請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/45159
標題: | 聲速與衰減係數於超音波乳房影像之估計 Estimation of Sound Velocity and Attenuation Coefficient for Breast Ultrasound Imaging |
作者: | Chen-Han Chang 張珵涵 |
指導教授: | 李百祺(Pai-Chi Li) |
關鍵字: | 飛行時間,衰減,有興趣區域,聲速,衰減係數,視訊信號分析,週期圖法,最小側邊差, Time-of-flight,Attenuation,Region-of-interest,Sound velocity,Attenuation coefficient,Video signal analysis,Periodogram,Minimum side difference, |
出版年 : | 2010 |
學位: | 碩士 |
摘要: | 近年來,乳癌一直是女性主要死因之一,因此,如何早期發現早期治療已經成為重要的研究方向。一直以來,乳房攝影術被視為最普及且能有效偵測出早期乳癌的工具,因其可偵測出小腫瘤,但輻射劑量一直是備受爭議的部分。此外,由於緻密的乳房組織之影像區域遮蔽住小腫瘤之影像區域,因此容易造成誤判,針對這點,超音波乳房影像可提供輔助。簡言之,超音波乳房影像擁有諸多優勢,例如:非侵入式、無游離輻射、便於攜帶、不受乳房密度之限制、即時等等。然而,使用傳統乳房超音波影像常受限於乳房組織之聲速不均勻性,造成折射現象、多重訊號干擾以及聚焦誤差,導致影像之對比解析度不足,因而造成診斷正確率下降。若使用其他組織特性,如聲速以及衰減係數,將可提供額外之診斷資訊,以提高診斷正確率。
本研究第一部份的研究目標為確定利用聲速與組織衰減於臨床上區分乳癌與健康組織之成效。此方法僅需使用由一線性陣列探頭所取得之原始通道資料,因此可實踐於現有之臨床系統。此研究中提出的方法於臨床資料上做測試,總共評估了十九個切片證實的案例。使用一影像設備包含一個中心頻率為5 MHz且擁有128通道的線性陣列探頭,來同時獲得灰階影像資料、飛行時間資料以及衰減資料。透過於灰階影像之手動圈選程序,可重建有興趣區域內外之聲速與衰減係數。為了降低組織不均勻性所造成的影像失真,我們使用基於回波信號所得之最佳化濾波器,而此回波信號來自於將水取代乳房組織做為介質所得。結果顯示若適當選取相對聲速之閾值(18.5 m/s)可將惡性腫瘤從纖維腺瘤與脂肪組織區分出來。儘管如此,在衰減係數方面,使得用單一閾值的來區分惡性與良性組織的方法失效。不過此研究中所敘述的方法在臨床上仍然有減少陰性切片以及改善乳癌偵測的潛力。 本研究的第二個部份主要是要評估衰減,實踐三個不同的方法上以及一個聯合的方法用以輔助判別惡性腫瘤,並以模擬數據評估這些方法。此三種方法依序為:視訊信號分析法、利用週期圖法之頻譜估計以及最小測邊差法。上述所有方法皆可直接利用現有B-mode影像設備來實踐。視訊信號分析法是觀察B-mode影像之灰階梯度變化,週期圖法是靠著估計波束成像資料頻譜所得之中心頻率而來,最小側邊差則是由計算有興趣區域正下方以及有興趣區域正下方之左右之平均灰階值而得來。在視訊信號分析法中,等效頻率是必要的,但通常只使用名義上的中心頻率。儘管如此,使用寬頻脈衝波來當作發射信號時,為了避免發生在視訊信號分析法中的重大誤差,精確估計等效頻率是必須的。在本論文中,我們提出修正的視訊信號分析法,此方法利用週期圖法來估計等效頻率。在透過上述修正後,尤其是當發射信號之頻寬較大時,發現視訊信號分析法之準確率可獲得改善。盼於未來對於提升診斷惡性腫瘤能有所裨益。 Recently, breast cancer has been one of the leading causes of death from cancer in females. Therefore, the study of its early detection has become an important issue now. Mammography has always been considered as the most common and the most effective screening method for detecting breast cancer because it can detect non-palpable and small tumors. However, the issue about ionization radiation is disputable. Besides, the images of dense breast region cover the images of small tumor region so that misdiagnosis happens. Toward this drawback, ultrasonic breast imaging can provide some aid. In short, ultrasonic breast imaging has many advantages, including noninvasive method, without ionizing radiation, portable, not limited to dense breast, and real-time. However, the use of conventional ultrasonic pulse-echo B-mode imaging to find breast tumors is also often limited by the image distortion caused by sound-velocity inhomogeneities in the breast tissue. Using other characteristics of tissue, such as sound velocity and attenuation coefficient, would provide diagnosticians additional information to increase the accuracy of diagnosis. The aim of the first part of this study was to determine the efficacy of using sound velocity and tissue attenuation to clinically discriminate breast cancer from healthy tissues. The method requires only raw channel data acquired by a linear transducer array and can therefore be implemented on existing clinical systems. In this study, these methods were tested on clinical data. A total of 19 biopsy-proven cases were evaluated. A imaging setup consisting of a 5-MHz, 128-channel linear array was used to simultaneously obtain B-mode image data, time-of-flight data and attenuation data. The sound velocity and attenuation coefficient can be reconstructed inside and outside a region of interest manually selected in the B-mode image. To reduce distortion caused by tissue inhomogeneities, an optimal filter derived from pulse-echo data—with water replacing the breast tissue—is applied. These results indicate that carcinoma (CA) can be discriminated from fibroadenoma (FA) and fat by choosing an appropriate threshold for the relative sound velocity (i.e., 18.5 m/s). However, the large variations in the attenuation within the same type of tissue make simple thresholding ineffective. Nevertheless, the method described in this study has the potential to reduce negative biopsies and to improve the accuracy of breast cancer detection in clinics. In the second part of this study, we implemented three different approaches to evaluate attenuation and one combined method in order to help the differentiation of breast cancer. Performance of these approaches is investigated based on simulation data. The three approaches are: video signal analysis (VSA), spectral estimation using periodogram (PER), and minimum side difference (MSD). Note that all approaches can readily be implemented using current B-mode imaging setup. First, VSA is to observe the gray-level gradient on a B-mode image. Second, PER is implemented by estimating the center frequency from the periodogram of the beamform data. Third, MSD is calculated using gray-level values of areas posterior to the region of interest, and left and right posterior to the ROI. In VSA, an effective frequency is required and typically the nominal center frequency is used. However, with a broadband pulse an accurate estimate of the effective frequency is needed to avoid large errors in VSA. In this study, we propose a modified VSA method in which PER is used to estimate the effective frequency. It is shown that accuracy of the VSA method can be improved by the proposed method particularly when the transmit bandwidth is large. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/45159 |
全文授權: | 有償授權 |
顯示於系所單位: | 電機工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-99-1.pdf 目前未授權公開取用 | 2.77 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。