Review of AI-Enabled In-Flow Imaging Technologies for Early Detection of Cyanobacterial Taste-and-Odor Metabolite Risks in Water Supply Systems

Taste and odor (T&O) compounds produced by cyanobacteria pose significant challenges for drinking-water management due to their extremely low detection thresholds. Increasing bloom frequency further elevates T&O risk, yet existing monitoring approaches are poorly suited to early warning. Chemical analytical methods provide reliable quantification of T&O compounds but are slow and detect odorants only after release. Fluorometric probes and remote sensing offer high-frequency signals but rely on pigment proxies without taxonomic resolution, while manual microscopy provides species-level information but is slow and requires taxonomic expertise. This review synthesizes how imaging-in-flow technologies can bridge these gaps by generating high-throughput, high-resolution images of individual cyanobacterial cells to support near real-time, taxonomically resolved monitoring of bloom composition. Commercial platforms (e.g., FlowCam and Imaging FlowCytobot) are critically evaluated alongside emerging approaches such as digital holographic microscopy and optofluidic time-stretch imaging, highlighting trade-offs that affect deployment, throughput, and image quality. Because operational value depends on rapid interpretation of image streams, artificial intelligence methods for automated analysis are examined, contrasting traditional feature-based classifiers with deep learning models. Gene-based molecular assays, particularly quantitative PCR, are discussed as a complementary layer linking observed taxa to odorant-production potential. Finally, future research directions toward integrated, proactive T&O monitoring frameworks are outlined.