Share：
Share this content in WeChat
X
[Chinese][PDF]8955
• Special Focus •
Prediction of distant metastasis in nasopharyngeal carcinoma by interpretable machine learning model based on multiparametric MRI radiomics and clinical factors
JIN Zhe  ZHANG Bin  ZHANG Lu  ZHANG Shuixing 

Cite this article as: Jin Z, Zhang B, Zhang L, et al. Prediction of distant metastasis in nasopharyngeal carcinoma by interpretable machine learning model based on multiparametric MRI radiomics and clinical factors[J]. Chin J Magn Reson Imaging, 2022, 13(11): 22-29. DOI:10.12015/issn.1674-8034.2022.11.005.


[Abstract] Objective To establish a machine learning prediction model based on multi-parametric MRI features and clinical variables, and evaluate its efficacy in predicting distant metastasis in nasopharyngeal carcinoma (NPC) before treatment.Materials and Methods MRI images of 1393 patients with pathologically confirmed NPC from three hospitals (1049 in the training cohort and 344 in the external validation cohort) from June 2010 to September 2017 were retrospectively analyzed. We used ITK-SNAP and Pyradiomics to delineate regions of interest and extract radiomic features, respectively. Features were selected using correlation analysis, univariate analysis and recursive feature elimination (RFE) method. The gradient boosting machine (GBM) algorithm was utilized to construct models. Receiver operating characteristic (ROC) curve and area under the curve (AUC) were used to compare the predictive efficacy of the models, and decision curve analysis (DCA) was used to assess the clinical utility. The SHapley Additive exPlanation (SHAP) algorithm was used to attribute interpretability to the optimal prediction model.Results Ten radiomic features were finally selected. GBM_R, GBM_C and GBM_RC models were constructed based on the three features combination: radiomic features, clinical variables, and radiomic features + clinical variables. The AUC values of the them on the training set were 0.938, 0.724, and 0.938, respectively. GBM_RC (hereafter, NPC-Wise) achieved the highest AUC value of 0.775 in the external validation set. The SHAP force plot provided a visualization of the direction and degree of influence of each feature on the predicting results of the model.Conclusions The interpretable machine learning prediction model NPC-Wise, based on multi-parametric MRI radiomic features and clinical variables, showed good performance in predicting the risk of distant metastasis in NPC, as well as providing individual-level interpretability with the SHAP algorithm, which can provide a valuable decision basis for personalized treatment.
[Keywords] nasopharyngeal carcinoma；distant metastasis；radiomics；Gradient Boosting Machine；interpretability；magnetic resonance imaging

JIN Zhe   ZHANG Bin   ZHANG Lu   ZHANG Shuixing*  

Department of Radiology, the First Affiliated Hospital of Jinan University, Guangzhou 510627, China

Zhang SX, E-mail: zsx7515@jnu.edu.cn

Conflicts of interest   None.

ACKNOWLEDGMENTS National Natural Science Foundation of China (No. 81871323).
Received  2022-08-02
Accepted  2022-11-14
DOI: 10.12015/issn.1674-8034.2022.11.005
Cite this article as: Jin Z, Zhang B, Zhang L, et al. Prediction of distant metastasis in nasopharyngeal carcinoma by interpretable machine learning model based on multiparametric MRI radiomics and clinical factors[J]. Chin J Magn Reson Imaging, 2022, 13(11): 22-29. DOI:10.12015/issn.1674-8034.2022.11.005.

