乳房彈性超音波之電腦輔助診斷

Abstract

根據統計，乳癌是全球女性因癌症死亡的第二大主因。在所有應用超音波來做乳房癌症篩檢的技術當中，彈性超音波是一種簡單而有效的方式。某些臨床上的研究指出，彈性超音波可藉由測量正常組織與病灶部位之組織的彈性程度差異，進而區分出良性及惡性的乳房腫瘤。然而，不同的放射科醫師對於一些腫瘤特徵如：腫瘤邊界、腫瘤寬度及腫瘤面程等會持有不同意見。因此不同的觀察者對於診斷結果會有不同的看法。基於上述的理由，本篇研究主要的目的就是致力於發展一套應用於彈性超音波，同時不受觀察者影響的電腦輔助診斷系統。目前已利用等階集合法實作自動腫瘤切割的技術，並利用模糊演算法對彈性圖中腫瘤內部的組織分類，進而由醫生所選的影像或是整段彈性超音波當中擷取腫瘤特徵進行診斷。此外，我們也設計一套量化影像的方式，藉由分析彈性超音波中之彈性圖裡的組織彈性資料分佈，去選擇最適合用於診斷之影像，同時研究如何將B-mode及彈性圖當中分別取出的特徵作結合，以利提升診斷之準確率。

According to statistics, breast cancer is the global second-leading cause of cancer death among women. Among all of ultrasonic techniques for breast ultrasound examination, elastography is an easily performed and efficient component of the ultrasound examination for breast. Some clinical studies had reported elastography is used to differentiate benign from malignant breast lesions based on evaluating the difference in tissue strain between normal and diseased tissue. However, diagnostic results are also observer dependent, i.e. different physicians may have different opinions on the lesion characteristics such as boundary, width, and area. Therefore the main purpose of this study is to investigate observer-independent computer-aided diagnostic schemes on ultrasound elastography. Currently we had developed an automatic tumor segmentation technique based on the level set algorithm. Tissues within the lesion on the elastogram were classified using the fuzzy c-means algorithm. Elastographic features were extracted from the physician-selected slice or from the entire sequence to diagnose tumors. In addition, an image quantification method based on analyzing the distribution of tissue strains will be used to automatically choose representative slice for diagnosis. Furthermore, features respectively extracted from B-mode image and elastogram will be combined to improve the diagnostic performance.