Application of supervised and semi-supervised learning prediction models to predict progression to cirrhosis in chronic hepatitis C

Abstract

In this study, we aim to examine the efficacy of deep learning methods in predicting the 1-year risk of developing cirrhosis in patients with chronic hepatitis C (CHC), as defined by transient elastography (TE), in comparison with conventional models, as well as to assess whether semi-supervised learning can improve performance relative to supervised learning when the labels are limited. We used the electronic health records of the 169,317 valid patients in the Veterans Health Administration system from 2000 to 2016. Predictor variables contained baseline characteristics, such as age, gender, race, hepatitis C virus genotype, and 26 liver-related longitudinal variables such as sustained virologic response and laboratory data. The response variable, developing cirrhosis, is defined as liver stiffness >12.5 kPa on TE within a 1-year window. Using baseline and longitudinal variables, we fitted four prediction models, including logistic regression (LR), random forest (RF), supervised recurrent neural network (RNN), and semi-supervised RNN (semi-RNN) and evaluated their performances. Both RNN (area under the receiver operating characteristic curve [AuROC] 0.744) and semi-RNN (AuROC 0.785) accurately predicted the risk of cirrhosis within 1 year and significantly outperformed RF (AuROC 0.731) and LR (AuROC 0.724). By enabling early identification of high-risk patients, these models hold promise for targeted interventions in clinical CHC treatment.

Keywords

Semi-supervised learning / Electronic health records / Longitudinal predictors

Funding

The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. Drs. Waljee, Beste, Ioannou, and Su are funded by IIR 16-024 from the United States (U.S.) Department of Veterans Affairs Health Services R&D (HSRD) Service.

Conflict of interest

Grace Su, MD, is an equity owner of Applied Morphomics and Prenovo. Dr. Su has a patent with the University of Michigan regarding image analysis of liver disease. Dr. Su has received funding from the National Institutes of Health, the Department of Veteran Affairs, and the Department of Defense. Dr. Su has received but there is no conflict of interest with the preparation of this manuscript. The remaining authors report they have no competing interests as well.

References

[1] Zeremski M, Dimova RB, Pillardy J, de Jong YP, Jacobson IM, Talal AH. Fibrosis progression in patients with chronic hepatitis C virus infection. J Infect Dis. 2016; 214(8):1164-1170. doi: 10.1093/infdis/jiw332

[2] Freeman AJ, Law MG, Kadlor JM, Dore GJ. Predicting progression to cirrhosis in chronic hepatitis C virus infection. J Infect Dis. 2003; 10(4):285-293. doi: 10.1046/j.1365-2893.2003.00436.x

[3] Waljee AK, Higgins PD. Machine learning in medicine: A primer for physicians. Am J Gastroenterol. 2010; 105(6):1224-1226. doi: 10.1038/ajg.2010.173

[4] Konerman MA, Beste LA, Van T, et al. Machine learning models to predict disease progression among veterans with hepatitis C virus. PLoS One. 2019; 14(1):e0208141. doi: 10.1371/journal.pone.0208141

[5] Singal AG, Mukherjee A, Elmunzer BJ, et al. Machine learning algorithms outperform conventional regression models in predicting development of hepatocellular carcinoma. Am J Gastroenterol. 2013; 108(11):1723-1730. doi: 10.1038/ajg.2013.332

[6] Chen YW, Luo J, Dong C, et al. Computer-aided diagnosis and quantification of cirrhotic livers based on morphological analysis and machine learning. Comput Math Methods Med. 2013; 2013:264809. doi: 10.1155/2013/264809

[7] Konerman MA, Zhang Y, Zhu J, et al. Improvement of predictive models of risk of disease progression in chronic hepatitis C by incorporating longitudinal data. Hepatology. 2015; 61:1832-1841. doi: 10.1002/hep.27750

[8] Weerakody PB, Wong KW, Wang G, Ela W. A review of irregular time series data handling with gated recurrent neural networks. Neurocomputing. 2021; 441:161-178. doi: 10.1016/j.neucom.2021.02.046

[9] Veterans Health Administration. Available from: https://www.va.gov/health [Last accessed on 2024 Dec 22].

