基于雙閾值的肺結(jié)節(jié)自動(dòng)檢測(cè).doc
約56頁(yè)DOC格式手機(jī)打開展開
基于雙閾值的肺結(jié)節(jié)自動(dòng)檢測(cè),全文56頁(yè) 約15000字圖文并茂論述翔實(shí)摘 要肺癌是目前發(fā)病率和死亡率較高的惡性腫瘤,其預(yù)后很大程度上取決于早期診斷和治療,而ct掃描是診斷肺癌的重要手段。但是,大量的ct圖像不僅導(dǎo)致醫(yī)生工作量增加,同時(shí)也增加了漏診和誤診的幾率,特別是對(duì)于10mm以內(nèi)spn的影像診斷更難。圖像處理技術(shù)如圖像...
內(nèi)容介紹
此文檔由會(huì)員 花季永駐 發(fā)布
基于雙閾值的肺結(jié)節(jié)自動(dòng)檢測(cè)
全文56頁(yè) 約15000字 圖文并茂 論述翔實(shí)
摘 要
肺癌是目前發(fā)病率和死亡率較高的惡性腫瘤,其預(yù)后很大程度上取決于早期診斷和治療,而CT掃描是診斷肺癌的重要手段。但是,大量的CT圖像不僅導(dǎo)致醫(yī)生工作量增加,同時(shí)也增加了漏診和誤診的幾率,特別是對(duì)于10mm以內(nèi)SPN的影像診斷更難。圖像處理技術(shù)如圖像分割、圖像的三維重建和顯示等,使得計(jì)算機(jī)輔助檢測(cè)成為可能,可以輔助醫(yī)生對(duì)病變感興趣區(qū)域進(jìn)行分析和判斷,從而提高醫(yī)療診斷的效率,減輕醫(yī)生的負(fù)擔(dān)。
本文介紹了一種基于CT圖像的肺結(jié)節(jié)計(jì)算機(jī)輔助自動(dòng)診斷系統(tǒng),我們將肺結(jié)節(jié)的自動(dòng)檢測(cè)分為肺實(shí)質(zhì)的提取、感興趣區(qū)域(ROI)的分割和ROI特征參數(shù)提取及分類判別幾個(gè)步驟,對(duì)普通CT圖像和高分辨率CT圖像的提取和檢測(cè)做了探討。肺實(shí)質(zhì)的提取使用高斯混合模型和區(qū)域生長(zhǎng)連通標(biāo)記等方法,對(duì)于結(jié)節(jié)與胸膜相連的情況也進(jìn)行了討論。得到肺實(shí)質(zhì)后我們使用自適應(yīng)區(qū)域生長(zhǎng)算法,結(jié)合“雙閾值”思想,對(duì)已經(jīng)分割好的肺部圖象進(jìn)行ROI的提取,并且對(duì)普通CT圖象和高分辨率CT圖象分別進(jìn)行了二維結(jié)節(jié)提取和三維結(jié)節(jié)提取。
結(jié)節(jié)區(qū)域和正常組織的分類判別是本論文討論的另一個(gè)主要內(nèi)容。目前的CAD系統(tǒng)普遍對(duì)小于10mm的小結(jié)節(jié)的檢測(cè)率不高,并且假陽(yáng)很高。結(jié)合醫(yī)學(xué)背景知識(shí),本文采用了多選擇過濾器的線性分類方法對(duì)結(jié)節(jié)進(jìn)行分類,用敏感性-特異性曲線(Receiver Operating Characteristic Curves:ROC)評(píng)估分類器性能。
關(guān)鍵詞: 計(jì)算機(jī)輔助檢測(cè) 圖像分割 肺結(jié)節(jié) 特征提取 雙閾值
ABSTRACT
Lung cancer is a malignant disease with high incidence and mortality, and its prognosis is to large degree determined by early diagnosis and treatment. Computed tomography (CT) are important means to discover and diagnose lung cancer. However, the large amount of CT images caused increasing work and inevitable false diagnosis rate,especially for those less than 10 mm. Image processing techniques, for example segmentation of images, reconstruction and rendering of suspicious objects, make it possible to carry on the computer aided detection (CAD), help doctors to analyze pathological changes and other regions of interest, and release the doctors’burden.
