Smartphone-Based Microalgae Monitoring Platform Using Machine Learning

There is a growing demand for microalgaemonitoring techniquessince microalgae are one of the most influential underwater organismsin aquatic environments. Specifically, such a technique should behand-held, rapid, and easily accessible in the field since currentmethods (benchtop microscopy, flow cytometry, or satellite imaging)require high equipment costs and well-trained personnel. This study'smain objective was to develop a field-deployable microalgae monitoringplatform using only a single smartphone and inexpensive acrylic colorfilms. It aimed to evaluate the morphological states of microalgaeincluding stress, cell concentration, and dominant species. Usinga smartphone's white LED flash and camera, the platform detectedfluorescence and reflectance intensities from microalgal samples invarious excitation and emission color combinations. Multidimensionalintensity data were evaluated from the smartphone images and usedto train a support vector machine (SVM) based machine learning modelto classify various morphological states. The SVM classification accuracieswere 0.84-0.96 in classifying four- to five-tier stress types,cell concentration, and dominant species and 0.99-1.00 in classifyingtwo-tier stress types and cell concentrations. Additional field sampleswere collected from the local pond and independently tested usingthe laboratory-collected training set, showing two-tier classificationaccuracies of 0.90-1.00. This platform enables accessible andon-site microalgae monitoring for nonexperts and can be potentiallyapplied to monitoring harmful algal blooms (HABs).