USSL Net: Focusing on Structural Similarity with Light U-Structure for Stroke Lesion Segmentation

doi:10.1007/s12204-022-2412-y

Abstract

Abstract: Automatic segmentation of ischemic stroke lesions from computed tomography (CT) images is of great significance for identifying and curing this life-threatening condition. However, in addition to the problem of low image contrast, it is also challenged by the complex changes in the appearance of the stroke area and the difficulty in obtaining image data. Considering that it is difficult to obtain stroke data and labels, a data enhancement algorithm for one-shot medical image segmentation based on data augmentation using learned transformation was proposed to increase the number of data sets for more accurate segmentation. A deep convolutional neural network based algorithm for stroke lesion segmentation, called structural similarity with light U-structure (USSL) Net, was proposed. We embedded a convolution module that combines switchable normalization, multi-scale convolution and dilated convolution in the network for better segmentation performance. Besides, considering the strong structural similarity between multi-modal stroke CT images, the USSL Net uses the correlation maximized structural similarity loss (SSL) function as the loss function to learn the varying shapes of the lesions. The experimental results show that our framework has achieved results in the following aspects. First, the data obtained by adding our data enhancement algorithm is better than the data directly segmented from the multi- modal image. Second, the performance of our network model is better than that of other models for stroke segmentation tasks. Third, the way SSL functioned as a loss function is more helpful to the improvement of segmentation accuracy than the cross-entropy loss function.

Key words: structural similarity, medical image segmentation, deep convolution neural network, automatic data enhancement algorithm

CLC Number:

TP391

JIANG Zhiguo (蒋志国), CHANG Qing∗ (常青). USSL Net: Focusing on Structural Similarity with Light U-Structure for Stroke Lesion Segmentation[J]. J Shanghai Jiaotong Univ Sci, 2022, 27(4): 485-497.

