Assessment of machine learning in radiomics for predicting the risk of clinically significant pancreatic fistulas after pancreatoduodenal resections using CT imaging

Despite the reduction in mortality after pancreatoduodenal resections, the incidence of postoperative complications remains high (20–60%). One of the most severe complications is postoperative pancreatic fistula. Predicting the risks of a complicated postoperative period remains an urgent problem. One of the key risk factors is the pancreatic structure, and modern approaches to predicting clinically relevant pancreatic fistula integrate not only morphological but also radiomic parameters of CT images using artificial intelligence methods.

Aim. To evaluate the potential of machine learning in radiomics for predicting clinically relevant pancreatic fistulas after pancreatoduodenal resection and to develop a clinical decision support system based on the “Virtual Biopsy” platform.

Methods. Retrospective analysis of data from 117 patients who underwent pancreatoduodenal resection (2016–2019) at the A.V. Vishnevsky National Medical Research Center of Surgery. Machine learning methods were applied to assess textural features of preoperative CT scans.

Results. Сlinically relevant pancreatic fistulas were recorded in 31 patients (26.5%). Clinically significant fistula alone was diagnosed in 11 patients (9.4%), while its combination with arrosive bleeding was observed in 20 patients (17.1%). The peak incidence of fistulas occurred on days 4–6, and of bleeding on days 8–14. The radiomic AdaBoost model demonstrated the highest efficacy (ROC AUC = 0.815), outperforming alternative approaches: Gradient Boosting (0.631), XGBoost (0.677), LightGBM (0.631), and Stacking (0.662). Integration of morphological features did not improve predictive capability, likely due to data noise. Models based on semantic parameters (max. ROC AUC = 0.653) confirmed limited clinical applicability.

Conclusion. Machine learning methods are effective in predicting clinically relevant pancreatic fistulas after pancreatoduodenal resections. Radiomic analysis extends the diagnostic potential of CT, demonstrating superior model accuracy metrics compared to classical semantic features alone.

1. Patyutko Yu.I., Kotelnikov A.G., Polyakov A.N., Podluzhnyi V.D. Evolution of surgery for pancreatic head and periampullary cancer. Annals of HPB Surgery. 2019; 24 (3): 45–53. http://doi.org/10.16931/1995-5464.2019345-53 (In Russian)

2. Angileri S.A., Gorga G., Tortora S. et al. Biliary injuries after pancreatic surgery: interventional radiology management. Gland Surg. 2019; 8 (2): 141–149. http://doi.org/10.21037/gs.2019.01.05

3. Kapoor V.K. Complications of pancreatoduodenectomy. Rozhl. Chir. 2016; 95 (2): 53–59.

4. Tonolini M., Ierardi A.M., Carrafiello G. Elucidating early CT after pancreatico-duodenectomy: a primer for radiologists. Insights Imaging. 2018; 9 (4): 425–436. http://doi.org/10.1007/s13244-018-0616-3

5. Chincarini M., Zamboni G.A., Pozzi Mucelli R. Major pancreatic resections: normal postoperative findings and complications. Insights. Imaging. 2018; 9 (2): 173–187. http://doi.org/10.1007/s13244-018-0595-4

6. Kabanov M.Yu., Sementsov K.V., Yakovleva D.M., Alekseev V.V. Current issues of pancreatic fistula development in patients after pancreaticoduodenectomy. Bulletin of the Pirogov National Medical and Surgical Center. 2017; 12 (2): 112–116. http://doi.org/10.25881/BPNMSC.2017.12.2.112.116 (In Russian)

7. Rayn V.Yu. Biliary fistula after pancreaticoduodenectomy. Surgery News. 2022; 30 (1), 95–101. http://doi.org/10.16931/1995-5464.2022-30-1-95-101 (In Russian)

8. Lowy A.M., Lee J.E., Pisters P.W. et al. Prospective, randomized trial of octreotide to prevent pancreatic fistula after pancreaticoduodenectomy for malignant disease. Ann. Surg. 1997; 226 (5): 632–641. http://doi.org/10.1097/00000658-199711000-00008

