Analysis of survival data of patient with breast cancer using artificial neural network and cox regression models

Sedehi, Morteza. and Amani, Firouz. and MOMENI DEHAGHI, Fatemeh. (2014) Analysis of survival data of patient with breast cancer using artificial neural network and cox regression models. JOURNAL OF ZABOL UNIVERSITY OF MEDICAL SCIENCES AND HEALTH SERVICES (JOURNAL OF ROSTAMINEH), 5 (4).

Full text not available from this repository.

Abstract

Introduction: In analyzing survival data using conventional methods of classical statistics requires some basic assumptions for data. Artificial neural networks as a modern modeling method can be used in situations where classic models have restricted application because their assumptions are not met. This study is compared survival of patients with breast cancer using artificial neural network and Cox regression models. Methods: This historical cohort study, include data from 161 patients with breast cancer in Ardabil province in the years 2002-2007 were diagnosed as having cancer. 68.9% of data dividing as training data set and 31.1% of data dividing as validation data set. Artificial neural networks and Cox regression models are fitted to data. Predictive accuracy and area under ROC used to compare models. Results: Between neural network models, models with SCG, OSS and LM learning algorithms with predictive accuracy of 94, 90 and 78 percent for validation data, had the highest efficiency respectively. Areas under ROC for these models are 0.991, 0.972 and 0.837 respectively and 0.869 for Cox regression model. Conclusion: This study shaw that if suitable architecture and algorithms are selected for artificial neural network model, this model will be more efficient than the Cox regression model to predict the survival situation of patients with breast cancer.

Item Type: Article
Uncontrolled Keywords: ARTIFICIAL NEURAL NETWORKS, BREAST CANCER, COX REGRESSION MODEL, SURVIVAL ANALYSIS
Subjects: QT physiology > physics.mathematics.engineering
Divisions: Faculty of Health > Department of Epidemiology
Depositing User: Users 1 not found.
Date Deposited: 03 Dec 2017 04:39
Last Modified: 24 Feb 2018 05:59
URI: http://eprints.skums.ac.ir/id/eprint/6497

Actions (login required)

View Item View Item