北京生物医学工程

基于CT影像组学特征的肾肿瘤组织学亚型分类

Classification of renal tumor histology subtypes based on radiomicsfeatures of CT images

作者：杨熠钱旭升周志勇沈钧康朱建兵戴亚康

单位：1中国科学技术大学（合肥 230027） 2中国科学院苏州生物医学工程技术研究所（苏州 215163） 3苏州大学附属第二医院（苏州 215163） 4南京医科大学附属苏州医院（苏州 215163） 5苏州科技城医院（苏州 215163）

关键词：乏脂肪血管平滑肌脂肪瘤；肾透明细胞癌；影像组学；机器学习

分类号：R318.04

出版年·卷·期（页码）：2020·39·1（15-20）

摘要：

目的在术前准确鉴别乏脂肪血管平滑肌脂肪瘤（fat-poor angiomyolipoma, fp-AML）和肾透明细胞癌（clear cell renal cell carcinoma, ccRCC）对制定正确的治疗方案是至关重要的。为了提高fp-AML和ccRCC的分类准确率，本文提出一种基于影像组学技术的分类模型。方法回顾性地收集苏州大学附属第二医院放射科18例fp-AML患者和42例ccRCC患者的CT图像。首先，从CT图像中提取430个影像组学特征。然后，分三步进行特征选择：计算皮尔森相关矩阵剔除冗余特征；使用Welch’s t检验确定具有显著差异的特征；利用序列浮动前向选择算法选择具有鉴别能力的特征。最后，建立k最近邻（k-nearest neighborhood, kNN）、随机森林（random forest, RF）、支持向量机（support vector machine, SVM）和AdaBoost四种分类器进行分类。结果 SVM分类器所构建的模型获得了最佳分类性能，正确率、敏感度、特异性、阳性预测值、阴性预测值和 ROC曲线下面积分别为91.67%、88.89%、92.86%、84.21%、95.12%和0.9418。结论本研究所构建的模型能提高fp-AML和ccRCC的分类准确率，能辅助医生进行fp-AML和ccRCC的鉴别诊断。

Objective Accurate preoperative differential diagnosis of fat-poor angiomyolipoma (fp-AML) and clear cell renal cell carcinoma (ccRCC) is essential for proper treatment planning. In order to increase the accuracy of discrimination of fp-AML from ccRCC, we develop a classification model based on radiomics technology. Methods The study retrospectively collected CT images of 18 cases with fp-AML and 42 cases with ccRCC from department of radiology, the Second Affiliated Hospital of Suzhou University. Firstly, 430 radiomics features were extracted from CT images. Then, the feature selection was carried by three steps: Pearson’s correlation matrices were calculated to remove redundant features, Welch’s t-test was utilized to determine the statistically significant features, and sequential forward floating selection method was used to select the discriminative features. Finally, k-nearest neighborhood, random forest, support vector machine and AdaBoost classifiers were built for classification. Results The model built by SVM classifier achieved the best classification performance, with accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and area under the receiver operating characteristic curves of 91.67%, 88.89%, 92.86%, 84.21%, 95.12%, and 0.9418. Conclusions The proposed model can increase the classification accuracy of discrimination of fp-AML from ccRCC, and has great potential in helping radiologists to discriminate fp-AML from ccRCC.

参考文献：

[1] Hsieh JJ, Purdue MP, Signoretti S, et al. Renal cell carcinoma [J]. Nature Reviews Disease Primers, 2017, 3: 17009.

[2] Wei J, Zhao J, Zhang X, et al. Analysis of dual energy spectral CT and pathological grading of clear cell renal cell carcinoma (ccRCC) [J]. Plos One, 2018, 13(5): e0195699.

[3] Prasad SR, Surabhi VR, C. O. Menias, et al. Chintapalli. Benign Renal Neoplasms in Adults: Cross-Sectional Imaging Findings [J]. American Journal of Roentgenology, 2008, 190(1): 158-164.

[4] Lane BR, Aydin H, Danforth TL, et al. Clinical Correlates of Renal Angiomyolipoma Subtypes in 209 Patients: Classic, Fat Poor, Tuberous Sclerosis Associated and Epithelioid [J]. The Journal of Urology, 2008, 80(3): 836-843.

