基於深度學習自動偵測缺血性中風於電腦斷層影像;Automated Infarct Detection in Computed Tomography Imaging Based on Deep Learning

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Electrical Engineering > Electronic Thesis & Dissertation > Item 987654321/84172

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/84172

Title:	基於深度學習自動偵測缺血性中風於電腦斷層影像;Automated Infarct Detection in Computed Tomography Imaging Based on Deep Learning
Authors:	楊景聿;Yang, Jing-Yu
Contributors:	電機工程學系
Keywords:	電腦斷層影像;深度學習;卷積神經網路;腦血管梗塞;computed tomography;deep learning;convolutional neural network;cerebral infarction
Date:	2020-07-28
Issue Date:	2020-09-02 18:26:53 (UTC+8)
Publisher:	國立中央大學
Abstract:	近年來，腦血管疾病居於台灣十大死因之列，因此腦血管疾病的醫療已成為社會的主要負擔。雖然預防和治療高血壓能有效降低腦血管疾病的發生，但其高死亡率和致殘率仍是台灣與全世界許多國家的醫療問題。本論文探討急性腦梗塞中風患者之電腦斷層影像(Computed Tomography)上梗塞區域的偵測，目前大部分的醫院都使用軟體輔助之半自動分割腦梗塞區域，處理與分析耗時過久且較容易產生不同量測者之偏差。大部分的軟體也都是運用在拍攝時間較久且成本較高核磁共振影像(Magnetic Resonance Imaging)上，且現在大部分發表有關電腦斷層掃描影像上病灶偵測的論文及技術都只侷限於分類影像是否有缺血性梗塞的病灶，而無法確切地找出病灶的位置，本論文的方法利用了影像對位處理、影像像素處理以及T分數統計的方式在病人與健康電腦斷層影像的特性發展新穎的腦梗塞區域偵測的方法當作預前處理，再運用深度學習的技術，能準確的偵測出病灶在CT影像上的所在位置。本論文同時也運用 MATLAB 製作了圖形使用者介面，讓此方法能在臨床上更有效率的幫助醫師在電腦斷層影像上的診斷。;In recent years, cerebrovascular diseases have been among the top ten causes of death in Taiwan, so the medical treatment of cerebrovascular diseases has become the main burden of society. The prevention and treatment of hypertension can effectively reduce the occurrence of cerebrovascular diseases but its high mortality rate and disability rate are still a serious medical problem in many countries. This paper investigates the detection of infarcted areas on computed tomography (Computed Tomography) in patients with acute cerebral infarction. Currently, most of the software is used in time-consuming and high-cost magnetic resonance imaging (Magnetic Resonance Imaging), and most of the papers and techniques which published about detection of lesions on computed tomography images are limited to classified images. However, the location of the lesion cannot be accurately identified. The method of this paper uses image alignment processing, image pixel processing and T score statistical methods to develop a new cerebral infarction area detection method in the characteristics of patient and healthy computerized tomographic images as preprocessing, and then use deep learning techniques to accurately detect the location of the lesion on the CT image. Besides, we also designed a graphical user interface by using MATLAB to help doctors diagnose the computed tomography image more effectively and accurately.
Appears in Collections:	[Graduate Institute of Electrical Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	123	View/Open

社群 sharing

Loading...