	SMITH R A , COKKINDES V , BROOKS D , et al. Cancer screening in the United States, 2011[J]. CA-Cancer J Clin, 2011, 61 (1): 8- 30. doi: 10.3322/caac.20096
6	DIDKOWSKA J , WOJCIECHOWSKA U . WHO position paper on mammography screening[J]. Oncology in Clinical Practice, 2014,
7	HEYWANG S H , HAHN D , SCHMIDT H , et al. MR imaging of the breast using gadolinium-DTPA[J]. J Comput Assist Tomogr, 1986, 10 (2): 199- 204. doi: 10.1097/00004728-198603000-00005
8	MORRIS E A . Breast cancer imaging with MRI[J]. Radiol Clin N Am, 2002, 40 (3): 443- 466. doi: 10.1016/S0033-8389(01)00005-7
9	BHOOSHAN N , GIGER M L , JANSEN S A , et al. Cancerous breast lesions on dynamic contrast-enhanced MR images: computerized characterization for image-based prognostic markers[J]. Radiology, 2010, 254 (3): 680- 690. doi: 10.1148/radiol.09090838
10	杨骞. 基于DCE-MRI影像的乳腺癌早期诊断研究[D]. 杭州: 杭州电子科技大学, 2014.
11	EADIE L H , TAYLOR P , GIBSON A P . A systematic review of computer-assisted diagnosis in diagnostic cancer imaging[J]. Eur J Radiol, 2011, 81 (1): e70- e76.
12	REIG B , HEACOCK L , GERAS K J , et al. Machine learning in breast MRI[J]. J Magn Reson Imaging, 2020, 52 (4): 998- 1018. doi: 10.1002/jmri.26852
13	TAN Z , HONG R R , YE S Z . A method to extract the region of interest in dynamic contrast-enhanced MR imaging of breast cancer[J]. Journal of Fuzhou University (Natural Science Edition), 2016, 44 (1): 33- 39.
	檀哲, 洪容容, 叶少珍. 一种提取乳腺癌DCE-MRI感兴趣区域的分割方法[J]. 福州大学学报(自然科学版), 2016, 44 (1): 33- 39.
14	QIAO M Y , SUO S T , CHENG F , et al. Three-dimensional breast tumor segmentation on DCE-MRI with a multilabel attention-guided joint-phase-learning network[J]. Comput Med Imag Grap, 2021, 90, 101909. doi: 10.1016/j.compmedimag.2021.101909
15	LIU T T , XU H Y , JIN W , et al. Medical image segmentation based on a hybrid region-based active contour model[J]. Comput Math Methods Med, 2014, 890725.
16	AL-FARIS A Q , NGAH U K , ISA N A M , et al. Computer-aided segmentation system for breast MRI tumour using modified automatic seeded region growing (BMRI-MASRG)[J]. J Digit Imaging, 2014, 27 (1): 133- 144. doi: 10.1007/s10278-013-9640-5
17	CHEN X, WILLIAMS B M, VALLABHANENI S R, et al. Learning active contour models for medical image segmentation[C]//CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2019.
18	KANNAN S R , RAMATHILAGAM S , DEVI P , et al. Improved fuzzy clustering algorithms in segmentation of DC-enhanced breast MRI[J]. J Med Syst, 2012, 36 (1): 321- 333. doi: 10.1007/s10916-010-9478-z
19	BAI P R , LIU Q Y , LEI L , et al. A novel region-based level set method initialized with mean shift clustering for automated medical image segmentation[J]. Comput Biol Med, 2013, 43 (11): 1827- 1832. doi: 10.1016/j.compbiomed.2013.08.024
20	PEREIRA S , PINTO A , ALVES V , et al. Brain tumor segmentation using convolutional neural networks in MRI images[J]. IEEE T Med Imaging, 2016, 35 (5): 1240- 1251. doi: 10.1109/TMI.2016.2538465
