Home    About Journal    Editorial Board    Instruction    Subscribe    Download    Messages Board    Contact Us    中文
  Apr. 5, 2025  
Quick Search
Chinese Journal of Biomedical Engineering  2020, Vol. 39 Issue (1): 26-32    DOI: 10.3969/j.issn.0258-8021.2020.01.04
Regular Papers Current Issue | Archive | Adv Search |
Intelligent Classification and Identification of Composite Colonies Based on Convolutional Neural Network
Yu Hui1#, Du Peipei1, Liu Xiang2, Liu Zheng1, Zhu Xianfeng1, Cao Yuzhen1*
1(Department of Biomedical Engineering,Tianjin University,Tianjin 300072,China)
2(Institute of Tianjin Food Safety Inspection Technology Research,Tianjin 300000,China)
Download: PDF (4576 KB)   HTML (1 KB) 
Export: BibTeX | EndNote (RIS)      
Abstract  To meet the requirement of intelligent morphology classification of compound colonies, a convolution neural network for colony classification was proposed in this work. All connected domains in the culture dish were obtained by level set evolution. Limit corrosion was used to determine the connected domains that seed points were greater than one. Obtaining the convex closures of the connected domains, the concave points were detected. By connecting the corresponding concave points, the connected domains with adhesion were segmented. Herein 600 single colony samples were normalized, and the data were expanded to 30,000 cases by rotating and flipping and superimposing random noise with signal-to-noise ratio not exceeding 5%. Samples of 70% were used as the training set for the network. The network was cross-verified by 10 folds. Samples of 30% were used as the test set. The weighted average accuracy of the four colonies reached 87.50%, of which the classification accuracy of spotted smooth colonies was 86.40%, that of circle like wavy colonies was 87.21%, that of oval colonies was 88.11%, and that of irregular other colonies was 87.25%. Compute unified device architecture (CUDA) was adopted to accelerate the algorithm. The running time was optimized to 1/10 of the original time. In terms of running speed and convenience, the algorithm far exceeded the traditional methods. The statistical results showed that the proposed method could effectively complete the task of intelligent classification and identification of compound colonies. This method showed good expansibility and self-learning function, which had reference value for intelligent analysis of biochemical samples based on images.
Key wordslevel set      division of adhesion      convolutional neural networks      colony morphology      compute unified device architecture(CUDA)     
Received: 18 April 2019     
PACS:  R318  
Service
E-mail this article
Add to my bookshelf
Add to citation manager
E-mail Alert
RSS
Articles by authors
Yu Hui
Du Peipei
Liu Xiang
Liu Zheng
Zhu Xianfeng
Cao Yuzhen
Cite this article:   
Yu Hui,Du Peipei,Liu Xiang, et al. Intelligent Classification and Identification of Composite Colonies Based on Convolutional Neural Network[J]. Chinese Journal of Biomedical Engineering, 2020, 39(1): 26-32.
URL:  
http://cjbme.csbme.org/EN/10.3969/j.issn.0258-8021.2020.01.04     OR     http://cjbme.csbme.org/EN/Y2020/V39/I1/26
Copyright © Editorial Board of Chinese Journal of Biomedical Engineering
Supported by:Beijing Magtech