In this paper, we present a computer-aided automatic diagnostic system for pulmonary nodules based on CT images. We separate the whole image analysis work to several steps: the extraction of pulmonary parenchyma, the segmentation of region of interests (ROI), the feature extraction and classification . The difference and the corresponding result of conventional CT and high resolution CT (HRCT) is discussed. We use a series of image processing methods such as computing of FGGM and the region growing to combine label. Especially we discuss the case of nodules attached with pleural surface. Then wo get ROI from the whole lungs parenchyma by using the proposed adaptive region growth algorithm and combining the technique of “Two” threshold. For concentional CT and high resolution CT(HRCT), we proposed adaptive 2-D and 3-D region growth algorithm.
Classification of nodules and normal areas is the other main content of the dissertation. The traditional system of CAD have a great deal of false positives lie in ROI,and bad detected result of SPN measuring less than 10mm.According to prior clinical knowledge,we use linear discriminant analysis with many selections filter to classify nodules . the discriminant scores are analyzed using Reciever Operating Characteristic Curves(ROC) method.
Keyword: CAD, Image Segmentation, Lung nodule Detection,pulmonary nodules, Feature Extraction,Two Threshold
目 錄
第一章 緒論…………………………………………………………………………1
1.1 課題背景及意義……………………………………………………………1
1.2 國(guó)內(nèi)外研究進(jìn)展及現(xiàn)狀……………………………………………………
1.2.1 計(jì)算機(jī)輔助檢測(cè)系統(tǒng)的國(guó)內(nèi)外研究現(xiàn)狀……………………………
1.2.2 計(jì)算機(jī)輔助診斷的基本原理…………………………………………
1.2.3 計(jì)算機(jī)輔助診斷在醫(yī)學(xué)影像學(xué)中的應(yīng)用……………………………
1.2.4 CT圖像中的肺結(jié)節(jié)檢測(cè)算法的研究現(xiàn)狀……………………………
1.3 論文的主要研究成果及章節(jié)安排…………………………………………
第二章 基于高斯混合模型的肺部自動(dòng)分割
2.1 引言……………………………………………………………………………
2.2 基于高斯混合模型自動(dòng)分割方法……………………………………………
2.2.1 高斯混合模型和EM算法………………………………………………
2.2.2 自適應(yīng)閾值的計(jì)算……………………………………………………
2.2.3 圖像的二值化…………………………………………………………
2.2.4 去除檢查床……………………………………………………………
2.2.5 區(qū)域生長(zhǎng)連通軀干和背景……………………………………………
2.2.6 肺實(shí)質(zhì)的識(shí)別…………………………………………………………
2.2.7 肺分割…………………………………………………………………
2.3 實(shí)驗(yàn)結(jié)果討論…………………………………………………………………
2.4 小結(jié)……………………………………………………………………………
第三章 基于雙閾值的肺部感興趣區(qū)域的提取………………………………………
3.1 肺結(jié)節(jié)的影像分析……………………………………………………………
3.1.1 概述……………………………………………………………………
3.1.2 肺結(jié)節(jié)CT征像分析及其病理相關(guān)性…………………………………
3.2 肺結(jié)節(jié)檢測(cè)算法的研究現(xiàn)狀…………………………………………………
3.2.1 引言……………………………………………………………………
3.2.2 肺結(jié)節(jié)檢測(cè)算法研究現(xiàn)狀……………………………………………
3.3 肺結(jié)節(jié)二維自動(dòng)提取算法……………………………………………………
3.3.1 肺結(jié)節(jié)提取算法………………………………………………………
3.3.2 種子區(qū)域的選取………………………………………………………
3.3.3 ROI提取的實(shí)驗(yàn)結(jié)果分析及算法
3.4 肺結(jié)節(jié)三維提取算法…………………………………………………………
3.4.1 薄層和厚層CT掃描圖像的結(jié)節(jié)提取算法……………………………
3.4.2 肺結(jié)節(jié)三維提取算法…………………………………………………
3.5 小結(jié)……………………………………………………………………………
第四章 肺結(jié)節(jié)特征提取和檢測(cè)………………………………………………………
4.1 引言……………………………………………………………………………
4.2 特征提取以及分類……………………………………………………………
4.3 實(shí)驗(yàn)結(jié)果分析…………………………………………………………………
4.4 小結(jié)……………………………………………………………………………
第五章 結(jié)束語(yǔ)…………………………………………………………………………
5.1 總結(jié)……………………………………………………………………………
5.2 今后的研究工作與展望………………………………………………………
致謝………………………………………………………………………………………
參考文獻(xiàn)…………………………………………………………………………………
部分參考文獻(xiàn)
[4] 薛以鋒, 鮑旭東, 等. 基于CT圖像的肺結(jié)節(jié)計(jì)算機(jī)輔助診斷系統(tǒng). 中國(guó)醫(yī)學(xué)物理學(xué)雜志. 2006, 23(2): 93-96.
