膀胱癌MRI图像存在肿瘤边界不清晰、肿瘤区域较小、肿瘤分布不连续等问题,现有的分割算法参数量庞大,计算复杂,且分割精度有待提高。因此,设计了一种多尺度特征融合的轻量化膀胱癌分割算法(pyramidal convolution lightweight network, PylNet),该算法在编码阶段设计的多尺度语义特征提取模块可提取不同尺度的肿瘤区域信息,确保对微小肿瘤信息提取的可靠性和全面性;在解码阶段设计的融合模块可以在保证分割精度的同时,极大地减少算法参数量和复杂度。实验结果表明,相较于FCN8s、DeepLabV3+、U-Net等算法,PylNet算法分割精度有一定的提高,Dice系数达88.40%,参数量是FCN8s的1/13,可实现对膀胱MRI的快速分割。
There exists serveral challenges inMRI images of bladder cancer, such as unclear tumor boundaries, small tumor areas, and discontinuous tumor distribution.The existing segmentation algorithms have huge parameters and complex calculations, and the accuracy and efficiency of the existing methods for bladder tumor segmentation need to be improved. In responseto the above issues, a lightweight bladder cancer segmentation algorithm based on multi-scale feature fusion (PylNet) was proposed in this paper. The algorithm can extract information of different scales through the multi-scale semantic feature extraction module. This module was designed in the coding stage to ensure the reliability and comprehensiveness of the extraction of information on small tumor regions. At the same time, the fusion module designed in another stage can quickly complete the tumor region segmentation, and the amount of parameters used is much less. Experimental results show that compared with FCN8s, SegNet, U-Net and other algorithms, the segmentation accuracy of this algorithm is improved to a certain extentwhere DSC reaches 88.40%, and the parameter amount is 1/13 of FCN8s, which a fast bladder MRI segmentation is achieved.
[1] JANSEN I,LUCAS M,BOSSCHIETER J,et al.Automated detection and grading of non-muscle-invasive urothelial cell carcinoma of the bladder[J].The American Journal of Pathology,2020,190(7):1483-1490.
[2] 熊巧,曾蜀雄,阳青松,等.多参数磁共振在膀胱癌病理分级和T分期中的研究新进展[J].临床泌尿外科杂志,2020,35(1):68-73.
XIONG Q,ZENG S X,YANG Q S,et al.Advances in multiparametric MRI in grading and staging of bladder cancer[J].Journal of Clinical Urology,2020,35(1):68-73.
[3] 韦荣超,吴承耀,张振声,等. 膀胱镜检查在膀胱癌诊断的研究进展[J].第二军医大学学报,2012,33(11):1257-1259.
WEI R C, WU C Y, ZHANG Z S, et al. Cystoscopy in the diagnosis of bladder cancer: recent progress[J].Academic Journal of Second Military Medical University, 2012,33(11):1257-1259.
[4] DANIELS M J,BARRY E,SCHOENBERG M,et al.Contemporary oncologic outcomes of second induction course BCG in patients with nonmuscle invasive bladder cancer[J].Urologic Oncology:Seminars and Original Investigations,2020,38(1):5.e9-5.e16.
[5] 王修德,谢红锋,顾峰.常规磁共振成像结合磁共振泌尿系水成像对膀胱癌及其分期的诊断价值[J].影像研究与医学应用,2021,5(1):55-56.
WANG X D,XIE H F,GU F.The diagnostic value of conventional magnetic resonance imaging combined with magnetic resonance hydrography of urinary system in bladder cancer and its staging[J].Journal of Imaging Research and Medical Applications,2021,5(1):55-56.
[6] 平秦榕,颜汝平,王剑松.膀胱癌早期诊断方法的临床价值[J].医学与哲学(B),2015,36(3):66-68,97.
PING Q R,YAN R P,WANG J S.The clinical value of early diagnostic methods of bladder cancer[J].Medicine & Philosophy (B),2015,36(3):66-68,97.
[7] 蔡亚洁,李畅,杜悦,等.基于深度学习的膀胱肿瘤MRI图像分级分期预测[J].电脑知识与技术,2021,17(1):29-31.
