Identification and enumeration of cyanobacteria species using a deep neural network

Authors: BAEK, SANGSOO - Ulsan National Institute Of Science And Technology (UNIST); PYO, JONGCHEOL - Ulsan National Institute Of Science And Technology (UNIST); Pachepsky, Yakov; PARK, YOUMGEUN - Konkuk University; LIGARAY, MAYZONEE - Ulsan National Institute Of Science And Technology (UNIST); AHN, CHI-YONG - Korea Research Institute Of Bioscience And Biotechnology; KIM, YONG-HYO - Hanyang University; CHUNJONG AHN - Apec Climate Center (APCC); CHO, KYUNGHWA - Ulsan National Institute Of Science And Technology (UNIST)

Submitted to: Harmful Algae
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/6/2020
Publication Date: 4/16/2020
Citation: Baek, S., Pyo, J., Pachepsky, Y.A., Park, Y., Ligaray, M., Ahn, C., Kim, Y., Chunjong Ahn, Cho, K. 2020. Identification and enumeration of cyanobacteria species using a deep neural network. Harmful Algae. 115:106395. https://doi.org/10.1016/j.ecolind.2020.106395.
DOI: https://doi.org/10.1016/j.ecolind.2020.106395

Interpretive Summary: Counting cells of cyanobacteria in recreation, irrigation, processing, and other types of water is critical for evaluating the danger of toxic water contamination and making management decisions to mitigate the harmful algae impacts on human health. Cell detection and counting is currently done by trained operators using microscopes. These operations are inevitably slow and prone to operator’s errors. We set to research if automation of cyanobacteria cell recognition and counting can be achieved with the modern artificial intelligence technology. We found that five cyanobacteria species could be identified and counted using the novel AI technology of deep learning with the 85 to 95 percent accuracy. Results of this work open the possibilities of fast assessment of cyanobacteria populations that is needed to make the operational management decisions in the wide variety of water resource uses facing harmful consequences of cyanobacteria presence in waters.

Technical Abstract: Cell classification and cell counting are essential for the detection, monitoring, forecasting, and management of harmful algae populations. Conventional methods of algae classification and cell counting are known to be time-consuming, labor-intensive, and subjective, depending on the expertise of observers. The objectives of this study were to classify and quantify five cyanobacteria using the deep learning techniques of a fast regional convolutional neural network (R-CNN) and convolutional neural network (CNN). Images of the five cyanobacteria in inland water samples from the Haman weir of Nakdong River and Baekje weir of Geum River were used to classify cyanobacteria species using the fast R-CNN model and quantify their respective cells with the CNN model. The distinctive morphological features of the five cyanobacteria were extracted by the fast R-CNN model, which achieved reasonable agreement with the manual classification results, yielding average precision (AP) values of 0.929, 0.973, 0.829, 0.890, and 0.890 for Microcystis aeruginosa, Microcystis wesenbergii, Dolichospermum, Oscillatoria, and Aphanizomenon, respectively. The CNN model for the Microcystis species obtained an R2 value of 0.775 and RMSE value of 26 cells for training, and an R2 of 0.854 and RMSE of 23 cells for validation. A slight underestimation for a population with <50 cells and minor overestimation for >250 cells were observed, owing to cell overlapping and the presence of out-of-focus regions in the input images. Further research is required to enhance the cell-count accuracy using the CNN model. Overall, this study was able to demonstrate the reliable performance of cyanobacteria classification and cell counting using deep learning approaches.