Optimization of selection criteria for liver transplantation patients with hepatocellular carcinoma using machine learning models

• Voskanyan S.E.
• Rudakov V.S.
• Sushkov A.I.
• Popov M.V.
• Bashkov A.N.
• Gubarev K.K.
• Artemiev A.I.
• Matkevich E.I.
• Grigorieva O.O.
• Lishchuk S.V.
• Kalachyan A.E.

Abstract

Background. The application of machine learning methods in medicine, particularly in transplantology, has significantly expanded over the past decade, offering new opportunities to improve patient selection criteria for hepatocellular carcinoma and enhance liver transplantation outcomes.

The aim of the study is to develop transplantation criteria for patients with hepatocellular carcinoma based on machine learning models.

Material and methods. The study includes data from 69 patients with hepatocellular carcinoma against the background of liver cirrhosis, who underwent liver transplantation at the Burnazyan Federal Medical Biophysical Center (FMBA of Russia) between 2010 and 2022. The assessment included levels of alpha-fetoprotein, maximum tumor nodule size, number of tumors, and presence of vascular invasion. Machine learning models, combined using a stacking method, were used to predict HCC recurrence: Model A (radiological data) and Model B (histological data). These models were compared with the “Milan Criteria,” “California Criteria,” and the “5-5-500 Rule” in terms of the proportion of patients meeting the criteria, sensitivity, specificity, F1-score, C-index, and survival rates.

Results. For radiological data, “Model A” showed the following results: proportion of patients meeting the criteria – 65%, sensitivity – 72%, specificity – 94%, F1-score – 0.81, C-index – 0.83, 5-year overall survival – 78%, recurrence-free survival – 85%. For the “Milan Criteria,” the values were: 36; 89; 65; 0.79; 0.77; 78; 91%. For the “California Criteria”: 55; 67; 71; 0.69; 0.69; 67; 81%, and for the “5-5-500 Rule” – 60; 78; 88; 0.82; 0.83; 79; 86% respectively. For histological data, “Model B” demonstrated: proportion of patients – 67%, sensitivity – 78%, specificity – 94%, F1-score – 0.85, C-index – 0.86, 5-year overall survival – 75%, recurrence-free survival – 84%. Corresponding indicators for the “Milan Criteria” were: 58; 61; 77; 0,67; 0,69; 68; 73%; for the “California Criteria”: 70; 56; 82; 0.65; 0,69; 69; 73% and for the “5-5-500 Rule” – 67; 72; 88; 0.79; 0.80; 72; 81%.

Conclusion. The developed machine learning models demonstrated high efficiency in predicting hepatocellular carcinoma recurrence, comparable to, and in some parameters exceeding, widely accepted clinical criteria. Their application, improvement, and validation on broader data can significantly enhance the accuracy of predictions and optimize treatment.

Keywords: hepatocellular carcinoma; liver transplantation; hepatocellular carcinoma recurrence; transplantation criteria; machine learning

