Liver Cell Nucleus Segmentation and Fibrosis Grading Recognition System for Liver Tissue Images

• Main Authors: Ying-Siou Wang, 王瀅琇
• Other Authors: Yung-Kuan Chan
• Format: Others
• Language: en_US
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/67760551619464785117

碩士 === 國立中興大學 === 資訊管理學系所 === 100 === This study proposes image segmentation and grading recognition system for liver fibrosis tissue images. It is a new idea to recognize the grades by extracting the object features in liver fibrosis tissue images. In this thesis, two objects of nucleus and vacuole for liver tissue images are retrieved by using the segmentation methods, and then the features are used to recognize which grade of liver fibrosis image is. In segmentation phase, some image processing methods are applied to segment the object regions of nucleuses and vacuoles. For nucleus regions segmentation, double thresholds are set to enhance the contrast. Run Length method makes the object regions become obviously and the noise be suppressed, and then morphological Opening operation is performed to split connecting objects. For vacuole regions segmentation, morphological Opening operation is also used and the mode filter is applied to fill up the dark holes in the objects to keep the completeness of regions. In recognition phase, the features of objects, such as the number, area, mean and standard deviation, will be evaluated. K-means Algorithm is performed to classify the grade of images, and let the distance of features between the image and group center is shortest. Furthermore, there are different weights generated by Genetic Algorithm for each feature, than it can used to find the optimal results in recognition. The goal of this research is to provide objective recognition results for medical experts by proposing this automatic image segmentation and grading system in liver fibrosis. Hence, the human judgment errors ,the costs of time, and human resources can be reduced. From the experimental results, the proposed segmentation method can achieve a good performance and the recognition method can efficiently recognize the grade of liver fibrosis tissue image.