[1]
Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries[J]. CA Cancer J Clin, 2021, 71(3): 209-249. DOI: 10.3322/caac.21660.
[2]
Ji MF, Sheng W, Cheng WM, et al. Incidence and mortality of nasopharyngeal carcinoma: interim analysis of a cluster randomized controlled screening trial (PRO-NPC-001) in Southern China[J]. Ann Oncol, 2019, 30(10): 1630-1637. DOI: 10.1093/annonc/mdz231.
[3]
Ou SI, Zell JA, Ziogas A, et al. Epidemiology of nasopharyngeal carcinoma in the United States: improved survival of Chinese patients within the keratinizing squamous cell carcinoma histology[J]. Ann Oncol, 2007, 18(1): 29-35. DOI: 10.1093/annonc/mdl320.
[4]
Chen YP, Chan ATC, Le QT, et al. Nasopharyngeal carcinoma[J]. Lancet, 2019, 394(10192): 64-80. DOI: 10.1016/S0140-6736(19)30956-0.
[5]
Wang WY, Twu CW, Chen HH, et al. Plasma EBV DNA clearance rate as a novel prognostic marker for metastatic/recurrent nasopharyngeal carcinoma[J]. Clin Cancer Res, 2010, 16(3): 1016-1024. DOI: 10.1158/1078-0432.CCR-09-2796.
[6]
Avanzo M, Stancanello J, El Naqa I. Beyond imaging: the promise of radiomics[J]. Phys Med, 2017, 38: 122-139. DOI: 10.1016/j.ejmp.2017.05.071.
[7]
Lambin P, Leijenaar RTH, Deist TM, et al. Radiomics: the bridge between medical imaging and personalized medicine[J]. Nat Rev Clin Oncol, 2017, 14(12): 749-762. DOI: 10.1038/nrclinonc.2017.141.
[8]
Gillies RJ, Kinahan PE, Hricak H. Radiomics: images are more than pictures, they are data[J]. Radiology, 2016, 278(2): 563-577. DOI: 10.1148/radiol.2015151169.
[9]
Chan AT, Grégoire V, Lefebvre JL, et al. Nasopharyngeal cancer: EHNS-ESMO-ESTRO Clinical Practice Guidelines for diagnosis, treatment and follow-up[J/OL]. Ann Oncol, 2012, 23(Suppl 7): vii83-vii85 [2022-08-01]. https://www.annalsofoncology.org/action/showPdf?pii=S0923-7534%2819%2937667-7. DOI: 10.1093/annonc/mds266.
[10]
Lin JC, Wang WY, Chen KY, et al. Quantification of plasma Epstein-Barr virus DNA in patients with advanced nasopharyngeal carcinoma[J]. N Engl J Med, 2004, 350(24): 2461-2470. DOI: 10.1056/NEJMoa032260.
[11]
Leung SF, Zee B, Ma BB, et al. Plasma Epstein-Barr viral deoxyribonucleic acid quantitation complements tumor-node-metastasis staging prognostication in nasopharyngeal carcinoma[J]. J Clin Oncol, 2006, 24(34): 5414-5418. DOI: 10.1200/JCO.2006.07.7982.
[12]
Pan JJ, Ng WT, Zong JF, et al. Proposal for the 8th edition of the AJCC/UICC staging system for nasopharyngeal cancer in the era of intensity-modulated radiotherapy[J]. Cancer, 2016, 122(4): 546-558. DOI: 10.1002/cncr.29795.
[13]
Bi Y, Xiang DX, Ge ZY, et al. An interpretable prediction model for identifying N7-methylguanosine sites based on XGBoost and SHAP[J]. Mol Ther Nucleic Acids, 2020, 22: 362-372. DOI: 10.1016/j.omtn.2020.08.022.
[14]
Tseng PY, Chen YT, Wang CH, et al. Prediction of the development of acute kidney injury following cardiac surgery by machine learning[J/OL]. Crit Care, 2020, 24(1): 478 [2022-08-01]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7395374/pdf/13054_2020_Article_3179.pdf. DOI: 10.1186/s13054-020-03179-9.
[15]
Lundberg SM, Nair B, Vavilala MS, et al. Explainable machine-learning predictions for the prevention of hypoxaemia during surgery[J]. Nat Biomed Eng, 2018, 2(10): 749-760. DOI: 10.1038/s41551-018-0304-0.
[16]
Zhang L, Wu XJ, Liu J, et al. MRI-based deep-learning model for distant metastasis-free survival in locoregionally advanced nasopharyngeal carcinoma[J]. J Magn Reson Imaging, 2021, 53(1): 167-178. DOI: 10.1002/jmri.27308.
[17]
Wu XJ, Dong D, Zhang L, et al. Exploring the predictive value of additional peritumoral regions based on deep learning and radiomics: a multicenter study[J]. Med Phys, 2021, 48(5): 2374-2385. DOI: 10.1002/mp.14767.
[18]
Zhang L, Dong D, Li HL, et al. Development and validation of a magnetic resonance imaging-based model for the prediction of distant metastasis before initial treatment of nasopharyngeal carcinoma: a retrospective cohort study[J]. EBioMedicine, 2019, 40: 327-335. DOI: 10.1016/j.ebiom.2019.01.013.
[19]
Peng LH, Hong XT, Yuan QY, et al. Prediction of local recurrence and distant metastasis using radiomics analysis of pretreatment nasopharyngeal [18F]FDG PET/CT images[J]. Ann Nucl Med, 2021, 35(4): 458-468. DOI: 10.1007/s12149-021-01585-9.
[20]
