结直肠癌免疫组化图像分级诊断方法

doi:10.16180/j.cnki.issn1007-7820.2024.12.004

电子科技 ›› 2024, Vol. 37 ›› Issue (12): 24-31.doi: 10.16180/j.cnki.issn1007-7820.2024.12.004

结直肠癌免疫组化图像分级诊断方法

莫卓锐¹, 黄强豪², 张琳², 曹雨齐², 葛维挺³, 余明晖¹

1.华中科技大学人工智能与自动化学院,湖北武汉 430074
2.浙江大学控制科学与工程学院,浙江杭州 310027
3.浙江大学医学院附属第二医院,浙江杭州 310009

收稿日期:2023-04-03 出版日期:2024-12-15 发布日期:2024-12-16
作者简介:莫卓锐(1996-),男,硕士研究生。研究方向:决策系统、深度学习、计算机视觉。
曹雨齐(1994-),女,博士,副研究员。研究方向:医学图像处理、深度学习。
余明晖(1971-),男,博士,副教授。研究方向:决策支持系统、系统建模与仿真。
基金资助:
浙江省“尖兵”“领雁”研发攻关计划(2022C03002);军科委基础加强项目(2019-JCJQ-ZD-334-12)

Grading and Diagnostic Method for Colorectal Cancer Immunohistochemical Images

MO Zhuorui¹, HUANG Qianghao², ZHANG Lin², CAO Yuqi², GE Weiting³, YU Minghui¹

1. School of Artificial Intelligence and Automation,Huazhong University of Science and Technology,Wuhan 430074,China
2. College of Control Science and Engineering,Zhejiang University,Hangzhou 310027,China
3. The Second Affiliated Hospital School of Medicine,Zhejiang University,Hangzhou 310009,China

Received:2023-04-03 Online:2024-12-15 Published:2024-12-16
Supported by:
"Pioneer" and "Leading Goose" R&D Program of Zhejiang(2022C03002);Basic Reinforcement Project of the Military Science and Technology Commission(2019-JCJQ-ZD-334-12)

摘要/Abstract

摘要：

人体组织病理学检查主要应用于各类肿瘤诊断和治疗,免疫组织化学技术在结直肠癌早期筛查中具有重要的临床意义。为准确判断结直肠癌抑癌基因p53表达程度,文中提出一种基于逐块释放微调策略迁移学习的分级诊断方法,通过图像预处理、有监督模型预训练以及逐块释放微调等步骤将细胞核分割模型的参数迁移至诊断框架中。生成的细胞核分割掩膜进行主成分分析(Principal Component Analysis, PCA)降维和支持向量机(Support Vector Machine, SVM)多元分类最终得到图像诊断结果。该方法在结直肠癌p53蛋白免疫组化(Immunohistochemistry, IHC)图像数据集上进行了验证,模型的Dice值可达到0.887 6,分级准确率达到90.28%。结果表明,所提方法能够对结直肠癌免疫组化图像有效分级,为医生阅片提供可靠的辅助信息。

关键词: 免疫组织化学, 结直肠癌, 病理诊断, 有监督学习, 迁移学习, 细胞核分割, 微调策略, 聚类

Abstract:

Human tissue pathology examination is mainly used for the diagnosis and treatment of various tumors. Immunohistochemical technique has important clinical significance in the early screening of colorectal cancer. In order to accurately determine the expression level of the tumor suppressor gene p53 in colorectal cancer, this study proposes a grading diagnostic method based on transfer learning with block-wise fine-tuning strategy. The parameters of the cell nucleus segmentation model are transferred to the diagnostic framework through image preprocessing, supervised model pre-training, and block-wise fine-tuning. The generated cell nucleus segmentation mask is subjected to PCA(Principal Component Analysis) dimensionality reduction and SVM(Support Vector Machine) multivariate classification to obtain the final image diagnosis result. The proposed method is verified on colorectal cancer p53 protein IHC(Immunohistochemistry) image dataset. Dice value of the model reaches 0.887 6 and classification accuracy reaches 90.28%. The results show that the proposed method can effectively grade the immunohistochemical images of colorectal cancer, and provide reliable auxiliary information for doctors to read the film.

Key words: immunohistochemistry, colorectal cancer, pathological diagnosis, supervised learning, transfer learning, cell segmentation, fine-tuning strategy, clustering

中图分类号:

TP391.41

莫卓锐, 黄强豪, 张琳, 曹雨齐, 葛维挺, 余明晖. 结直肠癌免疫组化图像分级诊断方法[J]. 电子科技, 2024, 37(12): 24-31.

MO Zhuorui, HUANG Qianghao, ZHANG Lin, CAO Yuqi, GE Weiting, YU Minghui. Grading and Diagnostic Method for Colorectal Cancer Immunohistochemical Images[J]. Electronic Science and Technology, 2024, 37(12): 24-31.

