Speaker
Description
Traditionally, the analysis of biological images has mostly relied on manual operations. In the context of this study, the experimental workflow involves at least the following steps: selecting samples, performing feature extraction, image segmentation to obtain regions of interest (ROI), generating plots and statistics, and conducting image visualization. However, with the rapid advancement of modern microscopy technologies and the resulting increase in data volume, manual processing would consume excessive time and labor, making it impossible to meet analytical demands.
To address this issue, computer-assisted AI technologies offer an effective solution for handling and interpreting large-scale image datasets. This study integrates two major open-source AI tools—Feature Extraction and Cellpose—to analyze fluorescence resonance energy transfer (FRET) images of live E. coli HB101 expressing the lead biosensor protein Met-lead 1.44 M1. Using these tools, the FRET signals and ratio data from each ROI are processed and converted into ratio images, which serve as the main analytical output of this research.
The image data were acquired using a portable lead biosensor device (pMet-lead) in combination with smartphone-based imaging. Fiji was used as the integrated platform for ratio calculation and AI tool integration. This platform combines machine learning–based feature extraction to automatically align YFP and CFP images, and deep learning–based Cellpose for automated ROI identification. After training, Cellpose demonstrated improved accuracy and sensitivity in ROI segmentation compared to pre-training performance. Custom scripts were implemented in Fiji to automate the workflow.
Overall, applying AI-driven automated analysis of biosensing images significantly improves the efficiency and accuracy of data processing. This approach can not only assist researchers but also enable the general public to easily use handheld sensing devices to detect specific substances in real time, achieving practical on-site testing. Furthermore, automation enhances data analysis efficiency and can be applied to a wide range of biological samples. In the future, integrating cloud computing could allow smartphone-captured images from the portable lead biosensor to be uploaded to a server for processing, with results then returned to the user’s device. Ultimately, the combination of mobile biosensors and smartphones will make on-site water quality monitoring more accessible and efficien
