文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > 1 Artificial Neural Networks in the Diagnosis of Prostate Cancer
侵权投诉

1 Artificial Neural Networks in the Diagnosis of Prostate Cancer

Methods for diagnosing prostate cancer suffer seriously from the problem of false-negative and false-positive results. In this study, we attempted to determine whether a special case of Artificial Neural Networks, Multi-Layer

Artificial Neural Networks in the Diagnosis of Prostate Cancer

Mehran Gomari (mehgom@cs.utu.fi)

Department of Computer Science

University of Turku

Abstract

Methods for diagnosing prostate cancer suffer seriously from the problem of false-negativeand false-positive results. In this study, we attempted to determine whether a special case ofArtificial Neural Networks, Multi-Layer Perceptrons >1@ >2@ >3@, would improve the diagnosisof prostate cancer by minimizing the false-negative and false-positive rates. We trainedseveral MLPs to predict biopsy results using two sets of data extracted from two prostatespecific antigen screening databases. An MLP predicted the biopsy results of a single test setwith 86.60% overall accuracy. Using the cross-validation method to estimate the performanceof the trained MLPs for the same database, an MLP predicted the biopsy results with 74.33%overall accuracy.

Keywords: artificial neural network, multi-layer perceptron, cross-validation; prostatespecific antigen, sensitivity, specificity.

Introduction

Prostate gland is the large gland round the neck of the bladder that is accessory to the malegenital organs. Prostate cancer is a frequent cancer among men in their 60s, especially in theindustrialized countries. According to the American Cancer Society, death from prostatecancer is increasing by 2-3% per year.

At the moment, there is no specific clinical factor by which physicians can say surely whetherthere is prostate cancer in question or not. However, the most used factor for detectingprostate cancer is measuring prostate-specific antigen (PSA) in the bloodstream. This antigenis produced only by the prostate gland. Its amount increases gradually with age and its leveljumps sharply when the patient has cancer. Although PSA can detect the disease in manycases, it suffers from the problem of false-positive (FP) and false-negative (FN) cases. Inother words, there are non-cancer cases with high PSA counts that PSA cut-off classifies ascancer cases (FP); on the other hand, there are cancer cases with low PSA counts that PSAcut-off classifies as non-cancer cases (FN). The problem is how the number of FPs and FNscan be minimized. Minimizing FPs and FNs is equal to maximizing sensitivity and specificity.Sensitivity is the percentage of recognized cancer cases among all cancer cases; andspecificity is the percentage of recognized non-cancer cases among all non-cancer cases.

One way to reduce the number of false cases according to Stenman et al. is to measure PSAcomplex with ,1-antichymotrypsin >4@. There are two predominant forms of PSA: PSA-Freeand PSA complex with ,1-antichymotrypsin (PSA-ACT). These two parts together form totalPSA or simply PSA. Stenman et al. have shown that the measurement of the ratio of PSA-

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email