[10] Corporate Data Warenouse (CDW): US Department of Veterans Affairs. Available from: https://www.hsrd.research.va.gov/for_researchers/vinci/cdw.cfm [Last accessed on 2014 Mar 28].

[11] Oliveira AC, El-Bacha I, Vianna MV, Parise ER. Utility and limitations of APRI and FIB4 to predict staging in a cohort of nonselected outpatients with hepatitis C. Ann Hepatol. 2016; 15(3):326-332. doi: 10.5604/16652681.1198801

[12] Lin ZH, Xin YN, Dong QJ, et al. Performance of the aspartate aminotransferase-to-platelet ratio index for the staging of hepatitis C-related fibrosis: An updated meta-analysis. Hepatology. 2011; 53(3):726-736. doi: 10.1002/hep.24105

[13] Ioannou GN, Tang W, Beste LA, et al. Assessment of a deep learning model to predict hepatocellular carcinoma in patients with hepatitis C cirrhosis. JAMA Netw Open. 2020; 3(9):e2015626. doi: 10.1001/jamanetworkopen.2020.15626

[14] Chung J, Gulcehre C, Cho KH, Bengio Y. Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv. Preprint posted online 2014. doi: 10.48550/arXiv.1412.3555

[15] Kingma DP, Ba J. Adam: A method for stochastic optimization. arXiv.Preprint posted online 2014. doi: 10.48550/arXiv.1412.6980

[16] Srivastava N, Hinton G, Krizhevsky A, Sutskever I, Salakhutdinov R. Dropout: A simple way to prevent neural networks from overfitting. J Mach Learn Res. 2014; 15(1):1929-1958. doi: 10.5555/2627435.2670313

[17] Tang W, Ma J, Waljee AK, Zhu J. Semi-supervised joint learning for longitudinal clinical events classification using neural network models. Stat. 2020; 9:e305. doi: 10.1002/sta4.305

[18] Stekhoven D, Bühlmann P. MissForest - non-parametric missing value imputation for mixed-type data. Bioinformatics. 2012; 28(1):112-118. doi: 10.1093/bioinformatics/btr597

[19] Boyd K, Eng KH, Page CD. Area under the precision-recall curve: Point estimates and confidence intervals. In: Blockeel H, Kersting K, Nijssen S, Železný F, editors. Machine Learning and Knowledge Discovery in Databases. Springer; 2013. p. 451-466. doi: 10.1007/978-3-642-40994-3_29

[20] Pedregosa F, Varoquaux G, Gramfort A, et al. Scikit-learn: Machine learning in Python. J Mach Learn Res. 2011; 12:2825-2830. doi: 10.5555/1953048.2078195

[21] Paszke A, Gross S, Chintala S, et al. Automatic Differentiation in Pytorch. Available from: https://openreview.net/pdf?id=BJJsrmfCZ [Last accessed on 2023 Feb 18].

[22] Graf E, Schmoor C, Sauerbrei W, Schumacher M. Assessment and comparison of prognostic classification schemes for survival data. Stat Med. 1999; 18(17-18):2529-2545. doi: 10.1002/(SICI)1097-0258(19990915/30)18:17/18<2529::AID-SIM274>3.0.CO;2-5

[23] Li J, Chen X, Hovy E, Jurafsky D. Visualizing and understanding neural models in NLP. arXiv. Preprint posted online 2016. doi: 10.48550/arXiv.1506.01066

[24] Becker U, Deis A, Sørensen TI, et al. Prediction of risk of liver disease by alcohol intake, sex, and age: A prospective population study. Hepatology. 1996; 23(5):1025-1029. doi: 10.1002/hep.510230513

[25] Liberal R, Grant C, Mieli-Vergani G, Vergani D. Autoimmune hepatitis: A comprehensive review. J Autoimmun. 2013; 41:126-139. doi: 10.1016/j.jaut.2012.11.002