Computational protocol: Assessing the Effect of Quantitative and Qualitative Predictors on Gastric Cancer Individuals Survival Using Hierarchical Artificial Neural Network Models

Similar protocols

Protocol publication

[…] In this study 330 patients with gastric cancer who had history of surgery at the Iran Cancer Institute over a 5-year period were enrolled and their postoperative life expectancy was determined. Those who were alive at the end of the study, and individuals who had missing data from the time they were lost to follow up, were excluded. During the study period, 239 patients died; the causes of which were not disease-related (in 13 cases) were considered as right censored from the date of death. Firstly, we randomly divided cases into two groups of 165 patients each to examine models and assess reproducibility; the training group and the study group. To ensure group similarity and random allocation, the chi-square test and multiple logistic regression analysis were used. Subgroups were compared using the log-rank test. Independent variables of age (continuous quantitative), gender (dichotomous: woman = 0, man = 1), history of substance abuse (dichotomous: no = 0, yes = 1), cancer site (trinomial: anterior = 2, cardia = 1, other = 0), type of pathology (dichotomous: adenocarcinoma = 1, other = 0), presence of metastasis (dichotomous: no = 0, yes = 1), T-stage (ordinal: 1-4), N-stage (ordinal: 0-3), M-stage (ordinal: 0-1), number of complementary treatments (discrete quantitative), and computed probabilities of survival by time were used. Subsequently, we calculated probabilities of survival past 6, 12, 18, 24, 36, 48, and 60 months using the Cox proportional hazards model and a NN. For this purpose, estimates and coefficients of the Cox model and the weights of the NN model (with 3, 5, and 7 nodes in the hidden layer) were determined in the training group and used to derive predictions in the study group. Predictions with these two methods were compared with those of the Kaplan-Meier product limit estimator as the gold standard. Comparisons were made with the Friedman and Kruskal-Wallis tests. Staging was done using the 6th edition of the TNM system. For statistical analyses, SPSS version 11, Matlab version 7.2, Statistica version 6.0, and S_PLUS 2000 were used and the level of significance was considered 0.05. […]

Pipeline specifications

Software tools SPSS, Statistica
Application Miscellaneous
Organisms Homo sapiens
Diseases Neoplasms, Stomach Neoplasms, Substance-Related Disorders