Deep learning-enabled ratiometric signal transduction for portable and intelligent colorimetric LAMP biosensing of Vibrio vulnificus.

Vibrio vulnificus is a highly lethal zoonotic pathogen associated with seafood consumption and aquaculture environments, demanding rapid and reliable on-site molecular diagnostics. Here, we report an intelligent biosensing platform utilizing a deep learning-enabled ratiometric signal transduction strategy for the automated detection of pathogens. This platform integrates loop-mediated isothermal amplification (LAMP) with a convolutional neural network (CNN)-based quantification framework to convert colorimetric biochemical reactions into robust digital outputs. To address the inherent limitations of conventional visual interpretation, our system employs a custom-designed bottom-illumination optical module that ensures uniform signal acquisition. A ratiometric analysis of the green-to-blue (G/B) channel intensity is utilized as a standardized transduction metric, effectively suppressing optical noise and variability caused by fluctuating environmental illumination. This AI-augmented sensing approach improves analytical sensitivity by one order of magnitude compared with traditional gel electrophoresis, achieving a detection limit of 10-5 ng/μL under extended amplification conditions with a quantitative correlation (R2 = 0.915). The deep learning model further provides reliable biosignal classification with a precision of 0.953 and a recall of 0.972. The entire assay requires only 5 μL of sample and delivers results within 30 min. By synergizing molecular engineering with intelligent signal processing, this work establishes a generalizable paradigm for converting subtle colorimetric changes into quantifiable and objective sensing data. This platform offers a scalable and robust route toward next-generation portable diagnostic sensors for field and resource-limited settings.