紋路分析在偵測乳房腫瘤的應用

不同於過去以乳房X光片的灰階值信息粹取灰階值強度統計圖（gray level histogram）和空間灰階值相關性（spatial gray level dependence）等相關參數作為描述腫瘤的特徵向量，本文改以針對像素的紋路單元之間灰階值變化的紋路頻譜（texture spectrum）和紋路特徵編碼（texture feature coding method）描述腫瘤的特徵向量，以達到腫瘤偵測的目的。由於紋路頻譜充分描述像素周圍八個鄰近像素之間一階灰階變化的幾何特性，而紋號特徵編碼則強調像素在水平及垂直二個方向灰階的二階變化，因此二者對捕捉腫瘤影像的灰階的微細變化，將遠勝於灰階值強度統計圖和空間灰階值相關的方法。從實驗結果顯示，在判斷乳房影像的腫瘤上，本文所採用的紋路特徵向量的描述方式，可以得到較佳的腫瘤判斷結果。因此紋路分析在腫瘤偵測及判斷上是一項值得採用的重要參數。

關鍵字

乳房攝影；腫瘤偵測；紋理特徵；紋路特徵編碼；紋路頻譜

並列摘要

In this article, feature selection based on the texture spectrum and texture feature coding method is considered to depict the characteristics of mass in mammograms for mass detection. It is quite unlike that of either the features of gray level histogram or the features of spatial gray level dependence that were often chosen as the feature vector of the mass in the past. The texture spectrum fully represents the first-order gray level change of a pixel in a texture unit (a cell composed by a center pixel and its eight connected pixels); meanwhile texture feature coding method emphasizes the second-order gray level change in the horizontal and vertical direction of a pixel in a texture unit. For the purpose of mass detection, both can capture the subtle of gray level change better than the conventional methods of gray level histogram and spatial gray level dependence. Experimental results evidently show that our proposed texture based feature vector can get better diagnostic judgment in mass detection and mass classification. It is worthy of note that texture analysis is recommended as important parameters in mass detection and mass classification in the computer-aided detection of mammograms.

並列關鍵字

Mammography ； Mass detection ； Texture feature ； Texture feature coding ； Texture specturm

國際替代計量

紋路分析在偵測乳房腫瘤的應用

全文下載

主題瀏覽