图/表 14

图1

图2

图3

图4

图5

图6

图7

图8

表1

表2

表3

图9

表4

图10

参考文献 23

[1]	Deo S V, Sharma J, Kumar S. GLOBOCAN report on global cancer burden:Challenges and opportunities for surgical oncologists[J]. Annals of Surgical Oncology, 2022, 29(11):6497-6500. doi: 10.1245/s10434-022-12151-6 pmid: 35838905
[2]	Fay K. Colorectal cancer risk assessment and precision approaches to screening:Brave new world or worlds apart[J]. Gastroenterology, 2023, 164(5):812-827.
[3]	Burnett-Hartman A N, Newcomb P A, Peters U. Challenges with colorectal cancer family history assessment-motivation to translate polygenic risk scores into practice[J]. Gastroenterology, 2020, 158(2):433-435. doi: S0016-5085(19)41483-2 pmid: 31682850
[4]	Zhang H, Wen W, Yan J. Application of immunohisto chemistry technique in hydro biological studies[J]. Aquaculture and Fisheries, 2017, 2(3):140-144.
[5]	Hesamian M H, Jia W. Deep learning techniques for medical image segmentation:Achievements and challenges[J]. Journal of Digital Imaging, 2019, 32(4):582-596.
[6]	Tajbakhsh N, Jeyaseelan L. Embracing imperfect datasets:A review of deep learning solutions for medical image segmentation[J]. Medical Image Analysis, 2020, 63(6):693-701.
[7]	Zhang P, Zhong Y, Deng Y, et al. A survey on deep learning of small sample in biomedical image analysis[J]. J-STAGE, 2019, 31(8):117-121.
[8]	Kassani S H, Kassani P H. Deep transfer learning based model for colorectal cancer histopathology segmentation: A comparative study of deep pretrained models[J]. International Journal of Medical Informatics, 2022, 159(3):669-682.
[9]	Raghu M, Zhang C. Transfusion:Understanding transfer learning for medical imaging[C]. Vancouver: Advances in Neural Information Processing Systems, 2019:3347-3357.
[10]	Chen L, Bentley P, Mori K, et al. Self-supervised learning for medical image analysis using image context restoration[J]. Medical Image Analysis, 2019, 58(11):101539-101550.
[11]	Pathak D, Krahenbuhl P. Context encoders:Feature learning by inpainting[C]. Las Vegas: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2016:2536-2544.
[12]	Doersch C, Gupta A, Efros A A. Unsupervised visual representation learning by context prediction[C]. Santiago: Proceedings of the IEEE International Conference on Computer Vision,2015:1422-1430.
[13]	Noroozi M, Favaro P. Unsupervised learning of visual representations by solving jigsaw puzzles[C]. Amsterdam:Computer Vision ECCV: The Fourtrrnth European Conference,2016:69-84.
[14]	Naik G R. Advances in principal component analysis:Research and development[M]. Kingswood: Springer,2017:1-252.
[15]	SteinwartI, Christmann A. Support vector machines[M]. Los Alamos: Springer Science & Business Media,2008: 7-12.
[16]	林昌, 周海峰, 陈武. 基于双边滤波的高斯金字塔变换Retinex图像增强算法[J]. 激光与光电子学进展, 2020, 57(16):161019-161023.
	Lin Chang, Zhou Haifeng, Chen Wu. Gaussian pyramid transform Retinex image enhancement algorithm based on bilateral filtering[J]. Laser & Optoelectronics Progress, 2020, 57(16):161019-161023.
[17]	Kapoor A, Singhal A. A comparative study of K-means, K-means++ and fuzzy C-means clustering algorithms[C]. Ghaziabad: The Third International Conference on Computational Intelligence & Communication Technology,2017:1-6.
[18]	Davri A, Birbas E. Deep learning on histopathological images for colorectal cancer diagnosis:A systematic review[J]. Diagnostics, 2022, 12(4):837-846.
[19]	Ronneberger O, Fischer P, Brox T. U-Net:Convolutional networks for biomedical image segmentation[C]. Munich:Medical Image Computing and Computer-Assisted Intervention: The Eighteenth International Conference,2015:234-241.
[20]	贝琛圆, 于海滨, 潘勉, 等. 基于改进U-Net网络的腺体细胞图像分割算法[J]. 电子科技, 2019, 32(11):18-22.
	Bei Chenyuan, Yu Haibin, Pan Mian, et al. Gland cell image segmentation algorithm based on improved U-Net network[J]. Electronic Science and Technology, 2019, 32(11):18-22.
[21]	Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. Transactions on Fundamentals of Electronics,Communications and Computer Sciences, 2014, 1409(5):1124-1131.
[22]	He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]. Las Vegas: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2016:770-778.
[23]	Huang G, Liu Z, Van D M L, et al. Densely connected convolutional networks[C]. Honolulu: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2017:4700-4708.

