| Autor: |
Kazuma KUROTANI, Yoshiharu MIYATA, Isamu NISHIDA |
| Jazyk: |
angličtina |
| Rok vydání: |
2024 |
| Předmět: |
|
| Zdroj: |
Journal of Advanced Mechanical Design, Systems, and Manufacturing, Vol 18, Iss 4, Pp JAMDSM0037-JAMDSM0037 (2024) |
| Druh dokumentu: |
article |
| ISSN: |
1881-3054 |
| DOI: |
10.1299/jamdsm.2024jamdsm0037 |
| Popis: |
The bone marrow examination is an essential investigation, from diagnosing many hematologic diseases to evaluating the effectiveness of treatment. Evaluation of bone marrow smears in bone marrow examination requires microscopic counting by a clinical laboratory technician. To obtain reliable test results, they must acquire extensive experience and high technical skills. However, this current method limits the number of examinations due to a chronic shortage of human resources. To solve this issue, various studies have been conducted on methods utilizing both deep learning and image processing technology for the automatic classification of blood cells. When incorporating these methods into actual clinical examinations, medical technologists must manually obtain the images of bone marrow smear surface at high magnification utilizing an optical microscope. However, this process significantly consumes time and labor. Thus, the purpose of this study is to develop an automatic imaging system for bone marrow smear surfaces by combining an imaging instrument and a method for determining a focused image. Moreover, based on the results of analyzing the color features of blood cells, this paper proposes a system that generates images suitable for automatic classification of blood cells by extracting individual blood cells utilizing image processing technology such as the watershed algorithm. To validate the effectiveness of the system, we conducted a case study. The results confirmed that the system could automatically obtain images of bone marrow smear surfaces, and 3114 blood cells were successfully extracted from 20 images (155 cells per image). The number of these that could be classified by the technician was 2871 (92.2%). |
| Databáze: |
Directory of Open Access Journals |
| Externí odkaz: |
|
