Application of AI software for chromosomal aberration analysis in occupational health surveillance and radiation biological dose estimation
10.20001/j.issn.2095-2619.20250409
- VernacularTitle:染色体畸变分析AI软件在职业健康检查和辐射生物剂量估算中的应用
- Author:
Yingyi PENG
1
;
Qiuying LIU
;
Zhifang LIU
;
Zongjun ZHANG
;
Xiaoyan CHEN
;
Kunjie HUANG
;
Qiying NONG
;
Na ZHAO
Author Information
1. Guangdong Province Hospital for Occupational Disease Prevention and Treatment, Guangzhou, Guangdong 510300, China
- Publication Type:Journal Article
- Keywords:
Chromosomal aberration;
Artificial intelligence;
Radiation worker;
Occupational health examination;
Radiation;
Biological dose estimation
- From:
China Occupational Medicine
2025;52(2):171-175
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the feasibility of applying artificial intelligence (AI) technology in chromosomal aberration (CA) analysis for occupational health surveillance of radiation workers and in biological dose estimation during nuclear emergency responses. Methods Peripheral blood samples from healthy volunteers were irradiated in vitro with X-rays and cobalt-60 (⁶⁰Co) γ rays. Chromosome slides were prepared using an automated harvesting and dropping device. The data training and outcome evaluation of CA analysis was performed on the AI software using chromosome images from occupational medical examination of radiation workers from the current lab or chromosome slides from blood samples irradiated with X-rays. The trained AI software was then used to assist in CA analysis and biological dose estimation among occupational medical examination of radiation workers, with results compared with manual reading and actual exposure doses. Results The trained AI software achieved a CA recognition accuracy of 95.11%. In the occupational health examination of radiation workers, the positive CA detection rate using AI + manual review was 2.25% higher than that in manual reviewing alone. The errors in biological dose estimation for ⁶⁰Co γ rays and X-rays using AI + manual review analysis were 11.86% and 7.33%, respectively, both within the acceptable 20.00% error margin. Conclusion AI + manual review can be effectively applied in CA analysis for occupational health examination and biological dose estimation during nuclear emergencies, significantly improving analysis efficiency.