References 41

[1]	SUDLOW C L, W ARLOW C P. Comparable studies of the incidence of stroke and its pathological types:Results from an international collaboration [J]. Stroke,1997, 28(3): 491-499.
[2]	DOYLE K P, SIMON R P, STENZEL-POORE M P.Mechanisms of ischemic brain damage [J]. Neuropharmacology, 2008, 55(3): 310-318.
[3]	GONZALEZ R G, HIRSCH J A, KOROSHETZ W J,et al. Acute ischemic stroke. imaging and intervention[J]. Journal of Neuroradiology, 2006, 33(3): 193.
[4]	ZAHARCHUK G, EL MOGY I S, FISCHBEIN N J,et al. Comparison of arterial spin labeling and bolusperfusion-weighted imaging for detecting mismatch in acute stroke [J]. Stroke, 2012, 43(7): 1843-1848.
[5]	MEZZAPESA D M, PETRUZZELLIS M, LUCIVEROV, et al. Multimodal MR examination in acute is-chemic stroke [J]. Neuroradiology, 2006, 48(4): 238-246.
[6]	DONAHUE J, WINTERMARK M. Perfusion CT and acute stroke imaging: Foundations, applications, and literature review [J]. Journal of Neuroradiology, 2015,42(1): 21-29.
[7]	GILLEBERT C R, HUMPHREYS G W, MANTINID. Automated delineation of stroke lesions using brain CT images [J]. Neuro Image: Clinical, 2014, 4: 540-548.
[8]	TURECKOV A A, RODRíGUEZ-S áNCHEZ A J.ISLES Challenge: U-shaped convolution neural net-work with dilated convolution for 3d stroke lesion seg-mentation [M]//Brainlesion: Glioma, multiple sclerosis, stroke and traumatic brain injuries. Cham:Springer, 2019: 319-327.
[9]	ABULNAGA S M, RUBIN J. Ischemic stroke lesion segmentation in CT perfusion scans using pyramidpooling and focal loss [M]//Brainlesion: Glioma, multiple sclerosis, stroke and traumatic brain injuries.Cham: Springer, 2019: 352-363.
[10]	ZHAO S, WU B X, CHU W Q, et al. Correlation maximized structural similarity loss forsemantic segmentation [EB/OL]. (2019-10-19).https://arxiv.org/abs/1910.08711.
[11]	AKKUS Z, GALIMZIANOV A A, HOOGI A, et al.Deep learning for brain MRI segmentation: State of the art and future directions [J]. Journal of Digital Imaging, 2017, 30(4): 449-459.
[12]	OLIVEIRA A, PEREIRA S, SIL V A C A. Augmenting data when training a CNN for retinal vessel segmentation: How to warp? [C]//2017 IEEE 5th Portuguese Meeting on Bioengineering. Coimbra: IEEE, 2017: 1-4.
[13]	PEREIRA S, PINTO A, AL VES V, et al. Brain tumorsegmentation using convolutional neural networks in MRI images [J]. IEEE Transactions on Medical Imaging, 2016, 35(5): 1240-1251.
[14]	NAZARI-F ARSANI S, NYMAN M, KARJALAINENT, et al. Automated segmentation of acute stroke lesions using a data-driven anomaly detection on diffusion weighted MRI [J]. Journal of Neuroscience Methods, 2020, 333: 108575.
[15]	ROTH H R, LU L, F ARAG A, et al. DeepOrgan:Multi-level deep convolutional networks for automatedpancreas segmentation [M]//Medical image computing and computer-assisted intervention - MICCAI 2015.Cham: Springer, 2015: 556-564.
[16]	ZHAO A, BALAKRISHNAN G, DURAND F, etal. Data augmentation using learned transformations for one-shot medical image segmentation [C]//2019IEEE/CVF Conference on Computer Vision and Pat-tern Recognition. Long Beach, CA: IEEE, 2019: 8535-8545.
[17]	MOESKOPS P, VIERGEVER M A, MENDRIK A M,et al. Automatic segmentation of MR brain images with a convolutional neural network [J]. IEEE Trans-actions on Medical Imaging, 2016, 35(5): 1252-1261.
[18]	MILLETARI F, NA V AB N, AHMADI S A. V-net:Fully convolutional neural networks for volumetric medical image segmentation [C]//2016 Fourth Inter-national Conference on 3D Vision. Stanford, CA:IEEE, 2016: 565-571.
[19]	DOSOVITSKIY A, FISCHER P, SPRINGENBERGJ T, et al. Discriminative unsupervised feature learning with exemplar convolutional neural networks [J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(9): 1734-1747.
[20]	MAIER O, MENZE B H, VON DER GABLENTZ J,et al. ISLES 2015: A public evaluation benchmark forischemic stroke lesion segmentation from multispectral MRI [J]. Medical Image Analysis, 2017, 35: 250-269.
[21]	WINZECK S, HAKIM A, MCKINLEY R, et al. ISLES2016 and 2017: Benchmarking ischemic stroke lesion outcome prediction based on multispectral MRI [J].Frontiers in Neurology, 2018, 9: 679.
