請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/63894
標題: | 定量乳房緻密度評估與分析 Quantitative Estimation and Analysis of Breast Density |
作者: | Jie-Fan Chang 張傑帆 |
指導教授: | 張瑞峰(Ruey-Feng Chang) |
關鍵字: | 乳房緻密度,斷層合成,核磁共振,數位X光攝影,超音波,深度學習,卷積神經網絡, breast density,tomosynthesis,MRI,digital mammography,ultrasound,deep learning,Convolutional Networks, |
出版年 : | 2020 |
學位: | 博士 |
摘要: | 前人研究結果顯示,相較於一般乳房密度的女性,具有高乳房密度的女性罹患乳癌的風險相對高出了四到六倍,然而,若採用常用的檢測方式,如:乳房X光攝影(Mammography),並無法正確的檢測乳房密度高的女性。本研究所開發的斷層合成(Tomosynthesis)和計算機輔助定量方法可以提供更可靠的乳房緻密度評估。本研究從斷層合成(Tomosynthesis)圖像中提取了纖維腺組織和整個乳腺組織,以進行3-D密度估計,乳房密度和乳房體積的百分比的計算應用於斷層合成(Tomosynthesis)、核磁共振(MRI)和乳房X光攝影(Mammography)。並比較分析結果表明,在斷層合成(Tomosynthesis)、核磁共振(MRI)和乳房X光攝影(Mammography)之間,乳房密度和體積存在顯著相關性。由於無輻射的乳房超音波(US)檢測可提供精確的篩查結果並檢測與突顯緻密乳房中的病灶組織,已被廣泛用於乳癌檢測,然而此檢測方式仍受限於缺乏位置信息和操作者的差異與依賴性等問題,因此仍可能會導致緻密度估計偏差,以及有結果不可重複性的問題。本研究提出的全乳房超音波(WBUS)系統能夠追踪探頭的3D空間位置。系統提供的精確位置信息可以解決有關重現性的問題,並大大降低對操作者依賴性。因此,本研究提出的全乳房超音波(WBUS)系統可以提供精確的位置信息,並有助於可靠的乳房密度評估。此外,本研究通過U-Net進行乳房區域的自動分割,並使用模糊c均值(FCM)分類器從乳房區域中提取纖維腺組織,最後計算並比較了U-Net預測區域的乳房密度和乳房面積的百分比及由專家所圈選的的乳房區域的百分比。發現由專家所圈選的結果與本研究所提出方法的結果之間,乳房密度、纖維腺組織面積,和乳房面積之間存在顯著且極高的相關性。根據所得結果,可以總結斷層合成(Tomosynthesis)可以通過本研究提出的方法提供有關乳房密度的有用的3D訊息。此外,本研究所提出的WBUS系統可做為臨床使用中乳房篩查和密度估計的替代影像系統。 Women with a high breast density are found to have fourfold to sixfold increase in the risk of developing the breast cancer. However, the detection accuracy in most commonly used modality, such as mammography, is limited when performed to women with high density breast. A newly developed tomosynthesis and computer-aided quantitative method could provide more reliable breast density evaluation. In this study, the fibroglandular tissues and whole breast tissues were extracted from tomosynthesis images for 3-D density estimation. Percentage of breast density and breast volume were computed from tomosynthesis, MRI and digital mammography. Comparison of the results showed that there are significant correlations on breast density as well as volume among tomosynthesis, MRI and digital mammography. As a radiation free modality, breast ultrasound (US) screening has been used for breast cancer detection which can provide precise screening results and detect masses in dense breasts, but it is limited due to the lack of position information and the issue of operator dependency which could cause bias in density estimation and non-reproducible results. The proposed whole-breast ultrasound (WBUS) system is able to track the 3-D spatial positon of the probe. The precise location information can solve the problem regarding reproducibility and decrease the operation dependency greatly. Thus, the proposed whole-breast ultrasound (WBUS) system can provide precise position information and facilitate reliable breast density evaluation. An automatic breast region segmentation by U-Net was conducted, and fibroglandular tissues were then extracted from breast region using fuzzy c-mean (FCM) classifier. Finally, the percentage of breast density and breast area of the U-Net predicted region and the breast region of the ground truth were calculated and compared. Significant and very high correlations of breast density, fibroglandular tissue area and breast area were found between the ground truth and the result of the proposed method for ultrasound images. Based on the results, it could be suggested that tomosynthesis could provide useful 3-D information on breast density through the proposed method. Moreover, the proposed WBUS system could have the potential to be an alternative modality for breast screening and density estimation in clinical use. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/63894 |
DOI: | 10.6342/NTU202000688 |
全文授權: | 有償授權 |
顯示於系所單位: | 資訊工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-109-1.pdf 目前未授權公開取用 | 4.06 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。