21	DOLZ J, AYED I B, DESROSIERS C. Dense multi-path U-Net for ischemic stroke lesion segmentation in multiple image modalities[C]//International MICCAI Brainlesion Workshop. Springer, Cham, 2018.
22	LEE P Q , GUIDA A , PATTERSON S , et al. Model-free prostate cancer segmentation from dynamic contrast-enhanced MRI with recurrent convolutional networks: A feasibility study[J]. Comput Med Imag Grap, 2019, 75, 14- 23. doi: 10.1016/j.compmedimag.2019.04.006
23	AMIT G, HADAD O, ALPERT S, et al. Hybrid mass detection in breast MRI combining unsupervised saliency analysis and deep learning[C]//International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2017.
24	BENJELLOUN M, ADOUI M E, LARHMAM M A, et al. Automated breast tumor segmentation in DCE-MRI using deep learning[C]//International Conference on Cloud Computing Technologies and Applications, 2018.
25	LONG J, SHELHAMER E, DARRELL T. Fully convolutional networks for semantic segmentation[C]//IEEE Conference on Computer Vision and Pattern Recognition. IEEE Computer Society, 2015: 3431-3440.
26	ZHANG J, SAHA A, ZHU Z, et al. Breast tumor segmentation in DCE-MRI using fully convolutional networks with an application in radiogenomics[C]//Computer-aided Diagnosis, 2018.
27	ZHANG J , SAHA A , ZHU Z , et al. Hierarchical convolutional neural networks for segmentation of breast tumors in MRI with application to radiogenomics[J]. IEEE T Med Imaging, 2019, 38 (2): 435- 447. doi: 10.1109/TMI.2018.2865671
28	ADOUI M E , MAHMOUDI S A , LARHMAM M A , et al. MRI breast tumor segmentation using different encoder and decoder CNN architectures[J]. Computers, 2019, 8 (3): 52. doi: 10.3390/computers8030052
29	LOFFE S , SZEGEDY C . Batch normalization: accelerating deep network training by reducing internal covariate shift[J]. arXiv preprint, 2015, 1502.03167.
30	RONNEBERGER O , FISCHER P , BROX T . U-Net: convolutional networks for biomedical image segmentation[M]. Spring Internation Publishing, 2015, 234- 241.
31	JIN Q G , MENG Z P , SUN C M , et al. RA-UNet: a hybrid deep attention-aware network to extract liver and tumor in CT scans[J]. Front Bioeng Biotechnol, 2020, 8, 605132. doi: 10.3389/fbioe.2020.605132
32	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
33	HU J , SHEN L , SUN G , et al. Squeeze-and-excitation networks[J]. arXiv preprint, 2017, 1709.01507.
34	SRIVASTAVA N , HINTON G , KRIZHEVSKY A , et al. Dropout: a simple way to prevent neural networks from overfitting[J]. J Mach Learn Res, 2014, 15 (1): 1929- 1958.
35	BADRINARAYANAN V , KENDALL A , CIPOLLA R . SegNet: a deep convolutional encoder-decoder architecture for image segmentation[J]. IEEE T Pattern Anal, 2017, 39 (12): 2481- 2495. doi: 10.1109/TPAMI.2016.2644615
36	ZHOU Z W, SIDDIQUEE M M R, TAJBAKHSH N, et al. UNet++: a nested U-Net architecture for medical image segmentation[C]//Deep Learning in Medical Image Anylysis and Multimodal Learning for Clinical Decision Support, 2018: 3-11.