[5] Matthew S. Brown, Michael F. McNitt-Gray, et al. Patient-specific models for lung nodule detection and surveillance in CT images. IEEE Trans Med Imag. 2001, 20(12): 1242-1250.
[6] Muhm J R, Miller W E, Fontana R S, et al. Lung cancer detected during a screening program using four-month chest radiography. Radiology. 1983, 148: 606-615.
全文56頁(yè) 約15000字 圖文并茂 論述翔實(shí)
摘 要
肺癌是目前發(fā)病率和死亡率較高的惡性腫瘤,其預(yù)后很大程度上取決于早期診斷和治療,而CT掃描是診斷肺癌的重要手段。但是,大量的CT圖像不僅導(dǎo)致醫(yī)生工作量增加,同時(shí)也增加了漏診和誤診的幾率,特別是對(duì)于10mm以內(nèi)SPN的影像診斷更難。圖像處理技術(shù)如圖像分割、圖像的三維重建和顯示等,使得計(jì)算機(jī)輔助檢測(cè)成為可能,可以輔助醫(yī)生對(duì)病變感興趣區(qū)域進(jìn)行分析和判斷,從而提高醫(yī)療診斷的效率,減輕醫(yī)生的負(fù)擔(dān)。
本文介紹了一種基于CT圖像的肺結(jié)節(jié)計(jì)算機(jī)輔助自動(dòng)診斷系統(tǒng),我們將肺結(jié)節(jié)的自動(dòng)檢測(cè)分為肺實(shí)質(zhì)的提取、感興趣區(qū)域(ROI)的分割和ROI特征參數(shù)提取及分類判別幾個(gè)步驟,對(duì)普通CT圖像和高分辨率CT圖像的提取和檢測(cè)做了探討。肺實(shí)質(zhì)的提取使用高斯混合模型和區(qū)域生長(zhǎng)連通標(biāo)記等方法,對(duì)于結(jié)節(jié)與胸膜相連的情況也進(jìn)行了討論。得到肺實(shí)質(zhì)后我們使用自適應(yīng)區(qū)域生長(zhǎng)算法,結(jié)合“雙閾值”思想,對(duì)已經(jīng)分割好的肺部圖象進(jìn)行ROI的提取,并且對(duì)普通CT圖象和高分辨率CT圖象分別進(jìn)行了二維結(jié)節(jié)提取和三維結(jié)節(jié)提取。
結(jié)節(jié)區(qū)域和正常組織的分類判別是本論文討論的另一個(gè)主要內(nèi)容。目前的CAD系統(tǒng)普遍對(duì)小于10mm的小結(jié)節(jié)的檢測(cè)率不高,并且假陽(yáng)很高。結(jié)合醫(yī)學(xué)背景知識(shí),本文采用了多選擇過濾器的線性分類方法對(duì)結(jié)節(jié)進(jìn)行分類,用敏感性-特異性曲線(Receiver Operating Characteristic Curves:ROC)評(píng)估分類器性能。
關(guān)鍵詞: 計(jì)算機(jī)輔助檢測(cè) 圖像分割 肺結(jié)節(jié) 特征提取 雙閾值
ABSTRACT
Lung cancer is a malignant disease with high incidence and mortality, and its prognosis is to large degree determined by early diagnosis and treatment. Computed tomography (CT) are important means to discover and diagnose lung cancer. However, the large amount of CT images caused increasing work and inevitable false diagnosis rate,especially for those less than 10 mm. Image processing techniques, for example segmentation of images, reconstruction and rendering of suspicious objects, make it possible to carry on the computer aided detection (CAD), help doctors to analyze pathological changes and other regions of interest, and release the doctors’burden.
In this paper, we present a computer-aided automatic diagnostic system for pulmonary nodules based on CT images. We separate the whole image analysis work to several steps: the extraction of pulmonary parenchyma, the segmentation of region of interests (ROI), the feature extraction and classification . The difference and the corresponding result of conventional CT and high resolution CT (HRCT) is discussed. We use a series of image processing methods such as computing of FGGM and the region growing to combine label. Especially we discuss the case of nodules attached with pleural surface. Then wo get ROI from the whole lungs parenchyma by using the proposed adaptive region growth algorithm and combining the technique of “Two” threshold. For concentional CT and high resolution CT(HRCT), we proposed adaptive 2-D and 3-D region growth algorithm.
