Scaling up phytolith preparation protocols for big data

Abstract

Analytical procedures in phytolith analysis are transforming with the potential of AI to automate phytolith identification and classification, facilitating the analysis of unprecedented large datasets. Digital microscope scanners are the most cost-effective and viable means to enable big data acquisition; however, they require new slide mounting methodologies for full automatization. Currently, phytolith extraction protocols are inadequate for processing the large numbers of samples it is becoming possible to analyse. This research optimises phytolith extraction workflows to process large numbers of samples and develops a new methodology for phytolith slide mounting adapted to digital scanning. This research develops workflows for optimising phytolith extraction protocols to handle large sets of samples, by reducing the human labour time required in established methodologies. Phytolith slide mounting procedures are tested to optimise the protocol for digital slide scanning. Human-labour time was reduced by up to 33% during the phytolith extraction process while maintaining comparable extract quality. Phytoliths mounted in Norland 165H and left for 24 hours before setting were best suited for digital scanning. Phytolith extraction and slide mounting workflows have been optimised to handle large data sets suitable for digital scanning and are fundamental in the development of automated phytolith identification.

Keywords: Phytoliths, Artificial Intelligence, Automation, Extraction, Digital scanning