CAI Y J,LI C,DU Y,et al.Deep learning based MRI image grading and staging prediction for bladder tumors[J].Computer Knowledge and Technology,2021,17(1):29-31.
[8] 丁力,戚聪聪.MR成像在膀胱癌诊断与分期中的价值研究[J].影像研究与医学应用,2020,4(7):65-66.
DING L,QI C C.The value of MR imaging in diagnosis and staging of bladder cancer[J].Journal of Imaging Research and Medical Applications,2020,4(7):65-66.
[9] 鲍中文,段继忠,杨俊东.基于小波树稀疏结构的磁共振成像快速重构算法[J].陕西师范大学学报(自然科学版),2020,48(6):1-9.
BAO Z W,DUAN J Z,YANG J D.Fast reconstruction algorithm of magnetic resonance imaging based on wavelet tree sparsity structure[J].Journal of Shaanxi Normal University (Natural Science Edition),2020,48(6):1-9.
[10] XIAO D,ZHANG G P,LIU Y,et al.3D detection and extraction of bladder tumors via MR virtual cystoscopy[J].International Journal of Computer Assisted Radiology and Surgery,2016,11(1):89-97.
[11] ZHENG H J, XU X P, ZHANG X, et al. The segmentation of bladder cancer using the voxel-features-based method[C]//The 25th Annual Meeting of SPIE Medical Imaging. San Diego, USA: SPIE,2019.
[12] ZHANG Y J,DUAN C J,YE D T,et al.The segmentation of MR bladder wall in 3D based on minimum closed set model[C]//2013 IEEE International Conference on Medical Imaging Physics and Engineering.Shenyang,China:IEEE,2013:319-323.
[13] MA X Y, HADJIISKI L, WEI J, et al.U-Net-based deep-learning bladder segmentation in CT urography[J]. Medical Physics, 2019, 46(4): 1752-1765.
[14] DOLZ J,XU X P,RONY J,et al.Multiregion segmentation of bladder cancer structures in MRI with progressive dilated convolutional networks[J].Medical Physics,2018,45(12):5482-5493.
[15] LIU J X,LIU L B,XU B L,et al.Bladder cancer multi-class segmentation in MRI with pyramid-in-pyramid network[C]//2019 IEEE 16th International Symposium on Biomedical Imaging.Venice,Italy:IEEE,2019:28-31.
[16] CHA K H,HADJIISKI L M,SAMALA R K,et al.Bladder cancer segmentation in CT for treatment response assessment:application of deep-learning convolution neural network-a pilot study[J].Tomography (Ann Arbor,Mich.),2016,2(4):421-429.
[17] 韩文忠,康莉,江静婉,等.深度全卷积网络对MRI膀胱图像的分割[J].信号处理,2019,35(3):443-450.
HAN W Z,KANG L,JIANG J W,et al.Segmentation of MRI bladder images by deep fully convolutional network[J].Journal of Signal Processing,2019,35(3):443-450.
[18] LI C Y,FAN Y X,CAI X D.PyConvU-Net:a lightweight and multiscale network for biomedical image segmentation[J].BMC Bioinformatics,2021,22(1):14.
[19] PASZKE A, GROSS S, MASSA F, et al. PyTorch: an imperative style, high-performance deep learning library[C]//Advances in Neural Information Processing Systems. Vancouver, Canada: NeurIPS, 2019: 8024-8035.
[20] 付顺兵,王朝斌,罗建,等.基于改进U-Net模型的脑肿瘤MR图像分割[J].西华师范大学学报(自然科学版),2021,42(2):202-208.
FU S B,WANG C B,LUO J,et al.Segmentation method of brain tumor MR image based on improved U-Net model[J].Journal of China West Normal University (Natural Sciences),2021,42(2):202-208.
[21] YU C Q,GAO C X,WANG J B,et al.BiSeNet V2:bilateral network with guided aggregation for real-time semantic segmentation[J].International Journal of Computer Vision,2021,129(11):3051-3068.