9. Braga M., Capretti G., Pecorelli N. et al. A prognostic score to predict major complications after pancreaticoduodenectomy. Ann. Surg. 2011; 254 (5): 702–707. http://doi.org/10.1097/SLA.0b013e31823598fb

10. Bassi C., Marchegiani G., Dervenis C. et al. The 2016 update of the International Study Group (ISGPS) definition and grading of postoperative pancreatic fistula: 11 years after. Surgery. 2017; 161 (3): 584–591. http://doi.org/10.1016/j.surg.2016.11.014

11. Fu S.J., Shen S.L., Li S.Q. et al. Risk factors and outcomes of postoperative pancreatic fistula after pancreatico-duodenectomy: an audit of 532 consecutive cases. BMC Surg. 2015; 15: 117. http://doi.org/10.1186/s12893-015-0011-7

12. Hallet J., Zih F.S., Deobald R.G. et al. The impact of pancreaticojejunostomy versus pancreaticogastrostomy reconstruction on pancreatic fistula after pancreaticoduodenectomy: meta-analysis of randomized controlled trials. HPB. 2015; 17 (2): 113–120. http://doi.org/10.1111/hpb.12299

13. Casadei R., Ricci C., Taffurelli G. et al. Are there preoperative factors related to a “soft pancreas” and are they predictive of pancreatic fistulas after pancreatic resection? Surg. Today. 2015; 45 (6): 708–714. http://doi.org/10.1007/s00595-014-1045-7

14. Galchina Y.S., Kondratyev E.V., Karmazanovsky G.G. et al. CT predictors of postoperative specific complications in pancreatic resections. Russian Electronic Journal of Radiology. 2020; 10 (2): 102–115. http://doi.org/10.21569/2222-7415-2020-10-2-102-115 (In Russian)

15. Galchina Yu.S., Karmazanovsky G.G., Kondratyev E.V. et al. Contrast-enhanced computed tomography in evaluation of pancreatic stump volume in pancreaticoduodenectomy. Pirogov Russian Journal of Surgery = Khirurgiya. Zhurnal imeni N.I. Pirogova. 2020; 6: 31–37. http://doi.org/10.17116/hirurgia202006131 (In Russian)

16. Galchina Y.S., Karmazanovsky G.G., Kalinin D.V. et al. Diagnostic criteria for a “soft” pancreas and their influence on the occurrence of pancreatic fistula after pancreatoduodenectomy. Annals of HPB Surgery. 2020; 25 (2): 113–123. http://doi.org/10.16931/1995-5464.20202113-123 (In Russian)

17. Lee S.E., Jang J.Y., Lim C.S. et al. Measurement of pancreatic fat by magnetic resonance imaging: predicting the occurrence of pancreatic fistula after pancreatoduodenectomy. Ann. Surg. 2010; 251 (5): 932–936. http://doi.org/10.1097/SLA.0b013e3181d65483

18. Watanabe H., Kanematsu M., Tanaka K. et al. Fibrosis and postoperative fistula of the pancreas: correlation with MR imaging findings-preliminary results. Radiology. 2014; 270 (3): 791–799. http://doi.org/10.1148/radiol.13131194

19. Räty S., Sand J., Lantto E. et al. Postoperative acute pancreatitis as a major determinant of postoperative delayed gastric emptying after pancreaticoduodenectomy. J. Gastrointest. Surg. 2006; 10 (8): 1131–1139. http://doi.org/10.1016/j.gassur.2006.05.012

20. El Nakeeb A., Hamdy E., Sultan A.M. et al. Isolated Roux loop pancreaticojejunostomy versus pancreaticogastrostomy after pancreaticoduodenectomy: a prospective randomized study. HPB (Oxford). 2014; 16 (8): 713–722. http://doi.org/10.1111/hpb.12210

21. Deng Y., Zhao B., Yang M. et al. Association between the incidence of pancreatic fistula after pancreaticoduodenectomy and the degree of pancreatic fibrosis. J. Gastrointest. Surg. 2018; 22 (3): 438–443. http://doi.org/10.1007/s11605-017-3660-2