Classification of nodules and normal areas is the other main content of the dissertation. The traditional system of CAD have a great deal of false positives lie in ROI,and bad detected result of SPN measuring less than 10mm.According to prior clinical knowledge,we use linear discriminant analysis with many selections filter to classify nodules . the discriminant scores are analyzed using Reciever Operating Characteristic Curves(ROC) method.
Keyword: CAD, Image Segmentation, Lung nodule Detection,pulmonary nodules, Feature Extraction,Two Threshold
目 錄
第一章 緒論…………………………………………………………………………1
1.1 課題背景及意義……………………………………………………………1
1.2 國(guó)內(nèi)外研究進(jìn)展及現(xiàn)狀……………………………………………………
1.2.1 計(jì)算機(jī)輔助檢測(cè)系統(tǒng)的國(guó)內(nèi)外研究現(xiàn)狀……………………………
1.2.2 計(jì)算機(jī)輔助診斷的基本原理…………………………………………
1.2.3 計(jì)算機(jī)輔助診斷在醫(yī)學(xué)影像學(xué)中的應(yīng)用……………………………
1.2.4 CT圖像中的肺結(jié)節(jié)檢測(cè)算法的研究現(xiàn)狀……………………………
1.3 論文的主要研究成果及章節(jié)安排…………………………………………
第二章 基于高斯混合模型的肺部自動(dòng)分割
2.1 引言……………………………………………………………………………
2.2 基于高斯混合模型自動(dòng)分割方法……………………………………………
2.2.1 高斯混合模型和EM算法………………………………………………
2.2.2 自適應(yīng)閾值的計(jì)算……………………………………………………
2.2.3 圖像的二值化…………………………………………………………
2.2.4 去除檢查床……………………………………………………………
2.2.5 區(qū)域生長(zhǎng)連通軀干和背景……………………………………………
2.2.6 肺實(shí)質(zhì)的識(shí)別…………………………………………………………
2.2.7 肺分割…………………………………………………………………
2.3 實(shí)驗(yàn)結(jié)果討論…………………………………………………………………
2.4 小結(jié)……………………………………………………………………………
第三章 基于雙閾值的肺部感興趣區(qū)域的提取………………………………………
3.1 肺結(jié)節(jié)的影像分析……………………………………………………………
3.1.1 概述……………………………………………………………………
3.1.2 肺結(jié)節(jié)CT征像分析及其病理相關(guān)性…………………………………
3.2 肺結(jié)節(jié)檢測(cè)算法的研究現(xiàn)狀…………………………………………………
3.2.1 引言……………………………………………………………………
3.2.2 肺結(jié)節(jié)檢測(cè)算法研究現(xiàn)狀……………………………………………
3.3 肺結(jié)節(jié)二維自動(dòng)提取算法……………………………………………………
3.3.1 肺結(jié)節(jié)提取算法………………………………………………………
3.3.2 種子區(qū)域的選取………………………………………………………
3.3.3 ROI提取的實(shí)驗(yàn)結(jié)果分析及算法
3.4 肺結(jié)節(jié)三維提取算法…………………………………………………………
3.4.1 薄層和厚層CT掃描圖像的結(jié)節(jié)提取算法……………………………
3.4.2 肺結(jié)節(jié)三維提取算法…………………………………………………
3.5 小結(jié)……………………………………………………………………………
第四章 肺結(jié)節(jié)特征提取和檢測(cè)………………………………………………………
4.1 引言……………………………………………………………………………
4.2 特征提取以及分類……………………………………………………………
4.3 實(shí)驗(yàn)結(jié)果分析…………………………………………………………………
4.4 小結(jié)……………………………………………………………………………
第五章 結(jié)束語(yǔ)…………………………………………………………………………
5.1 總結(jié)……………………………………………………………………………
5.2 今后的研究工作與展望………………………………………………………
致謝………………………………………………………………………………………
參考文獻(xiàn)…………………………………………………………………………………
部分參考文獻(xiàn)
[4] 薛以鋒, 鮑旭東, 等. 基于CT圖像的肺結(jié)節(jié)計(jì)算機(jī)輔助診斷系統(tǒng). 中國(guó)醫(yī)學(xué)物理學(xué)雜志. 2006, 23(2): 93-96.
[5] Matthew S. Brown, Michael F. McNitt-Gray, et al. Patient-specific models for lung nodule detection and surveillance in CT images. IEEE Trans Med Imag. 2001, 20(12): 1242-1250.
[6] Muhm J R, Miller W E, Fontana R S, et al. Lung cancer detected during a screening program using four-month chest radiography. Radiology. 1983, 148: 606-615.