Abstract:

BACKGROUND: The classification and identification of human chromosome is a basic mission of medical genetics. Analyzing and identifying human chromosome automatically with computer technology is an important research subject in human chromosome image analysis.
OBJECTIVE: To introduce basic principle, technique advantages and operating methods of artificial neural network and explore the applications of artificial neural network in chromosome automatic analysis.
METHODS: A computer-based online search of EBSCO database (http://search.ebscohost.com) and Wanfang database (http://www.wanfangdata.com.cn) was performed for articles about application of artificial neural network in chromosome automatic analysis system with the key words “chromosome, artificial neural network” in English and Chinese.
RESULTS AND CONCLUSION: The aim of studying chromosome automatic analysis is to reduce technician labor intensity and allow them free from repeated laboring. Finally, these systems are used in clinic for molecular cytogenetics identification and aristogenesis etc. Despite the significant research effort and progress of applications of artificial neural network in chromosome automatic analysis over these years, there are still some localizations. The classified chromosome database is needed, which is not easy to get for general researchers. In addition, the accuracy of the classification result is not good enough, which even cannot reach the level of well-trained cytology researchers. Moreover, the neural network has its inherent shortcomings such as huge training data and huge training time. Therefore, further research and development is required to improve and consummate the network through optimizing network structure, selecting effective characteristic, and reducing otiose operation.