乳癌是現今婦女最常罹患的癌症，然而隨著醫療科技的進步，若早期發現則治癒率可相對提高，在醫學影像的工具中可用來診斷乳房腫瘤的方法有許多，例如：乳房X光攝影、超音波影像或是核磁共振影像。其中，超音波影像是一個很好的選擇，它不具侵入性而且可以及時造影，相較與核磁共振影像成本較低，因此乳房超音波影像檢驗為目前普遍的診斷方式。但超音波影像有些特性需要克服，它通常包含大量的雜訊及斑點，若依賴醫師的臨床經驗來判斷腫瘤的位置及良惡性，很可能會因為影像本身的雜訊導致錯誤判斷，進而影響醫生的判讀結果，因此，本文提出一個腫瘤輪廓自動描繪法來輔助醫生判讀影像資訊。此方法的前處理先採用高斯濾波和各項擴散異性濾波降低影像的雜訊，再來調整對比度使得腫瘤和背景區分出來，並讓邊緣更加明顯，切割的方法採用區域成長法，此方法的優點是簡單且省時，後處理使用型態學的方法將切割的結果進行修補，使得切出來的腫瘤可以更貼緊邊緣，與醫生手繪的結果更為相似。所有的切割方法皆為三維，相較於二維切割可以考慮較多的關聯性，使結果更為準確。本研究總共使用30個病例進行實驗，分別為15個良性及15個惡性，最後實驗切割出的結果會與醫師手動描繪的腫瘤及自動切割方法VOCAL進行比較。 Breast cancer is the most common cancer in the woman. There is an upward trend in the number of such cases in the past years. Early diagnosis and early treatment is the most effective way of reducing mortality caused by breast cancer. Ultrasound is the common examination technology because of fast, cheap and noninvasive. The difficult part is that there are many noises and speckles on the ultrasonic images. Malignant and benign breast tumors are present different shape and size, thus this study propose a robust segmentation method to assist the physician on contouring tumor boundary. The proposed method first utilizes a pre-processing procedure to reduce the noises and speckle in imaging. After that, three-dimensional (3D) region growing method is applies to segment the tumor area. Finally, the proposed method made the area smoother and correctly though a post processing step. This study evaluated total of 30 tumor cases and four practical similarity measures (similarity index, overlap fraction, overlap value, and extraction fraction) are used to evaluate the result between the manually determined contours, an automatically method named virtual organ computer-aided analysis (VOCAL) and the proposed segmentation method.
