利用X光乳房攝影產生之紋理特徵影像在腫瘤偵測上之研究; The Mass Detection in Mammography Using Feature Images

NCUIR > College of Electrical Engineering & Computer Science > Executive Master of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/8758

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/8758

Title:	利用X光乳房攝影產生之紋理特徵影像在腫瘤偵測上之研究;The Mass Detection in Mammography Using Feature Images
Authors:	黃俊忠;Chun-Chung Huang
Contributors:	資訊工程學系碩士在職專班
Keywords:	自動目標物偵測及分類演算法;腫瘤偵測;特徵影像;正交次空間投影法;orthogonal subspace projection;automatic mixed pixel classification method;Tumor detection;feature images
Date:	2006-06-20
Issue Date:	2009-09-22 11:34:17 (UTC+8)
Publisher:	國立中央大學圖書館
Abstract:	乳癌是目前台灣女性所有癌症死亡率排名的第四位，對國內婦女造成相當大的威脅，而早期診斷、早期發現與早期治療是減少死亡率以及延長患者存活年限的最佳方法，目前以乳房X光攝影為早期乳癌診斷最有效的方法。由於腫瘤屬於乳房組織的一部分，所以必須找出腫瘤組織與正常組織之間的差異性，以作為其偵測及判斷的依據。本論文中，將利用腫瘤在各種特徵值上的不同表現，得到不同的特徵值影像，並透過正交次空間投影法（Orthogonal Subspace Projection；OSP）及自動目標物偵測及分類演算法(Automatic Mixed Pixel Classification；AMPC)之方法，以達到腫瘤偵測的目的。本論文使用的實驗影像資料，是由歐洲The Mammographic Image Analysis Society（MIAS）所提供的MIAS　MiniMammographic　Database。在這組資料庫中，有161位受試者的左右乳房影像，共322張數位影像。其中有115張影像是經過醫師判讀與病理證實具有腫瘤或鈣化等異常組織，其餘207張影像是正常的影像。根據實驗結果影像顯示，本論文所採取之特徵值適合描述腫瘤影像與正常影像之間的差異性，而利用本論文所提出的偵測方式，可以判斷出影像中是否有腫瘤存在，並標示出該腫瘤區域。應用在臨床醫學上，可以輔助醫師對於乳房腫瘤之診斷，提高診斷的正確率，以降低乳癌對於生命的危害。 The breast cancer is at the forth place of female cancer mortality rate in Taiwan. It poses a big threat to the domestic women. The early diagnosis, discover and treatment is the best way to reduce the mortality rate as well as increase the length of the surviving of the patients. The most effective for breast cancer diagnosis method at present is the X-ray mammography. Because of the tumor is part of the breast, we must distinguish the tumor from the normal tissue, and then we can detect it. In this thesis, the features of the tumor is used to obtain different feature images, and then we adopt Orthogonal Subspace Projection and Automatic Mixed Pixel Classification method to achieve the tumor detection. The data to be used for experiments is MIAS MiniMammographic Database provided by the Mammographic Image Analysis Society (MIAS) in Europe. The data we have 322 digital breast images from 161 participants. Among them, 115 images are confirmed with tumor or microcalcification after pathology, and the other 207 images only contain normal tissue. The experimental results show that the features adopted in this thesis are suitable to distinguish the difference between tumor and normal tissue. And the proposed detection methods, not only detect the tumor, but also indicate its position. For clinical medicine, our method may assist doctor’s diagnosis, to increase the accuracy of detecting tumors and reduce its threats to our lives.
Appears in Collections:	[Executive Master of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Size	Format

社群 sharing

Loading...