| Summary: | 碩士 === 國立中正大學 === 資訊工程研究所 === 89 === Spiculate is a kind of stellate distortion cased by the intrusion of the breast cancer. Its existence is an important clue to characterize malignant tumors. So far there are several methods proposed to detect tumors with spiculation on mammography. The objective of this paper is to identify such features on ultrasonic volume data. In the proposed method, each horizontal flat slice in the volume data is successively extracted and then operated as an 2D ultrasound image by the detecting method respectively. First, in each horizontal slice the modified ROSE operation is utilized for finding the region where spiculate lines concentrate. Second, we use four directions of short line segments to estimate the direction of each pixel. A pixel which concentrates the region is marked and denoted as part of speculation. Finally, we collect the marked pixels around the region, and then draw a histogram to analyze its distribution. This method was tested on 100 samples inclusive of 50 having tumors with spiculation and the recognition rate was over 90%.
|
|---|
