Computer-aided Pattern Recognition for Tuberculosis Bacteria
| Autor: | Yu-shu Chen, 陳玉書 |
|---|---|
| Rok vydání: | 2008 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 96 Tuberculosis disease is a major public health issues in Taiwan. As medical technology advances, in the prevention and treatment of tuberculosis also have a considerable effect, the mortality rate has decreased year by year, but there are still 15,000 new cases every year. In disease diagnosis, it’s required by professional doctors to diagnosis bacillus picture take by optical microscope, this way to spend a lot of time and cost of medical care for diseases diagnosis. Therefore, this research presents a method for recognizing tuberculosis bacteria at acid-fast stain based on features of bacillus. Our methodology can be divided into four steps: image preprocessing, image segmentation, feature extraction, and classification. First, image preprocessing consists of color transformation, color-to-grayscale transformation, bacillus preprocessing, and noise removal. In second step, we use chain code to obtain individual bacillus contours and to remove noise. Third is feature extraction, based on cytology, we extract thirteen features from every single bacillus. They are: area, perimeter, compactness, major & minor axis length, eccentricity, and Hu’s moments (φ1~7). The last step is classification, we use three classifying tools to classify these cells into two classes. Tools we used in this step are Back-Propagation Neural Network, Artificial NeuroMolecular System and Decision Tree. The results show that, this method can be successful recognize pulmonary tuberculosis from noise environment. But in the identification of Mycobacterium tuberculosis and non-Mycobacterium tuberculosis, furthermore pulmonary tuberculosis and tuberculosis complicated with diabetes disease among the different, and couldn’t use the pattern recognition technology to distinguish between effectively, perhaps in this area of medical decision support still be dependent on the area of medical research and analysis. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|