An Image-Based Entamoeba Automatic Detecting System

• Main Authors: Tsung-Ho Wu, 巫宗和
• Other Authors: 黃德成
• Format: Others
• Language: en_US
• Online Access: http://ndltd.ncl.edu.tw/handle/63461978634661162082

碩士 === 國立中興大學 === 資訊科學與工程學系所 === 97 === The amoeba is a parasite which can compromise the human’s health. In every year, there are a lot of cases inflect by amoeba and even cause death. Thus, pathologists pay more attention to find the method of diagnosing the amoeba. Traditionally, they use microscopy to execute the diagnosis in experiment. There are two points to be checked: one is whether to have amoeba, and the other is the number of cell nuclei. The pathologists use the results to diagnose whether he (or she) was inflected and his (or her) infection degree. However, the diagnosing quality and performance are heavily impacted by human’s behaviors such as eyesight, strength and professional knowledge etc. If the computer technique can help pathologist to diagnose, we believe that it will decrease the influence of human behaviors, and increase the diagnosis correctness and performance. Thus, we proposed an image-based amoeba automatic detecting system. The main purposes of our system are automatic detection whether he (or she) was inflected and his (or her) infection degree. The pathologist will do the diagnosis further after we have marked this probably images. Our proposed system has two stages: one is amoeba cell detecting and the other is cell nuclei detecting. In the first phase, we have to execute the preprocessing, enhancing and detecting for the amoeba images to obtain the image of cell. In the second phase, we also use the same concept to detect cell nuclei for cell images. In our experimental results, and we are able to obtain the number of cell or nuclei and their locations for 75 amoeba images. The TPR is 98.73% of detecting cells, and the TPR of detecting nuclei is 90.63%. Hence, our proposed system is efficiency for detecting cells and cell nuclei.