最新释放网络块	对应输出维度	微调后模型Dice值
Block₈	128×128	0.752 3
Block₇	64×64	0.822 4
Block₆	32×32	0.783 6
Block₅	16×16	0.692 8

添加卷积层数	对应卷积层参数	Dice系数
1	卷积层(1,(1,1), 激活函数='Softmax')	0.822 4
2	卷积层(32,(3,3), 步长=(1,1), 激活函数='ReLU') 卷积层(1,(1,1), 激活函数='Softmax')	0.862 8
3	卷积层(64,(3,3), 步长=(1,1), 激活函数='ReLU') 卷积层(32,(3,3), 步长=(1,1), 激活函数='ReLU') 卷积层(1,(1,1), 激活函数='Softmax')	0.831 2

微调操作	平均Dice系数
迁移预训练网络权重重构最后一层为卷积(1,(1,1))	0.624 9
释放Block₇~Block₈网络块并微调模型	0.822 4
在倒数第二层的位置添加卷积(32,(3,3))	0.862 8
结合结直肠癌IHC分级诊断标签筛选训练数据	0.887 6

序号	主成分个数	测试集占比	迭代误差精度	图像域	准确率 /%
1	150	0.2	1×10^-9	BGR	72.30
2	200	0.2	1×10^-9	BGR	84.50
3	220	0.2	1×10^-9	BGR	80.20
4	200	0.2	1×10^-6	BGR	76.60
5	200	0.2	1×10^-9	BGR	84.50
6	200	0.2	1×10^-12	BGR	86.10
7	200	0.2	1×10^-12	YUV	78.90
8	200	0.2	1×10^-12	HSV	90.28

结直肠癌免疫组化图像分级诊断方法

Grading and Diagnostic Method for Colorectal Cancer Immunohistochemical Images

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 14

参考文献 23

相关文章 15

Metrics

本文评价

推荐阅读 0

[1]	曹春萍, 许志华. 融合片内语义和片间结构特征的自监督CT图像分类方法[J]. 电子科技, 2024, 37(7): 43-52.
[2]	吴家葆, 曾国辉, 张振华. 基于k-sums分段聚类的动态组合学习光伏短期功率预测[J]. 电子科技, 2024, 37(4): 69-76.
[3]	崔晶楠, 黄春艳, 李艳玲. 改进YOLOv5s算法的钢材表面缺陷检测[J]. 电子科技, 2024, 37(12): 48-55.
[4]	张铭坤,蔡文郁,张帅. 基于LiDAR扫描角度修正的障碍目标定位方法[J]. 电子科技, 2024, 37(1): 41-47.
[5]	赖彬,王森. 利用深度可变形轮廓模型进行结构体裂纹视觉检测[J]. 电子科技, 2023, 36(9): 35-40.
[6]	欧敬逸,田颖,向鑫,宋启哲. 基于迁移BN-CNN框架的小样本工业过程故障诊断[J]. 电子科技, 2023, 36(7): 49-55.
[7]	张明,章国宝,朱宏伟. 融合IMU信息的三维激光SLAM方法[J]. 电子科技, 2023, 36(6): 27-33.
[8]	陈纯,舒慧生,阚秀,孙维周. 基于分峰思想的煤岩显微组分识别与统计分析[J]. 电子科技, 2023, 36(4): 9-20.
[9]	王桥,胡春燕,李菲菲. 基于深度迁移学习与多尺度特征融合的场景识别方法[J]. 电子科技, 2023, 36(11): 19-27.
[10]	张崇崇,黄亚宇. GA-BP神经网络对片烟结构的预测研究[J]. 电子科技, 2022, 35(6): 35-42.
[11]	程顺达,程颖,孙士江. 基于机器学习的肿瘤智能辅助诊断方法[J]. 电子科技, 2022, 35(5): 56-59.
[12]	金霄,吴飞,鄢松,陆雯霞,张忠艺. 基于GAWK-means的地铁车站指纹定位方法[J]. 电子科技, 2022, 35(2): 34-39.
[13]	王夏霖,阚秀,范艺璇. 基于LSTM-Attention的P300事件相关电位识别分类研究[J]. 电子科技, 2022, 35(12): 10-16.
[14]	方凯,史志才,贾媛媛. 基于混合聚类的k-匿名数据发布算法[J]. 电子科技, 2022, 35(12): 78-83.
[15]	林静,胡德敏,王揆豪. 差分隐私模糊聚类位置保护方法[J]. 电子科技, 2022, 35(11): 64-71.