22. Kim S.Y., Kim H., Cho J.Y. et al. Quantitative assessment of pancreatic fat by using unenhanced CT: pathologic correlation and clinical implications. Radiology. 2014; 271 (1): 104–112. http://doi.org/10.1148/radiol.13122883

23. Roberts K.J., Storey R., Hodson J. et al. Pre-operative prediction of pancreatic fistula: is it possible? Pancreatology. 2013; 13 (4): 423–428. http://doi.org/10.1016/j.pan.2013.04.322

24. Hashimoto Y., Sclabas G.M., Takahashi N. et al. Dual-phase computed tomography for assessment of pancreatic fibrosis and anastomotic failure risk following pancreatoduodenectomy. J. Gastrointest. Surg. 2011; 15 (12): 2193–2204. http://doi.org/10.1007/s11605-011-1687-3

25. Kondratyev E.V., Shmeleva S.A., Ustalov A.A. et al. Theoretical basics of abdominal CT radiomics: a review. Diagnostic Radiology and Radiotherapy. 2025; 16 (1): 33–46. http://doi.org/10.22328/2079-5343-2025-16-1-33-46 (In Russian)

26. Vasilev Y.A., Vladzimirskyy A.V., Omelyanskaya O.V. et al. Review of meta-analyses on the use of artificial intelligence in radiology. Medical Visualization. 2024; 28 (3): 22–41. http://doi.org/10.24835/1607-0763-1425 (In Russian)

27. Gelezhe P.B., Blokhin I.A., Semenov S.S., Caruso D. Magnetic resonance imaging radiomics in prostate cancer radiology: what is currently known? Digital Diagnostics. 2021; 2 (4): 441–452. http://doi.org/10.17816/DD70170 (In Russian)

28. Gelezhe P.B., Blokhin I.A., Semenov S.S. Radiomics for the diagnosis and treatment of prostate cancer. Medical Physics. 2022; 1: 21. (In Russian)

29. Kambakamba P., Mannil M., Herrera P.E. et al. The potential of machine learning to predict postoperative pancreatic fistula based on preoperative, non-contrast-enhanced CT: a proof-of-principle study. Surgery. 2020; 167 (2): 448–454. http://doi.org/10.1016/j.surg.2019.09.019

30. Lin Z., Tang B., Cai J. et al. Preoperative prediction of clinically relevant postoperative pancreatic fistula after pancreaticoduodenectomy. Eur. J. Radiol. 2021; 139: 109693. http://doi.org/10.1016/j.ejrad.2021.109693

31. Suvorov V.A., Panin S.I., Kovalenko N.V. et al. Prediction of pancreatic fistula after pancreatoduodenectomy using machine learning. Siberian Journal of Oncology. 2023; 22 (6): 25–34. http://doi.org/10.21294/1814-4861-2023-22-6-25-34 (In Russian)

32. Van Griethuysen J.J.M., Fedorov A., Parmar C. et al. Computational radiomics system to decode the radiographic phenotype. Cancer Res. 2017; 77 (21): e104–e107. http://doi.org/10.1158/0008-5472.CAN-17-0339

33. Fedorov A., Beichel R., Kalpathy-Cramer J. et al. 3D Slicer as an image computing platform for the quantitative imaging network. Magn. Reson. Imaging. 2012; 30 (9): 1323–1341. http://doi.org/10.1016/j.mri.2012.05.001

34. Perri G., Marchegiani G., Partelli S. et al. Preoperative risk stratification of postoperative pancreatic fistula: a risk-tree predictive model for pancreatoduodenectomy. Surgery. 2021; 170 (6): 1596–1601. http://doi.org/10.1016/j.surg.2021.06.046

35. Shi Y., Gao F., Qi Y. et al. Computed tomography-adjusted fistula risk score for predicting clinically relevant postoperative pancreatic fistula after pancreatoduodenectomy: training and external validation of model upgrade. EBioMedicine. 2020; 62: 103096. http://doi.org/10.1016/j.ebiom.2020.103096