ARTFICIAL INTELLIGENCE, SYSTEM ANALYSIS AND SIMULATION MODELING IN OPTIMIZATION OF TREATMENT FOR ESOPHAGEAL CANCER PATIENTS AFTER COMPLETE ESOPHAGECTOMIES

1. ARTFICIAL INTELLIGENCE, SYSTEM ANALYSIS AND SIMULATION MODELING IN OPTIMIZATION OF TREATMENT FOR ESOPHAGEAL CANCER PATIENTS AFTER COMPLETE ESOPHAGECTOMIES Oleg Kshivets, MD, PhD Department of Surgery, Siauliai Public Hospital & Cancer Center, Lithuania

2. Abstract: ARTFICIAL INTELLIGENCE, SYSTEM ANALYSIS AND SIMULATION MODELING IN OPTIMIZATION OF TREATMENT FOR ESOPHAGEAL CANCER PATIENTS AFTER COMPLETE ESOPHAGECTOMIES Oleg Kshivets Department of Surgery, Siauliai Public Hospital, Siauliai, Lithuania OBJECTIVE: The search of optimal treatment plan for esophageal cancer (EC) patients (ECP) with stage T1-4N1M1a was realized. We examined the clinicomorphologic factors associated with the low- and high-risk of generalization of ECP after complete en block (R0) esophagectomies (E) through left and right thoracoabdominal incision. METHODS: We analyzed data of 126 consecutive ECP (age=56.8±7.9 years; tumor size=5.4±2.5 cm) radically operated and monitored in 1975-2007 (males=98, females=28; E Ivor-Lewis=89, E Garlock=37; combined E with resection of diaphragm, pericardium, lung, liver, etc=40; lymphadenectomy D2=59, D3=67; adenocarcinoma=93, squamos=31, mix=2; T1=25, T2=38, T3=29, T4=34; N0=55, N1=23, M1a=48; M1b=0; G1=46, G2=39, G3=41; stage I=22, stage IIA=24, stage IIB=13, stage III=19, stage IVA=48; only surgery-S=97, adjuvant chemoimmunoradiotherapy-AT=29: 5-FU + thymalin/taktivin + radiotherapy 45-50Gy). Variables selected for 5-year survival (5YS) study were input levels of 45 blood parameters, sex, age, TNMG, cell type, tumor size. Survival curves were estimated by the Kaplan-Meier method. Differences in curves between groups of ECP were evaluated using a log-rank test. Multivariate Cox modeling, multi-factor clustering, discriminant analysis, structural equation modeling, Monte Carlo, bootstrap simulation and neural networks computing were used to determine any significant dependence. RESULTS: General cumulative 5YS was 50.5%, 10-year survival – 38.3%. 64 ECP (50.8%) were alive, 39 ECP (31%) lived more than 5 years (life span: LS=3544.3±1712.5 days) and 17 ECP - 10 years (LS=5000.1±1639 days) without any features of EC progressing. 55 ECP (43.7%) died because of LC during first 5 years after surgery (LS=621.4±366 days). AT significantly improved ECP 5YS after E (P=0.023 by log-rank test). Cox modeling displayed that 5YS of ECP after complete E significantly depended on: T, N, histology, stage, combined procedures, AT, age, blood cell subpopulations (P=0.000-0.039). Neural networks computing, genetic algorithm selection and bootstrap simulation revealed relationships between 5YS of ECP and N (rank=1), sex (2), EC growth (3), T (4), histology (5), combined procedures (6), G (7), blood residual nitrogen (8), hemorrhage time (9), blood chlorides (10), AT (11), stick neutrophils (12), tumor size (13), thrombocytes (14), monocytes (15). Correct prediction of ECP survival after radical procedures was 100% by neural networks computing (area under ROC curve=1.0; error=0.001). CONCLUSIONS: Optimal treatment strategies for ECP are: 1) screening and early detection of EC; 2) availability of very experienced surgeons because of complexity radical procedures; 3) aggressive en block surgery for completeness; 4) precise prediction; 5) AT for ECP with unfavorable prognosis.

7. Left Thoracoabdominal Esophagectomy (Garlock)…………… 37

8. Garlock Procedures…. 37

9. Right Thoracoabdominal Esophagectomy (Ivor Lewis)... 89

10. Ivor Lewis Procedures…… 89

11. One-Stage Esophagogastroplasty Intrapleural Anastomosis………….. 61 Neck Anastomosis…………………... 65

14. General E sophageal Cancer Patients Survival after Complete Esophagectomies (Kaplan-Meier) ( n=1 26)

15. Results of Univariate Analysis in Prediction of E sophageal Cancer Patients Survival ( n=1 26)

16. Results of Univariate Analysis in Prediction of E sophageal Cancer Patients Survival ( n=1 26)

17. Results of Univariate Analysis in Prediction of E sophageal Cancer Patients Survival ( n=1 26)

18. Results of Cox Regression Modeling in Prediction of Esophageal Cancer Patients Survival after Complete Esophagectomies (n=126 )

19. Results of Cox Regression Modeling in Prediction of Esophageal Cancer Patients Survival after Complete Esophagectomies (n=126 )

20. Neural Networks in Prediction of Esophageal Cancer Patients Survival after Complete Esophagectomies (n=126 )

21. Results of Neural Networks Computing in Prediction of Esophageal Cancer Patients Survival after Complete Esophagectomies (n=126 )

22. Results of Bootstrap Simulation in Prediction of Esophageal Cancer Patients Survival after Complete Esophagectomies (n=126 )

23. Holling-Tenner Models of Esophageal Cancer Cell Population and Cytotoxic Cell Population Dynamics

24. Esophageal Cancer Dynamics

25. Results of Kohonen Self-Organizing Neural Networks Computing in Prediction of Esophageal Cancer Patients Survival after Complete Esophagectomies (n=126)

26. Networks between Clinicopathologic, Biochemic, Hemostasis & Hematologic Data and 5-Year Survival of Esophageal Cancer Patients after Esophagectomies (n=94)

27. Prognostic SEPATH-Model of Esophageal Cancer Patients Survival after Complete Esophagectomies, n=126