乳癌是現今婦女最常罹患的癌症，隨著醫學研究的發展與進步，若早期發現並接受治療能夠提高乳癌的治癒率。在醫學影像常用來診斷乳房腫瘤的工具有乳房X光攝影、超音波影像與核磁共振影像，對於電腦輔助分析系統而言，準確的乳腺體積和乳腺密度已被證明可以幫助醫生有效地預測罹癌的風險，在乳房磁振造影(MRI)上進行乳腺輪廓描繪，是一個相當重要的步驟。隨著乳房磁振造影愈來愈廣泛被使用，自動切割乳腺組織變得重要，在臨床上應用很迫切。因此，本研究提出一個乳腺輪廓自動描繪演算法，來輔助醫生判讀乳房磁振造影的乳腺組織資訊。本研究先採用各項擴散異性濾波法進行乳房磁振造影的前處理，以降低影像的雜訊，再調整影像對比度，使得乳腺區域和乳房區域分離並讓邊緣更加明顯，影像切割的主要技術採用三維區域成長法取得乳腺組織，最後再使用型態學的方法將切割的結果進行後處理修補乳腺區域，使得切出來的乳腺組織可以更為精確。本研究總共使用10個病例進行實驗，最後實驗切割出的結果會與醫師手動描繪的乳腺區域進行比較，並計算四個指數(SI、OF、OV、EF)評估相似性。 Breast cancer is the most common cancer in woman. The development and progress of medical research, if early detection and treatment can improve the cure rate of breast cancer. There are many ways to diagnose breast tumors in medical imaging tools, such as mammography, ultrasonography and magnetic resonance imaging (MRI). In computer aided analysis of MRI, contouring of breast fibroglandular region is an important step. Accurate volume of fibroglandular tissue and breast density should help physicians to effective predict the risk of cancer. As breast MRI becomes more widespread used, a functional automatic method for extracting fibroglandular breast tissue is essential and its clinical application is becoming urgent. This study proposes a robust segmentation method to assist the physician on contouring breast fibroglandular region. The proposed method first utilizes the anisotropic diffusion filtering to reduce the noises and speckle in MRI images. Three-dimensional (3D) region growing method is applied to segment the breast fibroglandular area. Finally, the proposed method obtains the area smoother and correctly though a post processing step. All segmentation methods are three-dimensional, compared to two-dimensional segmentation can be considered more relevance, the results more accurate. This study evaluated total of 10 breast 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, and the proposed segmentation method.