References

1. Federal State Statistics Service (Rosstat). Healthcare in Russia. Statistical report. 2021.
2. Wang J., Bao J., Wang R., Hong J., Zhang L., Que Q., et al. The predictive value of the modified AFP model for liver transplantation outcomes in multinodular hepatocellular carcinoma patients. World J Surg Oncol. 2023; 21: 104. DOI: https://doi.org/10.1186/s12957-023-02994-y
3. To J., Ghosh S., Zhao X., Pasini E., Fischer S., Sapisochin G., et al. Deep learning-based pathway-centric approach to characterize recurrent hepatocellular carcinoma after liver transplantation. Hum Genomics. 2024; 18: 58. DOI: https://doi.org/10.1186/s40246-024-00624-6
4. Granov A.M., Granov D.A., Zherebtsov F.K., Gerasimova O.A., Borovik V.V., Osovskikh V.V., et al. Liver transplantation. A single center experience of 100 cases. Vestnik transplantologii i iskusstvennykh organov [Bulletin of Transplantology and Artificial Organs]. 2012; 14 (4): 11–6. DOI: https://doi.org/10.15825/1995-1191-2012-4-11-16 (in Russian)
5. Voskanyan S.A., Sushkov A.I., Artem’ev A.I., Zabezhinsky D.A., Naydenov E.V., Bashkov A.N., et al. Salvage liver transplantation for hepatocellular carcinoma. Khirurgiya. Zhurnal imeni N.I. Pirogova [Surgery. The Journal named after I. Pirogov]. 2019; (10): 21–8. DOI: https://doi.org/10.17116/hirurgia201910121 (in Russian)
6. Olisov O.D., Novruzbekov M.S., Gulyaev V.A., Lutsyk K.N. The role of calcineurin inhibitors in the progression of hepatocellular carcinoma after liver transplantation. Transplantologiya [Transplantology]. 2022; 14 (3): 292–300. DOI: https://doi.org/10.23873/2074-0506-2022-14-3-292-300 (in Russian)
7. Voskanyan S.E., Rudakov V.S., Sushkov A.I., Popov M.V., Bashkov A.N., Gubarev K.K., et al. New prognostic model for liver transplantation outcomes in hepatocellular carcinoma. Transplantologiya [Transplantology]. 2024; 16 (3): 278–90. DOI: https://doi.org/10.23873/2074-0506-2024-16-3-278-290 (in Russian)
8. Liu X., Liu Y., Zhang W., Hong Y., Meng J., Wang J., et al. Deep learning for prediction of hepatocellular carcinoma recurrence after resection or liver transplantation: a discovery and validation study. Hepatol Int. 2022; 16: 577–89. DOI: https://doi.org/10.1007/s12072-022-10321-y
9. Kwong A., Hameed B., Syed S., Ho R., Mard H., Arshad S., et al. Machine learning to predict waitlist dropout among liver transplant candidates with hepatocellular carcinoma. Cancer Med. 2022; 11: 1535–41. DOI: https://doi.org/10.1002/cam4.4538
10. Tran V.G., Moris D., Markovic D., Zaribafzadeh H., Henao R., Lai Q., et al. Development and validation of a REcurrent Liver cAncer Prediction ScorE (RELAPSE) following liver transplantation in patients with hepatocellular carcinoma: analysis of the US multicenter HCC transplant consortium. Liver Transpl. 2023; 29 (7): 683–97. DOI: https://doi.org/10.1097/LVT.0000000000000145
11. Gotlieb N., Azhie A., Sharma D., Spann A., Suo N.-J., Tran J., et al. The promise of machine learning applications in solid organ transplantation. NPJ Digit Med. 2022; 5 (1): 89. DOI: https://doi.org/10.1038/s41746-022-00637-2
12. Wolpert D.H. Stacked generalization. Neural Networks. 1992; 5 (2): 241–59.
13. Xu X., Mo Q., Wang Z., Zhao Y., Li C. Classification of sleeping position using enhanced stacking ensemble learning. Entropy. 2024; 26: 817. DOI: https://doi.org/10.3390/e26100817
14. Peduzzi P., Concato J., Kemper E., Holford T.R., Feinstein A.R. A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol. 1996; 49 (12): 1373–9. DOI: https://doi.org/10.1016/s0895-4356(96)00236-3
15. Vittinghoff E., McCulloch C.E. Relaxing the rule of ten events per variable in logistic and Cox regression. Am J Epidemiol. 2007; 165: 710–8. DOI: https://doi.org/10.1093/aje/kwk052
16. van Smeden M., de Groot J.A.H., Moons K.G.M., Collins G.S., Altman D.G., Eijkemans M.J.C., et al. No rationale for 1 variable per 10 events criterion for binary logistic regression analysis. BMC Med Res Methodol. 2016; 16: 163. DOI: https://doi.org/10.1186/s12874-016-0267-3
17. Pavlou M., Ambler G., Qu C., Seaman S.R., White I.R., Omar R.Z. An evaluation of sample size requirements for developing risk prediction models with binary outcomes. BMC Med Res Methodol. 2024; 24: 146. DOI: https://doi.org/10.1186/s12874-024-02268-5
18. Schima W., Heiken J.P. LI-RADS v2017 for liver nodules: how we read and report. Cancer Imaging. 2018; 18: 14. DOI: https://doi.org/10.1186/s40644-018-0149-5
19. Hong C.W., Chernyak V., Choi J.-Y., Lee S., Potu C., Delgado T., et al. A multicenter assessment of interreader reliability of LI-RADS version 2018 for MRI and CT. Radiology. 2023; 308 (1): e239018. DOI: https://doi.org/10.1148/radiol.239018
20. Wang K., Liu Y., Chen H., Yu W., Zhou J., Wang X. Fully automating LI-RADS on MRI with deep learning-guided lesion segmentation, feature characterization, and score inference. Front Oncol. 2023; 13. DOI: https://doi.org/10.3389/fonc.2023.1153241
21. Bhatti A.B.H., Naqvi W., Khan N.Y., Zia H.H., Dar F.S., Khan Z.A., et al. Living donor liver transplantation for advanced hepatocellular carcinoma including macrovascular invasion. J Cancer Res Clin Oncol. 2022; 148: 245–53. DOI: https://doi.org/10.1007/s00432-021-03665-9
22. Kim S.-C., Kim D.-W., Cho E.J., Lee J.-Y., Kim J., Kwon C., et al. A circulating cell-free DNA methylation signature for the detection of hepatocellular carcinoma. Mol Cancer. 2023; 22: 164. DOI: https://doi.org/10.1186/s12943-023-01872-1
23. Hao Y., Yang Q., He Q., Hu H., Weng Z., Su Z., et al. Identification of DNA methylation signatures for hepatocellular carcinoma detection and microvascular invasion prediction. Eur J Med Res. 2022; 27: 276. DOI: https://doi.org/10.1186/s40001-022-00910-w
24. Ding T., Li X., Mo J., Alexander G., Li J. Recurrence risk stratification of hepatocellular carcinomas based on immune gene expression and features extracted from pathological images. PLoS Comput Biol. 2023; 19: e1011716. DOI: https://doi.org/10.1371/journal.pcbi.1011716

All articles in our journal are distributed under the Creative Commons Attribution 4.0 International License (CC BY 4.0 license)