國立高雄大學圖資館 |

語系: 繁體中文

說明(常見問題)

圖資館首頁

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

使用基因演算法變數篩選與SVM分類器於PET/CT上孤立肺結節之診斷 =...

周書賢

使用基因演算法變數篩選與SVM分類器於PET/CT上孤立肺結節之診斷 = Diagnosis of solitary pulmonary nodule in PET/CT using GA for feature extraction and SVM classifiers

紀錄類型:	書目-語言資料,印刷品 : 單行本
並列題名:	Diagnosis of solitary pulmonary nodule in PET/CT using GA for feature extraction and SVM classifiers
作者:	周書賢,
其他團體作者:	國立高雄大學
出版地:	[高雄市]
出版者:	撰者;
出版年:	2013[民102]
面頁冊數:	64面部份彩圖，表格 : 30公分;
標題:	PET/CT
標題:	Positron emission tomography combined computed t
電子資源:	http://handle.ncl.edu.tw/11296/ndltd/79487482359088369291
附註:	參考書目：面45-48
附註:	102年10月31日公開
其他題名:	使用基因演算法變數篩選與SVM分類器於PETCT上孤立肺結節之診斷
摘要註:	醫學上的電腦斷層掃描技術(Computed Tomography，CT)方便醫師以肉眼判斷病患身體內部的異常部份，以診斷出腫瘤的形狀與位置。現常與正電子發射掃描計算機斷層掃描(Positron Emission Tomography，PET)並用，以求更快速、準確找出腫瘤。對於肺部腫瘤，診斷主要方法是在CT上針對肺結節(lung nodule)的形狀、大小與位置做判斷，也以PET上的標準攝取值(Standardized uptake value，SUV值)數值大小判斷。此篇論文針對68位病患(40位癌症患者，28位為非癌症之良性結節患者)，察看診斷報告擷取65顆孤立結節，作門檻值與取輪廓作完處理，再進行良性與惡性分類，再以統計與紋理的特徵(Texture Features)提取變數，輸入基因演算法(Genetic Algorithm)作變數篩選，並用支持向量機(Support Vector Machine，SVM)分類，由電腦自行訓練與辨識良性結節與孤立的惡性腫瘤，找出最適當的重要特徵變數。使用5-fold cross validation與5次重覆實驗，GLCM四個方向所有的變數篩選出22個變數他們的敏感度是72.41%特異度72.22%準確率72.30%。若使用十四個變數敏感度特異度準確率分別是79.31%，72.22%和75.38%。若篩選五個變數，它們的敏感度是79.31%，特異度是80.56%，準確率是80.00%。這篇論文顯示變數組合的紋理特徵有助於提高分類良性與惡性腫瘤的準確率。 The Computed Tomography (CT) scan technique in medical science enables physicians to diagnose the abnormal parts within the body of the patients with unaided eyes. It is a diagnostic tool used to detect the shape and position of the tumors. It is now jointly used with Positron Emission Tomography (PET) scan for a fast and accurate way to find out the tumors. The diagnostic method for lung tumors is to determine the shape, size and position of pulmonary nodules from the image of CT scan, taking into account the Standardized uptake value (SUV) from PET. This study was done on 68 patients (40 cancer patients, 28 non-cancer patients with benign nodules). It investigated the diagnostic report of 65 solitary nodules, and classified them into benign or malignant tumors after treating for threshold value and contouring. The tumors were subsequently computed according to their texture features to find out all the variables, applying with Genetic Algorithm (GA) for variable screening and using Support Vector Machine (SVM) for their classification. The most appropriate variable characteristics for the determination of benign and malignant tumors were found out using computer automatic training. The final result of this study was obtained from the 5-times-repeating experiment in the 5-fold cross-validation. When 22 variables were screened out of all the variables in the four directions in GLCM, their sensitivity was found to be 72.41%, specificity 72.22% and accuracy 72.30%. If 14 variables were used instead, the sensitivity, specificity and accuracy were 79.31%, 72.22% and 75.38% respectively. When 5 variables were screened, they showed a sensitivity of 79.31%, specificity 80.56% and accuracy 80.00%. This study indicates texture features from the combination of variables are useful for enhancing the accuracy for classifying benign and malignant tumors.

使用基因演算法變數篩選與SVM分類器於PET/CT上孤立肺結節之診斷 = Diagnosis of solitary pulmonary nodule in PET/CT using GA for feature extraction and SVM classifiers
周, 書賢

使用基因演算法變數篩選與SVM分類器於PET/CT上孤立肺結節之診斷 = Diagnosis of solitary pulmonary nodule in PET/CT using GA for feature extraction and SVM classifiers / 周書賢撰 - [高雄市] : 撰者, 2013[民102]. - 64面 ; 部份彩圖，表格 ; 30公分.
參考書目：面45-48102年10月31日公開.
PET/CTPositron emission tomography combined computed t

