Texture analysis and machine learning based on T2 weighted image in distinguishing renal angiomyolipoma without visible fat and renal cell carcinoma

Share this content in WeChat

• CLINICAL ARTICLE •

LIU Zhenhao BAI Xu YE Huiyi GUO Aitao LIN Mingquan ZUO Panli WANG Haiyi

Cite this article as: Liu ZH, Bai X, Ye HY, et al. Texture analysis and machine learning based on T2 weighted image in distinguishing renal angiomyolipoma without visible fat and renal cell carcinoma[J]. Chin J Magn Reson Imaging, 2021, 12(2): 38-42. DOI:10.12015/issn.1674-8034.2021.02.009.

Abstract

[Abstract] Objective To distinguish between renal angiomyolipoma without visible fat (AMLwvf) and renal cell carcinoma (RCC) using T2WI texture analysis and machine learning. Materials andMethods 80 cases of renal tumors were analyzed retrospectively, including AMLwvf (n=20), clear cell renal cell carcinoma (n=20), papillary renal cell carcinoma (n=20) and chromophobe renal cell carcinoma (n=20). Lesions were delineated on software by two radiologists to extract the corresponding volumes of interest (VOI) and then 93 features were generated. The Kruskal Wallis test showed that there was no significant difference between renal carcinoma subtypes, so renal carcinoma subtypes were combined into one group (renal carcinoma, n=60). Univariable analysis was carried out through Mann-Whitney U test and Holm-Bonferroni method to find the best features and analyze the diagnostic performance. Modeling with multiple features: after the primary selection of features by Pearson correlation coefficient, the C5.0 node of IBM SPSS modeler software calculated the relative importance ranking of features. Top 2, 3, 4 and 5 most important features were used to form 4 feature subsets. Decision tree C5.0 model was built with or without boosting. The differentiation and generalization ability of each model was evaluated to find the best one as the final model.Results Univariable analysis: after Holm-Bonferroni correction, four different features were screened: minimum, 10 percentile, difference variance and contrast. The area under the curve was 0.888, 0.837, 0.789 and 0.777, respectively. The range of positive predictive value was 50.00%—69.57%. Modeling with multiple features: 8 decision tree C5.0 models were constructed. The area under the curve of final model was 0.950. The sensitivity, specificity, positive predictive value, negative predictive value and accuracy of final model were 90.00%, 100%, 100%, 96.77% and 97.5%, respectively. The accuracy based on cross validation is 95.0%.Conclusions Univariable analysis based on T2WI has limited clinical application value because of its low positive predictive value. Decision tree C5.0 model has high accuracy and good generalization ability to distinguish AMLwvf and RCC, which is helpful to make reasonable treatment plan in clinic.

[Keywords] kidney neoplasms；magnetic resonance imaging；texture analysis；machine learning；angiomyolipoma without visible fat

Contributor Information

LIU Zhenhao^{1,
2} BAI Xu¹ YE Huiyi¹ GUO Aitao³ LIN Mingquan⁴ ZUO Panli⁵ WANG Haiyi^1*

¹ Department of Radiology, the first medical center of Chinese PLA General Hospital, Beijing 100853, China

² Department of Radiology, Affiliated Hospital of Changzhi Institute of Traditional Chinese Medicine, Changzhi 046000, China

³ Department of Pathology, the first medical center of Chinese PLA General Hospital; Beijing 100853, China

⁴ Department of Electronic Engineering, City University of Hong Kong, Hong Kong 999077, China

⁵ Innovation and Collaboration Center, Huiying Medical Technology (Beijing) Co., Ltd, Beijing 100192, China

Wang HY, E-mail: wanghaiyi301@outlook.com

Conflicts of interest None.

Publication Information

ACKNOWLEDGMENTS This work was part of National Natural Science Foundation of China (No. 81471641).

Received 2020-09-14

Accepted 2021-01-12

DOI: 10.12015/issn.1674-8034.2021.02.009

Text

References

Halpenny D, Snow A, McNeill G, et al. The radiological diagnosis and treatment of renal angiomyolipoma-current status[J]. Clin Radiol, 2010, 65(2): 99-108. DOI: 10.1016/j.crad.2009.09.014.

Ljungberg B, Bensalah K, Canfield S, et al. EAU guidelines on renal cell carcinoma: 2014 update[J]. Eur Urol, 2015, 67(5): 913-924. DOI: 10.1016/j.eururo.2015.01.005.

Li Q, Liu ZH, Wang HY, et al. Application of multimodal MRI features in nuclear grading of renal clear cell carcinoma[J]. Natl Med J China, 2019, 99(23): 1767-1772. DOI: 10.3760/cma.j.issn.0376-2491.2019.23.003.

Nie P, Yang G, Wang Z, et al. A CT-based radiomics nomogram for differentiation of renal angiomyolipoma without visible fat from homogeneous clear cell renal cell carcinoma[J]. Eur Radiol, 2020, 30(2): 1274-1284. DOI: 10.1007/s00330-019-06427-x.

Simpfendorfer C, Herts BR, Motta-Ramirez GA, et al. Angiomyolipoma with minimal fat on MDCT: can counts of negative-attenuation pixels aid diagnosis?[J]. AJR Am J Roentgenol, 2009, 192(2): 438-443. DOI: 10.2214/ajr.08.1180.

Woo S, Kim SY, Cho JY, et al. Differentiation between papillary renal cell carcinoma and fat-poor angiomyolipoma: a preliminary study assessing detection of intratumoral hemorrhage with chemical shift MRI and T2*-weighted gradient echo[J]. Acta Radiol, 2018, 59(5): 627-634. DOI: 10.1177/0284185117723371.

