| Summary: | 碩士 === 國立中興大學 === 資訊科學研究所 === 93 === There are many types of cells in peripheral blood and body fluid. Each type of cells should always maintain a certain proportion for a healthy person. When a person becomes ill, each type of blood cells may present a different proportion due to different disease. The most obvious change is in white blood cell (WBC). Therefore, WBC differential count (DC) provides very important information for body checkout on disease diagnoses. Today, complete blood count (CBC) is usually performed by using blood count instruments. However, blood count instruments are not appropriate to use for body fluid because of various thickness and too few in quantity. Thus, the differential counts for WBC in body fluid are usually carried out by medical inspection specialist through the help of microscope.
This research uses image processing technique to find the nucleus of WBC and then segment WBC from the smear images. According to the characteristics of various kinds of WBCs and knowledge obtained from clinical inspection specialist, complete recognition of leukocytes image and achieve automated differential blood count system by constructing a decision tree which that uses the different feature vector in the different branch point.
This research uses about 1,289 origin smear images. The smear images include peripheral blood, cerebrospinal fluid (CSF), synovial fluid, pleural fluid and peritoneal fluid. We compare the recognition results with the inspection results of senior medical inspection specialist in laboratory exam department of hospital. The recognition accuracy is up to 98.06%.
|
|---|
