Breast Tumor Identification with Gabor Filter on Sonogram
| Autor: | Chia-Chen Chang, 張家甄 |
|---|---|
| Rok vydání: | 2010 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 98 Evolution and use of ultrasonic technology are widely applied to the medical image. The ultrasound is the most popular way for making a diagnosis; however, a noisy ultrasound image will reduce its overall diagnostic efficiency. In order to real-time ultrasound scans for segmenting and identifying the tumors, the main purpose of my study is to develop an adaptive initial contouring method for segmenting tumors in ultrasound images, and using my proposed texture extraction method to provide a physician some tumor diagnosis information. In my research methodology, the image preprocessing is by the Gaussian low-pass filter to eliminate the speckle in ultrasound image. Such as method of image enhancement, my study used histogram equalization in the specific region of gray level distribution of the US images. Next, my paper used Level Set and Fast Marching Method with adaptive initial contouring to segment possible tumor blocks. Then according to twelve orientations and five kinds of frequency of Gabor filter, my research proposed a mean expectation method to extract feature of possible tumors. In addition, my proposed feature extraction was combined with Auto-correlation. Finally, input the statistical information to neural network to perform training. After obtaining knowledge bases, the tumors can be identified effectively. It is therefore clinically useful to improve the diagnostic efficiency and offer a second reading to assist physicians in medical diagnosis. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|