Wobker SE, Williamson SR. Modern pathologic diagnosis of renal oncocytoma[J]. J Kidney Cancer VHL, 2017, 4(4): 1-12. DOI: 10.15586/jkcvhl.2017.96.

Woo S, Suh CH, Cho JY, et al. Diagnostic performance of CT for diagnosis of fat-poor angiomyolipoma in patients with renal masses: A systematic review and meta-analysis[J]. AJR Am J Roentgenol, 2017, 209(5): 297-307. DOI: 10.2214/ajr.17.18184.

Chen LS, Zhu ZQ, Wang ZT, et al. Chemical shift magnetic resonance imaging for distinguishing minimal-fat renal angiomyolipoma from renal cell carcinoma: a meta-analysis[J]. Eur Radiol, 2018, 28(5): 1854-1861. DOI: 10.1007/s00330-017-5141-0.

Lassel EA, Rao R, Schwenke C, et al. Diffusion-weighted imaging of focal renal lesions: a meta-analysis[J]. Eur Radiol, 2014, 24(1): 241-249. DOI: 10.1007/s00330-013-3004-x.

Chung MS, Choi HJ, Kim MH, et al. Comparison of T2-weighted MRI with and without fat suppression for differentiating renal angiomyolipomas without visible fat from other renal tumors[J]. AJR Am J Roentgenol, 2014, 202(4): 765-771. DOI: 10.2214/ajr.13.11058.

Hodgdon T, McInnes MD, Schieda N, et al. Can quantitative CT texture analysis be used to differentiate fat-poor renal angiomyolipoma from renal cell carcinoma on unenhanced CT images?[J]. Radiology, 2015, 276(3): 787-796. DOI: 10.1148/radiol.2015142215.

Silverman SG, Pedrosa I, Ellis JH, et al. Bosniak classification of cystic renal masses, version 2019: An update proposal and needs assessment[J]. Radiology, 2019, 292(2): 475-488. DOI: 10.1148/radiol.2019182646.

You ZY. Classification optimization method for unbalanced samples[M]. Wuhan: Huazhong Univ Sci & Technol, 2018. DOI: 10.7666/d.D01542247.

Raman SP, Chen Y, Schroeder JL, et al. CT texture analysis of renal masses: pilot study using random forest classification for prediction of pathology[J]. Acad Radiol, 2014, 21(12): 1587-1596. DOI: 10.1016/j.acra.2014.07.023.

Moch H, Cubilla AL, Humphrey PA, et al. The 2016 WHO classification of tumours of the urinary system and male genital organs-Part A: renal, penile, and testicular tumours[J]. Eur Urol, 2016, 70(1): 93-105. DOI: 10.1016/j.eururo.2016.02.029.

Zhang YY, Luo S, Liu Y, et al. Angiomyolipoma with minimal fat: differentiation from papillary renal cell carcinoma by helical CT[J]. Clin Radiol, 2013, 68(4): 365-370. DOI: 10.1016/j.crad.2012.08.028.

Zhou M, Chen Y, Liu J, et al. A predicting model of bone marrow malignant infiltration in (18)F-FDG PET/CT images with increased diffuse bone marrow FDG uptake[J]. J Cancer, 2018, 9(10): 1737-1744. DOI: 10.7150/jca.24836.

Wang R, Wang Y, Li J, et al. Block-regularized m × 2 cross-validated estimator of the generalization error[J]. Neural Comput, 2017, 29(2): 519-554. DOI: 10.1162/NECO_a_00923.

You MW, Kim N, Choi HJ. The value of quantitative CT texture analysis in differentiation of angiomyolipoma without visible fat from clear cell renal cell carcinoma on four-phase contrast-enhanced CT images[J]. Clin Radiol, 2019, 74(7): 547-554. DOI: 10.1016/j.crad.2019.02.018.

Li H, Li A, Zhu H, et al. Whole-tumor quantitative apparent diffusion coefficient histogram and texture analysis to differentiation of minimal fat angiomyolipoma from clear cell renal cell carcinoma[J]. Acad Radiol, 2019, 26(5): 632-639. DOI: 10.1016/j.acra.2018.06.015.

Choi HJ, Kim JK, Ahn H, et al. Value of T2-weighted MR imaging in differentiating low-fat renal angiomyolipomas from other renal tumors[J]. Acta Radiol, 2011, 52(3): 349-353. DOI: 10.1258/ar.2010.090491.

Moriyama S, Yoshida S, Tanaka H, et al. Intensity ratio curve analysis of small renal masses on T2-weighted magnetic resonance imaging: Differentiation of fat-poor angiomyolipoma from renal cell carcinoma[J]. Int J Urol, 2018, 25(6): 554-560. DOI: 10.1111/iju.13561.

Semelka RC, Chew W, Hricak H, et al. Fat-saturation MR imaging of the upper abdomen[J]. AJR Am J Roentgenol, 1990, 155(5): 1111-1116. DOI: 10.2214/ajr.155.5.2120945.

Shuman WP, Baron RL, Peters MJ, et al. Comparison of STIR and spin-echo MR imaging at 1.5 T in 90 lesions of the chest, liver, and pelvis[J]. AJR Am J Roentgenol, 1989, 152(4): 853-859. DOI: 10.2214/ajr.152.4.853.

Daniels DL, Kneeland JB, Shimakawa A, et al. MR imaging of the optic nerve and sheath: correcting the chemical shift misregistration effect[J]. AJNR Am J Neuroradiol, 1986, 7(2): 249-253.

PREV The preliminary study on the parameter optimization of IVIM-DWI pulse sequence in gastric cancer patients with 3.0 T MR

NEXT Study of amplitude of low-frequency fluctuations and brain functional connectivity in patients with end-stage renal disease of Uygur population in Southern of Xinjiang

LINK

Tel & Fax: +8610-67113815 E-mail: editor@cjmri.cn