[5] Yan L, Liu Z, Wang G, et al. Angiomyolipoma with Minimal Fat : Differentiation From Clear Cell Renal Cell Carcinoma and Papillary Renal Cell Carcinoma by Texture Analysis on CT Images [J]. Academic Radiology, 2015, 22(9): 1115.

[6] 许玥. 基于CT图像纹理分析鉴别乏脂肪肾血管平滑肌脂肪瘤与肾透明细胞癌的初步研究 [D]. 沈阳：中国医科大学, 2017.

Xu Y. Preliminary study on differential diagnosis of renal angiomyolipoma and renal clear cell carcinoma based on CT image texture analysis [D]. Shenyang: China Medical University, 2017.

[7] Feng Z, Rong P, Cao P, et al. Machine learning-based quantitative texture analysis of CT images of small renal masses: Differentiation of angiomyolipoma without visible fat from renal cell carcinoma [J]. European Radiology, 2018, 28(4): 1625-1633.

[8] Lee HS, Hong H, Jung DC, et al. Differentiation of fat-poor angiomyolipoma from clear cell renal cell carcinoma in contrast-enhanced MDCT images using quantitative feature classification [J]. Medical Physics, 2017, 44(7): 3604-3614.

[9] Lambin P, Leijenaar RTH, Deist TM, et al. Radiomics: the bridge between medical imaging and personalized medicine [J]. Nature Reviews Clinical Oncology, 2017, 14: 749.

[10] Bernardi ED, Buda A, Guerra L, et al. Radiomics of the primary tumour as a tool to improve 18F-FDG-PET sensitivity in detecting nodal metastases in endometrial cancer [J]. EJNMMI Research, 2018, 8(1): 86.

[11] Ferreira Junior JR, Koenigkam-Santos M, Cipriano FEG, et al. Radiomics-based features for pattern recognition of lung cancer histopathology and metastases [J]. Computer Methods and Programs in Biomedicine, 2018, 159: 23-30.

[12] Sapate SG, Mahajan A, Talbar SN, et al. Radiomics based detection and characterization of suspicious lesions on full field digital mammograms [J]. Computer Methods and Programs in Biomedicine, 2018, 163: 1-20.

[13] Meng X, Xia W, Xie P, et al. Preoperative radiomic signature based on multiparametric magnetic resonance imaging for noninvasive evaluation of biological characteristics in rectal cancer [J]. European Radiology, 2019, 29(6): 3200-3209.

[14] Inza I, Larra?aga P, and Saeys Y. A review of feature selection techniques in bioinformatics [J]. Bioinformatics, 2007, 23(19): 2507-2517.

[15] Zwanenburg A, Leger S, Vallières M, et al. Image biomarker standardisation initiative - feature definitions [J]. arXiv preprint arXiv: 1612.07003, 2016.

[16] Lubner MG, Smith AD, Sandrasegaran K, et al. CT Texture Analysis: Definitions, Applications, Biologic Correlates, and Challenges [J]. RadioGraphics, 2017, 37(5): 1483-1503.

[17] Harikumar R and Vinoth kumar B. Performance analysis of neural networks for classification of medical images with wavelets as a feature extractor [J]. International Journal of Imaging Systems and Technology, 2015, 25(1): 33-40.

[18] Pearson K. Note on regression and inheritance in the case of two parents [J]. Proceedings of the Royal Society of London, 1895, 58: 240-242.

[19] Derrick B and White P. Why Welch's test is Type I error robust [J]. The Quantitative Methods in Psychology, 2016, 12(1): 30-38.

[20] Aha DW and Bankert RL. A comparative evaluation of sequential feature selection algorithms [M]// Learning from Data: Artificial Intelligence and Statistics V. D. Fisher and H.-J. Lenz, Eds. New York: Springer New York, 1996: 199-206.

[21] Bugdol MD, Bugdol MN, Lipowicz AM, et al. Prediction of menarcheal status of girls using voice features [J]. Computers in Biology and Medicine, 2018, 100: 296-304.

[22] Fallahpour S, Lakvan EN, and Zadeh MH. Using an ensemble classifier based on sequential floating forward selection for financial distress prediction problem [J]. Journal of Retailing and Consumer Services, 2017, 34: 159-167.

[23] Chapelle O, Vapnik V, Bousquet O, et al. Choosing Multiple Parameters for Support Vector Machines [J]. Machine Learning, 2002, 46(1): 131-159.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录