Digital imaging-in-flow (FlowCAM) and probabilistic machine learning to assess the sonolytic disinfection of cyanobacteria in sewage wastewater
Published in Journal of Hazardous Materials, 2024
Recommended citation: Jaffari, Z. H., Na, S., Abbas, A., Park, K. Y., & Cho, K. H. (2024). Digital imaging-in-flow (FlowCAM) and probabilistic machine learning to assess the sonolytic disinfection of cyanobacteria in sewage wastewater. Journal of Hazardous Materials, 468, 133762. https://doi.org/10.1016/j.jhazmat.2024.133762
Toxic blue-green algae in sewage have to be brought down to safe levels before the water is released, which means knowing not just how many cells are left but whether those cells are still alive. An instrument that photographs and counts individual cells as they flow past made it possible to watch directly what treatment with ultrasound does to them. Once the conditions were tuned, the treatment destroyed more than 80% of the cells and left a further 10–15% visibly injured — a distinction that plain counting would have missed entirely. For the first time, these image-based measurements were used to train models that predict disinfection performance in advance. Examining those models showed that the length of the ultrasound treatment alone accounts for roughly half the effect.