| Summary: | 碩士 === 國立中興大學 === 資訊管理學系所 === 99 === This paper proposed an image segmentation and grading system for medical tissue images, which are related to the brain tumor, called astrocytoma. In this study, the features of brain tissue image are retrieved by using the segmentation methods to future recognizing which grade of brain tumor is. The segmentation methods are broken to three sub-methods including nuclear atypia, necrosis, and mitoses cells. After segmentation, the features, such as cell number, average area, and standard deviation will be obtained. There are different weights generated by Genetic Algorithm for each feature help in judging which grade of the processing image is.
During the segmentation process, each kind of cells is segmented according to their properties in an image; for example, nuclear atypia is with lower intensities, necrosis is with higher intensities as well as mitoses is with smoother area. Using the Genetic Algorithm, the optimal parameters will be used in the experiments.
It is the first idea to recognize the tumor grade in medical tissue image. Due to dealing with the huge amount of digital images is a time consuming task, we propose an automatic image segmentation and grading method to save time and reduce the human judgment errors. The experiments mainly focus on the accuracy of both results of segmentation and image grading. On average, the accuracy of image grading is higher than 85%.
|
|---|