结构		输入	输出	结构		输入	输出
block1	conv	128²*1	128²*32	block4	conv	16²*128	16²*64
	BN	128²*32	128²*32		BN	16²*64	16²*64
	max-pooling	128²*32	64²*32		max-pooling	16²*64	8²*64
block2	conv	64²*32	64²*64	block5	conv	8²*64	8²*32
	BN	64²*64	64²*64		BN	8²*32	8²*32
	max-pooling	64²*64	32²*64		max-pooling	8²*32	4²*32
block3	conv	32²*64	32²*128	flatten		4²*32	512
	BN	32²*128	32²*128	dense		512	64
	max-pooling	32²*128	16²*128	dense		64	2

结构^*	输入	输出
conv	128²*1	128²*64
conv	128²*64	128²*64
max-pooling	128²*64	64²*64
RA_block	64²* 64	64²*64
max-pooling	64²*64	32²*64
conv	32²*64	32²*128
conv	32²*128	32²*128
max-pooling	32²*128	16²*128
RA_block	16²*128	16²*128
max-pooling	16²*128	8²*128
conv	8²*256	8²*256
conv	8²*256	8²*256
up-conv	8²*256	16²*128
merge	16²128, 16²128	16²*256
RA_block	16²*256	16²*256
up-conv	16²*256	32²*128
merge	32²128, 32²128	32²*256
conv	32²*256	32²*128
conv	32²*128	32²*128
up-conv	32²*128	64²*64
merge	64²64, 64²64	64²*128
RA_block	64²*128	64²*128
up-conv	64²*128	128²*64
merge	128²64, 128²64	128²*128
conv	128²*128	128²*64
conv	128²*64	128²*2
conv	128²*2	128²*1

结构^*	输入^#	输出^#
conv	a²*c	a²*c
conv	a²*c	a²*c
GAP	a²*c	c
Dense+ReLU	c	c/4
Dense+Sigmoid	c/4	c
multiply	a²*c, c	a²*c
add	a²c, a²c	a²*c

方法		Dice系数		敏感度	特异性	IoU	耗时/s
方法		无后处理	有后处理	敏感度	特异性	IoU	耗时/s
U-Net		0.9001	0.9057	0.9469	0.9994	0.8425	1.8931
ResU-Net		0.9056	0.9142	0.9296	0.9988	0.8495	1.7236
U-Net++		0.9175	0.9198	0.9201	0.9985	0.8601	2.0370
RAU-Net		0.9202	0.9254	0.9566	0.9990	0.8624	1.7936

CBP5-Net+ RAU-Net	大肿瘤	0.9506	0.9579	0.9580	0.9983	0.9001
	小肿瘤	0.9167	0.9197	0.9466	0.9988	0.8535	1.8162
	平均	0.9336	0.9388	0.9523	0.9985	0.8768

[1]	Ying-shan WANG, Ao-qi DENG, Jin-ling MAO, Zhong-qi ZHU, Jie SHI, Guang YANG, Wei-wei MA, Qing LU, Hong-zhi WANG. Automatic Segmentation of Knee Joint Synovial Magnetic Resonance Images Based on 3D VNetTrans [J]. Chinese Journal of Magnetic Resonance, 2022, 39(3): 303-315.
[2]	Zhen-yu WANG, Ying-shan WANG, Jin-ling MAO, Wei-wei MA, Qing LU, Jie SHI, Hong-zhi WANG. Magnetic Resonance Images Segmentation of Synovium Based on Dense-UNet++ [J]. Chinese Journal of Magnetic Resonance, 2022, 39(2): 208-219.
[3]	Ying LIU,Yi-yun GUO,Jing-cong CHEN,Hao-wei ZHANG. Automatic Precise Segmentation of Cerebellopontine Angle Tumor Based on Faster-RCNN and Level-Set Method [J]. Chinese Journal of Magnetic Resonance, 2021, 38(3): 381-391.
[4]	Qin-yi SHI,Fang YAN,Yang YANG,Yue-fu CHEN,Xiao-lang LIN,Yuan-jun WANG. Image Segmentation of Tooth and Alveolar Bone with the Level Set Model [J]. Chinese Journal of Magnetic Resonance, 2021, 38(2): 182-193.
[5]	GONG Jin-chang, WANG Yu, WANG Yuan-jun. A Method for Segmentation of Glioma on Multimodal Magnetic Resonance Images Based on Wavelet Fusion and Deep Learning [J]. Chinese Journal of Magnetic Resonance, 2020, 37(2): 131-143.
[6]	XIAO Liang, LOU Yu-kun, ZHOU Hang-yu. A U-Net Network-Based Rapid Construction of Knee Models for Specific Absorption Rate Estimation [J]. Chinese Journal of Magnetic Resonance, 2020, 37(2): 144-151.
[7]	ZHANG Xue-ying, WANG Cheng-long, XIE Hai-bin, ZHANG Cheng-xiu, MA Chao, LU Jian-ping, YANG Guang. Parallel Segmentation and Tracking Algorithm for Magnetic Resonance Angiography Images Based on GPU [J]. Chinese Journal of Magnetic Resonance, 2016, 33(4): 570-580.