高分辨率皮肤黑色素瘤图像的两阶段式分割算法

doi:10.19678/j.issn.1000-3428.0066329

摘要/Abstract

摘要：

皮肤黑色素瘤切片图像分辨率过大且病理特征表现形式多样，现有很多分割算法结果不精准同时消耗巨大显卡内存。针对该问题，提出一种低显存消耗的两阶段式精细分割算法。该算法第一阶段采用全局分割网络对以ResNet50为骨干的特征金字塔结构进行改进，图像特征提取过程中使用全局金字塔平均池化模块增强图像全局语义信息的提取，并采用多尺度特征融合分支将高层特征图的语义信息融入到低层特征图中，增强低层特征图语义信息的表征能力。第二阶段采用一种全局到局部的精细分割策略，以全局分割结果为基准对图像进行剪裁，得到一个较小的候选区域，将其输入到局部分割网络中，局部分割网络仅处理候选区域内的像素并与全局网络对应层共享图像特征，精细分割结果的同时减少显存的消耗。在经典数据集ISIC2018上的实验结果显示，该算法的准确度和IOU分别达到93.5%和82.1%，相较于对比的经典分割算法精度最高且占用的显卡内存减少了22.8%~36.9%，能有效适用于高分辨率皮肤病灶图像的分割任务。

关键词: 两阶段式分割, ResNet50, 特征金字塔结构, 全局金字塔平均池化模块, 多尺度特征融合分支

Abstract:

Skin melanoma imaging faces a few challengesthe required resolution is often large, the pathological features of skin melanoma are diverse, and many segmentation algorithms are not sufficiently accurate and require huge amounts of GPU memory. Aiming at this problem, this paper proposes a low GPU memory consumption algorithm based on two-stage fine segmentation. The global segmentation network adopted in the first stage of the algorithm improves the feature pyramid structure with ResNet50 as the backbone. During image feature extraction, the Global Pyramid Average Pooling Module(GPAPM)is used to enhance the extraction of image global semantic information. The multi-scale feature fusion branch is adopted to integrate the semantic information of the high-level feature map into the lower-level feature map to enhance the representation ability of the semantic information of the low-level feature map. In the second stage, a global to local fine segmentation strategy is adopted. The image is clipped based on the global segmentation results to reduce the candidate area, and the image is then input into the local segmentation network. The local segmentation network only processes pixels in the candidate region and shares image features with the corresponding layer of the global network. The GPU memory consumption is minimized while the segmentation results are improved. The algorithm is verified on the classic dataset ISIC2018. The experimental results show that the accuracy and IOU of the algorithm reach 93.5% and 82.1%, respectively, and compared with other classical segmentation algorithms, the algorithm has the highest accuracy and reduces the GPU memory by 22.8% to 36.9%, indicating its effective applicability to the segmentation of high resolution skin melanoma images.

Key words: two-stage segmentation, ResNet50, feature pyramid structure, Global Pyramid Average Pooling Module(GPAPM), multi-scale feature fusion branch

贵向泉, 张馨月, 李立. 高分辨率皮肤黑色素瘤图像的两阶段式分割算法[J]. 计算机工程, 2023, 49(11): 267-274.

Xiangquan GUI, Xinyue ZHANG, Li LI. Two-Stage Segmentation Algorithm of High Resolution Skin Melanoma Image[J]. Computer Engineering, 2023, 49(11): 267-274.

