Automatically Generate a Simplified Chest Atlas from the Chest Computed Tomography Using Morphology, Image Segmentation and K-means

• Main Authors: Yue-Yang Tsai, 蔡岳洋
• Other Authors: 殷堂凱
• Format: Others
• Language: en_US
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/dravsf

碩士 === 國立高雄大學 === 資訊工程學系碩士班 === 97 === With the improvement of technology, computers have given human beings a lot of improvement. These improvements include in our lifestyles and entertainment. Especially in the medical field, computers have given doctors a much more accurate diagnosis of patients. This paper deals with different patient’s chest CT (Computerized Tomography), which will give an analysis according to the human anatomy. When CTs are taken there will be outside influences, for example platforms will be photographed as well. Using a method similar to the method of opening which is a recursive erosion and dilation techniques, this would allow outside influences to be eliminated. After this, Sobel edge detection is used to get the outline of the pictures. Use Otsu’s method to automatically get the gray level threshold, and eliminate muscles or organs which have a lower gray level. This is due to that muscles and organs have lower gray levels. This would leave cavities and bone cavities to appear on the CT. Finally, use K-Means to cluster. Each cluster will be a center point. Thus, the center points will give a clear view of the locations of all parts. In conclusion, the mean of total Euclidean distances between the labeled body parts and the ideal body positions is 10.5582 pixels.