運用深度學習方法預測癌症種類及存活死亡與治癒復發;Deep learning in predicting outcomes of cancer type, overall survival and disease free survival

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/77702

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/77702

Title:	運用深度學習方法預測癌症種類及存活死亡與治癒復發;Deep learning in predicting outcomes of cancer type, overall survival and disease free survival
Authors:	陳柏傑;Chen, Po-Chieh
Contributors:	資訊工程學系
Keywords:	機器學習;癌症;醫學預測;深度學習;Maching Learning;Cancer;Medical prediction;Deep learning
Date:	2018-08-07
Issue Date:	2018-08-31 14:53:19 (UTC+8)
Publisher:	國立中央大學
Abstract:	深度類神經網路(DNN)在各個不同的領域，舉凡聲音，影像等領域中皆有不凡的表現，而目前在生醫工程方面也陸續有深度學習方法的應用，而本論文中，我們使用來自The Cancer Genome Atlas(TCGA)上，針對DNA去做次世代定序而得出的RNA-seq資料，因其通量數高，且背景雜質低的特性，所以能最正確的偵測基因表現量。在本篇論文裡，主要大方向為三： 1. 使用基因表現來判定癌症種類之預測 2. 分別對肺癌、乳癌、腦癌之存活，死亡分析其分類 3. 分別對肺癌、乳癌、腦癌之治癒，復發分析其分類在本篇論文裡，實驗了一部份機器學習之方法，如決策樹(Decision Tree)，以及支援向量機(Support Vector Machine)，XGBoost(Extreme Gradient Boost)，也加入深度學習之方法，深度神經網路(Deep Neural Network)、自編碼器(Auto-encoder)及變異型自編碼器(Variational Auto-encoder)，去比較各個方法的辨識率。 ;Deep neural networks (DNN) have extraordinary performances in various of fields such as sound and image processing. Recently, deep learning methods are applied in the field of biomedical engineering. In this paper, we use the RNA-sequencing data from TCGA (The Cancer Genome Atlas), which is sequenced from RNA data and generated by NGS (Next Generation Sequencing). Due to its high flux number and low background impurities, the most accurate detection of gene expression become possible. In this paper, we have tree main directions: 1. Classification of cancer types based on RNA- sequencing data. 2. To predict the OS (Overall Survival) of lung, breast, and brain cancer. 3. To predict the DFS (Disease Free Survival) of lung, breast, and brain cancer. In this paper, we have experimented with a number of methods of machine learning, such as Decision tree, Support Vector Machine and XGBoost (Extreme gradient boost), as well as deep learning methods, including DNN (Deep neural network), autoencoder and VAE (Variational Autoencoder). Our goal is to perform all of these methods and compare the recognition rate of each method.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	161	View/Open

社群 sharing

Loading...