Make Intelligent of Gastric Cancer Diagnosis Error in Qazvin's Medical Centers: Using Data Mining Method.

Asghar Mortezagholi,Omid Khosravizadeh,Rohollah Kalhor,Younes Shafigh,Mohammad Bagher Menhaj

doi:10.31557/apjcp.2019.20.9.2607

Abstract

Objective:Gastric cancer is one of the most common types of cancers, which will result in irreparable harm in the case of misdiagnosis or late diagnosis. The purpose of this study is to investigate the capability of data mining techniques and disease risk factor characteristics to predict and diagnose the gastric cancer. Methods:In this retrospective descriptive-analytic study, we selected 405 samples from two groups of patient and healthy participants. A total of 11 characteristics and risk factors were examined. we used four Machine learning methods, Include support vector machine (SVM), decision tree (DT), naive Bayesian model, and k nearest neighborhood (KNN) to classify the patients with gastric cancer. The evaluation criteria to investigate the model on the database of patients with gastric cancer included Recall, Precision, F-score, and Accuracy. Data was analyzed using MATLAB® software, version 3.2 (Mathworks Inc., Natick, MA, USA). Results:Based on the results achieved from the evaluation of four methods, the accuracy rates of SVM, DT, naive Bayesian model, and KNN algorithms were 90.08, 87.89, 87.60, and 87.60 percent, respectively. The findings showed that the highest level of F-Score was related to the SVM (91.99); whereas, the lowest rate was associated with the KNN algorithm (87.17). Conclusion:According to the findings, the SVM algorithm showed the best results in classification of Test samples. So, this intelligent system can be used as a physician assistant in medical education hospitals, where the diagnosis processes are performed by medical students.

Highlights

Cancer is a disease in which the cells proliferate abnormally and uncontrollably and can involve their adjacent tissues
The findings showed that the highest level of F-Score was related to the support vector machine (SVM) (91.99); whereas, the lowest rate was associated with the k nearest neighborhood (KNN) algorithm (87.17)
We studied the four learning methods of SVM, decision tree (DT), naive Bayesian model, and KNN to classify the patients with gastric cancer

Summary

Introduction

Cancer is a disease in which the cells proliferate abnormally and uncontrollably and can involve their adjacent tissues. Gastric cancer is the fourth most common cancer in the world after lung, breast, and intestinal cancers (Biglarian et al, 2011) and is generally ranked as the second cause of cancer death worldwide (Billiar et al, 2009) This cancer should reach its advanced levels to show the symptoms such as: 1. Data mining is used to solve problems with no or very complex algorithmic solutions, such as the problems related to clinical diagnosis, analysis of medical images, and survival prediction. Considering the advancement of machine learning algorithms and the difficulty of diagnosing the gastric cancer by clinical and pharmacological parameters, we need to use computer systems such as data mining more than ever. This study has been done to determine the possibility of predicting gastric cancer by using data mining techniques and risk factors

Materials and Methods

Findings

Discussion