[22]	YAHIAOUI A F Z, BESSAID A. Segmentation of is-chemic stroke area from CT brain images [C]//2016International Symposium on Signal, Image, Video and Communications. Tunis: IEEE, 2016: 13-17.
[23]	ABRAHAM N, KHAN N M. A novel focal Tversky loss function with improved attention U-net for lesion segmentation [C]//2019 IEEE 16th International Symposium on Biomedical Imaging. Venice: IEEE, 2019:683-687.[
24	] SHEN D G, WU G R, SUK H I. Deep learning in med-ical image analysis [J]. Annual Review of Biomedical Engineering, 2017, 19: 221-248.
[25]	PINHEIRO G R, VOLTOLINE R, BENTO M, etal. V-net and U-net for ischemic stroke lesion seg-mentation in a small dataset of perfusion data[M]//Brainlesion: Glioma, multiple sclerosis, strokeand traumatic brain injuries. Cham: Springer, 2019:301-309.
[26]	ANAND V K, KHENED M, ALEX V, et al. Fullyautomatic segmentation for ischemic stroke using CT perfusion maps [M]//Brainlesion: Glioma, multiplesclerosis, stroke and traumatic brain injuries. Cham:Springer, 2019: 328-334.
[27]	CLèRIGUES A, V AL VERDE S, BERNAL J, et al.A c u t e i s c h e m i c s t r o k e l e s i o n c o r e s e g m e n t a t i o n i n C Tperfusion images using fully convolutional neural net-works [J]. Computers in Biology and Medicine, 2019,115: 103487.
[28]	LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection [C]//2017 IEEE International Conference on Computer Vision. Venice: IEEE, 2017:2999-3007.
[29]	REKIK I, ALLASSONNIèRE S, CARPENTER T K,et al. Medical image analysis methods in MR/CT-imaged acute-subacute ischemic stroke lesion: Segmentation, prediction and insights into dynamic evolution simulation models. A critical appraisal [J]. NeuroImage: Clinical, 2012, 1(1): 164-178.
[30]	KABIR Y, DOJAT M, SCHERRER B, et al. Multimodal MRI segmentation of ischemic stroke lesions[C]//2007 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society.Lyon: IEEE, 2007: 1595-1598.
[31]	KAMNITSAS K, LEDIG C, NEWCOMBE V F J, et al. Efficient multi-scale 3D CNN with fully connectedCRF for accurate brain lesion segmentation [J]. Medical Image Analysis, 2017, 36: 61-78.
[32]	ZHOU Y J, HUANG W J, DONG P, et al. D-UNet: A dimension-fusion U shape network for chronicstroke lesion segmentation [J]. IEEE/ACM Transactions on Computational Biology and Bioinformatics,2021, 18(3): 940-950.
[33]	LUO P, REN J M, PENG Z L. Differentiable learning-to-normalize via switchable normalization [EB/OL]. (2018-06-28). https://arxiv.org/abs/1806.10779.
[34]	DOLZ J, AYED I B, DESROSIERS C. Dense multi-path U-Net for ischemic stroke lesion segmentation in multiple image modalities [M]//Brainlesion: Glioma,multiple sclerosis, stroke and traumatic brain injuries.Cham: Springer, 2019: 271-282.
[35]	LOU M, QI Y L, LI X R, et al. Aggregated pyramidattention network for mass segmentation in mammograms [J]. Multimedia Tools and Applications, 2021.https://doi.org/10.1007/s11042-021-10940-x.
[36]	LOU M, QI Y L, MENG J, et al. DCANet: Dualcontextual affinity network for mass segmentation in whole mammograms [J]. Medical Physics, 2021, 48(8):4291-4303.
[37]	SALEHI S S M, ERDOGMUS D, GHOLIPOUR A.Tversky loss function for image segmentation using3D fully convolutional deep networks [M]//Machinelearning in medical imaging. Cham: Springer, 2019:379-387.
[38]	SUDRE C H, LI W Q, VERCAUTEREN T, et al. Generalised Dice overlap as a deep learning loss function for highly unbalanced segmentations [M]//Deep learn-ing in medical image analysis and multimodal learning for clinical decision support. Cham: Springer, 2017:240-248.
[39]	PISANA F, HENZLER T, SCH ?NBERG S, et al.Noise reduction and functional maps image quality im-provement in dynamic CT perfusion using a new K-means clustering guided bilateral filter (KMGB) [J].Medical Physics, 2017, 44(7): 3464-3482.
[40]	CHEN L C, PAPANDREOU G, KOKKINOS I, etal. DeepLab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully con-nected CRFs [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018, 40(4): 834-848.
[41]	C? I C? E K ?, ABDULKADIR A, LIENKAMP S S, etal. 3D U-net: Learning dense volumetric segmentationfrom sparse annotation [C]//Medical image computing and computer-assisted intervention - MICCAI 2016.Cham: Springer, 2016: 424-432.