使用基因演算法變數篩選與SVM分類器於PET/CT上孤立肺結節之診斷 = Diagnosis of solitary pulmonary nodule in PET/CT using GA for feature extraction and SVM classifiers
LDR:04755nam0a2200301 450 001 389695
005 20170214093142.0
009 389695
010 0 $b 精裝
010 0 $b 平裝
100 $a 20170214d2013 k y0chiy05 b
101 1 $a eng $d chi $d eng
102 $a tw
105 $a ak am 000yy
200 1 $a 使用基因演算法變數篩選與SVM分類器於PET/CT上孤立肺結節之診斷 $d Diagnosis of solitary pulmonary nodule in PET/CT using GA for feature extraction and SVM classifiers $z eng $f 周書賢撰
210 $a [高雄市] $c 撰者 $d 2013[民102]
215 0 $a 64面 $c 部份彩圖，表格 $d 30公分
300 $a 參考書目：面45-48
300 $a 102年10月31日公開
314 $a 指導教授：殷堂凱博士
328 $a 碩士論文--國立高雄大學資訊工程學系碩士班
330 $a 醫學上的電腦斷層掃描技術(Computed Tomography，CT)方便醫師以肉眼判斷病患身體內部的異常部份，以診斷出腫瘤的形狀與位置。現常與正電子發射掃描計算機斷層掃描(Positron Emission Tomography，PET)並用，以求更快速、準確找出腫瘤。對於肺部腫瘤，診斷主要方法是在CT上針對肺結節(lung nodule)的形狀、大小與位置做判斷，也以PET上的標準攝取值(Standardized uptake value，SUV值)數值大小判斷。此篇論文針對68位病患(40位癌症患者，28位為非癌症之良性結節患者)，察看診斷報告擷取65顆孤立結節，作門檻值與取輪廓作完處理，再進行良性與惡性分類，再以統計與紋理的特徵(Texture Features)提取變數，輸入基因演算法(Genetic Algorithm)作變數篩選，並用支持向量機(Support Vector Machine，SVM)分類，由電腦自行訓練與辨識良性結節與孤立的惡性腫瘤，找出最適當的重要特徵變數。使用5-fold cross validation與5次重覆實驗，GLCM四個方向所有的變數篩選出22個變數他們的敏感度是72.41%特異度72.22%準確率72.30%。若使用十四個變數敏感度特異度準確率分別是79.31%，72.22%和75.38%。若篩選五個變數，它們的敏感度是79.31%，特異度是80.56%，準確率是80.00%。這篇論文顯示變數組合的紋理特徵有助於提高分類良性與惡性腫瘤的準確率。 The Computed Tomography (CT) scan technique in medical science enables physicians to diagnose the abnormal parts within the body of the patients with unaided eyes. It is a diagnostic tool used to detect the shape and position of the tumors. It is now jointly used with Positron Emission Tomography (PET) scan for a fast and accurate way to find out the tumors. The diagnostic method for lung tumors is to determine the shape, size and position of pulmonary nodules from the image of CT scan, taking into account the Standardized uptake value (SUV) from PET. This study was done on 68 patients (40 cancer patients, 28 non-cancer patients with benign nodules). It investigated the diagnostic report of 65 solitary nodules, and classified them into benign or malignant tumors after treating for threshold value and contouring. The tumors were subsequently computed according to their texture features to find out all the variables, applying with Genetic Algorithm (GA) for variable screening and using Support Vector Machine (SVM) for their classification. The most appropriate variable characteristics for the determination of benign and malignant tumors were found out using computer automatic training. The final result of this study was obtained from the 5-times-repeating experiment in the 5-fold cross-validation. When 22 variables were screened out of all the variables in the four directions in GLCM, their sensitivity was found to be 72.41%, specificity 72.22% and accuracy 72.30%. If 14 variables were used instead, the sensitivity, specificity and accuracy were 79.31%, 72.22% and 75.38% respectively. When 5 variables were screened, they showed a sensitivity of 79.31%, specificity 80.56% and accuracy 80.00%. This study indicates texture features from the combination of variables are useful for enhancing the accuracy for classifying benign and malignant tumors.
510 1 $a Diagnosis of solitary pulmonary nodule in PET/CT using GA for feature extraction and SVM classifiers $z eng
517 1 $a 使用基因演算法變數篩選與SVM分類器於PETCT上孤立肺結節之診斷 $z chi
610 0 $a PET/CT $a 標準攝取值(SUV) $a 支持向量機 $a 基因演算法 $a 紋理特徵
610 1 $a Positron emission tomography combined computed t $a Standardized Uptake Value (SUV) $a SVM $a Genetic Algorithms $a Texture Features
681 $a 008M/0019 $b 464103 7757 $v 2007年版
700 1 $a 周 $b 書賢 $4 撰 $3 614555
712 0 2 $a 國立高雄大學 $b 資訊工程學系碩士班 $3 353878
801 0 $a tw $b NUK $c 20131007 $g CCR
856 7 $z 電子資源 $2 http $u http://handle.ncl.edu.tw/11296/ndltd/79487482359088369291