Ren XY, Yang XJ, Cheng B, et al. HOPX hypermethylation promotes metastasis via activating SNAIL transcription in nasopharyngeal carcinoma[J/OL]. Nat Commun, 2017, 8: 14053 [2022-08-01]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5296651/pdf/ncomms14053.pdf. DOI: 10.1038/ncomms14053.
[21]
Huang CL, Sun ZQ, Guo R, et al. Plasma Epstein-Barr virus DNA load after induction chemotherapy predicts outcome in locoregionally advanced nasopharyngeal carcinoma[J]. Int J Radiat Oncol Biol Phys, 2019, 104(2): 355-361. DOI: 10.1016/j.ijrobp.2019.01.007.
[22]
An X, Wang FH, Ding PR, et al. Plasma Epstein-Barr virus DNA level strongly predicts survival in metastatic/recurrent nasopharyngeal carcinoma treated with palliative chemotherapy[J]. Cancer, 2011, 117(16): 3750-3757. DOI: 10.1002/cncr.25932.
[23]
Peng H, Dong D, Fang MJ, et al. Prognostic value of deep learning PET/CT-based radiomics: potential role for future individual induction chemotherapy in advanced nasopharyngeal carcinoma[J]. Clin Cancer Res, 2019, 25(14): 4271-4279. DOI: 10.1158/1078-0432.CCR-18-3065.
[24]
Qiang MY, Li CF, Sun YY, et al. A prognostic predictive system based on deep learning for locoregionally advanced nasopharyngeal carcinoma[J]. J Natl Cancer Inst, 2021, 113(5): 606-615. DOI: 10.1093/jnci/djaa149.
[25]
Zhong LZ, Dong D, Fang XL, et al. A deep learning-based radiomic nomogram for prognosis and treatment decision in advanced nasopharyngeal carcinoma: a multicentre study[J/OL]. EBioMedicine, 2021, 70: 103522 [2022-08-01]. https://www.thelancet.com/action/showPdf?pii=S2352-3964%2821%2900315-7. DOI: 10.1016/j.ebiom.2021.103522.
[26]
Haibe-Kains B, Adam GA, Hosny A, et al. Transparency and reproducibility in artificial intelligence[J/OL]. Nature, 2020, 586(7829): E14-E16 [2022-08-01]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8144864/pdf/nihms-1699061.pdf. DOI: 10.1038/s41586-020-2766-y.
[27]
Zhang B, Tian J, Pei S, et al. Machine learning-assisted system for thyroid nodule diagnosis[J]. Thyroid, 2019, 29(6): 858-867. DOI: 10.1089/thy.2018.0380.
[28]
Park VY, Han K, Seong YK, et al. Diagnosis of thyroid nodules: performance of a deep learning convolutional neural network model vs. radiologists[J/OL]. Sci Rep, 2019, 9(1): 17843 [2022-08-08]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6882804/pdf/41598_2019_Article_54434.pdf. DOI: 10.1038/s41598-019-54434-1.
[29]
Li XC, Zhang S, Zhang Q, et al. Diagnosis of thyroid cancer using deep convolutional neural network models applied to sonographic images: a retrospective, multicohort, diagnostic study[J]. Lancet Oncol, 2019, 20(2): 193-201. DOI: 10.1016/S1470-2045(18)30762-9.
[30]
Peng S, Liu YH, Lv WM, et al. Deep learning-based artificial intelligence model to assist thyroid nodule diagnosis and management: a multicentre diagnostic study[J/OL]. Lancet Digit Health, 2021, 3(4): e250-e259 [2022-08-01]. https://www.thelancet.com/journals/landig/article/PIIS2589-7500(21)00041-8/fulltext. DOI: 10.1016/S2589-7500(21)00041-8.
[31]
Doupe P, Faghmous J, Basu S. Machine learning for health services researchers[J]. Value Health, 2019, 22(7): 808-815. DOI: 10.1016/j.jval.2019.02.012.
[32]
Li K, Yao S, Zhang Z, et al. Efficient gradient boosting for prognostic biomarker discovery[J]. Bioinformatics, 2022, 38(6):1631-1638. DOI: 10.1093/bioinformatics/btab869.
[33]
Golden CE, Rothrock MJ, Mishra A. Comparison between random forest and gradient boosting machine methods for predicting Listeria spp. prevalence in the environment of pastured poultry farms[J]. Food Res Int, 2019, 122: 47-55. DOI: 10.1016/j.foodres.2019.03.062.
[34]
Thorsen-Meyer HC, Nielsen AB, Nielsen AP, et al. Dynamic and explainable machine learning prediction of mortality in patients in the intensive care unit: a retrospective study of high-frequency data in electronic patient records[J/OL]. Lancet Digit Health, 2020, 2(4): e179-e191 [2022-08-01]. https://www.thelancet.com/journals/landig/article/PIIS2589-7500(20)30018-2/fulltext. DOI: 10.1016/S2589-7500(20)30018-2.

PREV The value of quantitative parameters on magnetization transfer imaging of lacrimal glands in distinguishing the clinical activity of thyroid-associated ophthalmopathy
NEXT Prediction of tumor cell proliferation activity in nasopharyngeal carcinoma by nomogram based on multiparametric MRI radiomics combined with clinic-radiological features
  


LINK


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