https://www.ecice06.com/CN/Y2023/V49/I11/267

图/表 10

图1 全局金字塔平均池化模块

Fig.1 Global pyramid average pooling module

图2 语义嵌入分支模块示意图

Fig.2 Diagram of semantic embedding branch module

图3 全局分支网络模型

Fig.3 Global branch network model

图4 两阶段式精细分割策略示意图

Fig.4 Schematic diagram of two-stage fine segmentation strategy

图5 ISIC2018数据集部分图像

Fig.5 Partial images of ISIC2018 dataset

图6 实验结果可视化对比图

Fig.6 Visualization comparison chart of experimental results

参考文献 26

1	胡露, 郑德义, 李伟人. 皮肤恶性黑色素瘤的外科治疗进展. 医学综述, 2022, 28(9): 1695- 1700. URL
	HU L, ZHENG D Y, LI W R. Advances in surgical treatment of cutaneous malignant melanoma. Medical Recapitulate, 2022, 28(9): 1695- 1700. URL
2	XIMENES V, FRANCISCO F, MEDEIROS A G, et al. Automatic skin lesions segmentation based on a new morphological approach via geodesic active contour. Cognitive Systems Research, 2019, 55, 44- 59. doi: 10.1016/j.cogsys.2018.12.008
3	CELEBI M E, WEN Q A, HWANG S, et al. Lesion border detection in dermoscopy images using ensembles of thresholding methods. Skin Research and Technology, 2013, 19(1): 252- 258.
4	WONG A, SCHARCANSKI J, FIEGUTH P. Automatic skin lesion segmentation via iterative stochastic region merging. IEEE Transactions on Information Technology in Biomedicine, 2011, 15(6): 929- 936. doi: 10.1109/TITB.2011.2157829
5	ZHOU H Y, SCHAEFER G, SADKA A H, et al. Anisotropic mean shift based fuzzy c-means segmentation of dermoscopy images. IEEE Journal of Selected Topics in Signal Processing, 2009, 3(1): 26- 34. doi: 10.1109/JSTSP.2008.2010631
6	IMTIAZ I, AHMED I, JEON G, et al. An efficient image processing and machine learning based technique for skin lesion segmentation and classification[C]//Proceedings of 2021 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference. Washington D. C., USA: IEEE Press, 2021: 1499-1505.
7	ATTIA M, HOSSNY M, NAHAVANDI S, et al. Skin melanoma segmentation using recurrent and convolutional neural networks[C]//Proceedings of the 14th International Symposium on Biomedical Imaging. Washington D. C., USA: IEEE Press, 2017: 292-296.
8	LONG J, SHELHAMER E, DARRELL T, et al. Fully convolutional networks for semantic segmentation[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2015: 3431-3440.
9	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. Berlin, Germany: Springer, 2015: 234-241.
10	BADRINARAYANAN V, KENDALL A, CIPOLLA R. SegNet: a deep convolutional encoder-decoder architecture for image segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(12): 2481- 2495. doi: 10.1109/TPAMI.2016.2644615
11	王雪. 基于U-Net多尺度和多维度特征融合的皮肤病变分割方法. 吉林大学学报(理学版), 2021, 59(1): 123- 127. URL
	WANG X. Skin lesion segmentation method based on U-Net with multi-scale and multi-dimensional feature fusion. Journal of Jilin University(Science Edition), 2021, 59(1): 123- 127. URL
12	HUSSAIN R, BASAK H. RecU-Net: improved utilization of receptive fields in U-Net for skin lesion segmentation[C]//Proceedings of the 18th India Council International Conference. Washington D. C., USA: IEEE Press, 2022: 1-6.
13	RAJINIKANTH V, KADRY S, DAMAŠEVIČIUS R, et al. Skin melanoma segmentation using VGG-UNet with Adam/SGD optimizer: a study[C]//Proceedings of the 3rd International Conference on Intelligent Computing Instrumentation and Control Technologies. Washington D. C., USA: IEEE Press, 2022: 982-986.
14	NINH Q C, TRAN T T, TRAN T T, et al. Skin lesion segmentation based on modification of SegNet neural networks[C]//Proceedings of the 6th NAFOSTED Conference on Information and Computer Science. Washington D. C., USA: IEEE Press, 2020: 575-578.
15	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[EB/OL]. [2022-09-10]. https://arxiv.org/abs/1706.03762.
16	WU H S, PAN J Q, LI Z Y, et al. Automated skin lesion segmentation via an adaptive dual attention module. IEEE Transactions on Medical Imaging, 2021, 40(1): 357- 370. doi: 10.1109/TMI.2020.3027341
17	AGHDAM E K, AZAD R, ZARVANI M, et al. Attention Swin U-Net: cross-contextual attention mechanism for skin lesion segmentation[EB/OL]. [2022-09-10]. https://arxiv.org/abs/2210.16898.
18	齐永锋, 侯璐璐, 段友放. 基于DenseNet-BC网络的皮肤镜下皮肤损伤分割. 计算机工程与科学, 2020, 42(6): 1060- 1067. URL
	QI Y F, HOU L L, DUAN Y F. Dermatological skin lesion segmentation based on DenseNet-BC network. Computer Engineering and Science, 2020, 42(6): 1060- 1067. URL
19	YAO Y K, ZHANG Y C, YAO X K, et al. Image segmentation of skin lesions based on attention mechanism and deep convolutional neural network[C]//Proceedings of International Conference on Intelligent Transportation, Big Data and Smart City. Washington D. C., USA: IEEE Press, 2021: 659-662.
20	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2016: 770-778.
21	ZHANG Z L, ZHANG X Y, PENG C, et al. ExFuse: enhancing feature fusion for semantic segmentation[C]//Proceedings of European Conference on Computer Vision. Berlin, Germany: Springer, 2018: 273-288.
22	CHEN W Y, JIANG Z Y, WANG Z Y, et al. Collaborative global-local networks for memory-efficient segmentation of ultra-high resolution images[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2020: 8916-8925.
23	LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2017: 936-944.
24	LIN T Y, GOYAL P, GIRSHICK R, et al. Focal Loss for dense object detection[C]//Proceedings of IEEE International Conference on Computer Vision. Washington D. C., USA: IEEE Press, 2017: 2999-3007.
25	CHEN L C, PAPANDREOU G, SCHROFF F, et al. Rethinking atrous convolution for semantic image segmentation[EB/OL]. [2022-09-10]. https://arxiv.org/abs/1706.05587.
26	ZHAO H S, SHI J P, QI X J, et al. Pyramid scene parsing network[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2017: 6230-6239.

[1]	邹慧海, 侯进. 改进SSD算法的道路小目标检测研究[J]. 计算机工程, 2022, 48(5): 281-288.
[2]	谢瑞, 邵堃, 霍星, MITHUN Md Masud Parvej. 一种改进的DIQaM_FR/NR图像质量评价模型[J]. 计算机工程, 2020, 46(8): 258-263,270.

选择文件类型/文献管理软件名称

选择包含的内容