Volume 13, Issue 3 (May-June 2014)
Payesh 2014, 13(3): 285-291 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Mahmood Reza Gohari, Parisa Mokhtari, Mohammad Amin Pourhoseingholi, Akbar Biglarian. Artificial Neural Network in survival analysis of gastric cancer patients. Payesh 2014; 13 (3) :285-291
URL: http://payeshjournal.ir/article-1-300-en.html
URL: http://payeshjournal.ir/article-1-300-en.html
1- Department of Biostatistics, Hospital management Research Center, Tehran University of Medical Sciences (TUMS), Tehran, Iran
2- Department of Biostatistics, Tehran University of Medical Sciences (TUMS), Tehran, Iran
3- Department of Health System Research, Research Center of Gastroenterology and Liver Disease, Shahid Beheshti university of Medical Sciences, Tehran, Iran
4- Department of Biostatistics, University of Social Welfare and Rehabilitation Sciences (USWRS), Tehran, Iran
2- Department of Biostatistics, Tehran University of Medical Sciences (TUMS), Tehran, Iran
3- Department of Health System Research, Research Center of Gastroenterology and Liver Disease, Shahid Beheshti university of Medical Sciences, Tehran, Iran
4- Department of Biostatistics, University of Social Welfare and Rehabilitation Sciences (USWRS), Tehran, Iran
Abstract: (5153 Views)
Objective (s): Gastric cancer (GC) is the most prevalent cancer in men and third for women. The aim of this study was to analyze the survival of GC patients and then the prediction of death in these patients using a neural network model.
Methods: This was a historical cohort study with data selected from 232 registered gastric cancer patients who underwent surgery between 2002 and 2007 at a referral center for gastrointestinal cancers, in Tehran, Iran. Prognostic factors of survival using Kaplan-Meier and Cox proportional hazards models were estimated. To analyze the data using the neural network, the data were divided randomly into two groups, training and testing sets, and homogeneous survival times were tested using log rank test. To select the best network structure, the characteristics of network models modified and the best model was selected using ROC analysis. The analysis was performed using SPSS version 19.
Results: The median survival time of GC patients was 25.1 months. One and three year survival times of GC patients were 0.769 (CI 95%: 0.711, 0.833) and 0.256 (CI 95%: 0.185, 0.373), respectively. The prediction error was 37.7% and the area under the ROC was 0.732 at final model. True prediction for censorship and death were 71.4% and 50%, respectively.
Conclusion: Neural network is an appropriate approach for predicting the survival of GC patients and its predictions can be used to classify these patients.
Methods: This was a historical cohort study with data selected from 232 registered gastric cancer patients who underwent surgery between 2002 and 2007 at a referral center for gastrointestinal cancers, in Tehran, Iran. Prognostic factors of survival using Kaplan-Meier and Cox proportional hazards models were estimated. To analyze the data using the neural network, the data were divided randomly into two groups, training and testing sets, and homogeneous survival times were tested using log rank test. To select the best network structure, the characteristics of network models modified and the best model was selected using ROC analysis. The analysis was performed using SPSS version 19.
Results: The median survival time of GC patients was 25.1 months. One and three year survival times of GC patients were 0.769 (CI 95%: 0.711, 0.833) and 0.256 (CI 95%: 0.185, 0.373), respectively. The prediction error was 37.7% and the area under the ROC was 0.732 at final model. True prediction for censorship and death were 71.4% and 50%, respectively.
Conclusion: Neural network is an appropriate approach for predicting the survival of GC patients and its predictions can be used to classify these patients.
type of study: Descriptive |
Accepted: 2018/11/28 | ePublished ahead of print: 2014/04/7 | Published: 2014/05/15
Accepted: 2018/11/28 | ePublished ahead of print: 2014/04/7 | Published: 2014/05/15
| Rights and Permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |