Prediction of breast cancer relapse time in continuous scale based on type-2 TSK fuzzy model

Recently , microarray analysis and gene expression profiles have been widely applied in diagnosis and classification of different types of cancer such as liver, colon or breast cancer. As the number of breast cancer cases increased dramatically in many contries including Malaysia in recent decades,...

Full description

Saved in:
Bibliographic Details
Main Author: Mahmoudian, Sayed Hamid
Format: Thesis
Language:English
Published: 2010
Online Access:http://psasir.upm.edu.my/id/eprint/41139/1/FK%202010%2075R.pdf
http://psasir.upm.edu.my/id/eprint/41139/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Recently , microarray analysis and gene expression profiles have been widely applied in diagnosis and classification of different types of cancer such as liver, colon or breast cancer. As the number of breast cancer cases increased dramatically in many contries including Malaysia in recent decades, different types of studies have been done to control the disease or reduce the cost of their treatments. Gene expression profiles, which can screen the behavior of a large number of genes simultaneously, have been used in some studies to extract the significant genes related to breast cancer. Tumor classification, Estrogen Receptor status recognition or survival analysis has been usually considered as important objectives in these studies. Due to the fact that studies in survival analysis of breast cancer can reduce the cost of treatments and side effects of the adjuvant therapy, different methods for predicting the outcome of the disease have been proposed by previous researcher. The two major objectives of this research are to propose a fuzzy classifier to discriminate breast cancer tumors into two classes, which are high risk and low risk by some interpretable rules similar to linguistic words, and to predict the relapse time of breast cancer by TSK fuzzy models in continuous scale. For this reason, breast cancer dataset has been applied for training the models and two other independent samples have been used for validating the results. In addition, K-fold Cross Validation, B632 and B632+ methods have been used for error estimation. In the first objective of the thesis, a lemma has been proven and a new hybrid algorithm based on Fuzzy Association Rule Mining has been proposed to gather some selected genes and generate fuzzy rules for classification. In the second one, a method for generating the fuzzy rules to discriminate the samples of breast cancer into the different groups have been proposed and applied to predict the relapse time of samples in continuous scale while handling the uncertainties in linguistic terms of the rules. The relapse time of two available independent samples of breast cancer have been predicted by the model and the results show the superiority of the proposed model with respect to the previous study. Finally 46 significant genes and 16 fuzzy rules have been introduced which can be used in a Type-2 TSK fuzzy model as a predictor.