Nanosensors Convert H2O2 to Machine-Learnable Thermal Signatures in Plants

Abstract

Abiotic stresses, such as drought and salt stress, can significantly affect plant growth, posing substantial threats to agricultural productivity. As a central signaling mediator in plant adversity response mechanisms, real-time monitoring of the spatiotemporal dynamics of hydrogen peroxide (H 2 O 2 ) Current monitoring is essential for research in plant phenology. In this study, we developed an innovative detection method, which uses nanosensors to convert endogenous H 2 O 2 fluctuations at sub-micromolar concentrations into infrared thermal signals that can be learned by a machine, and processes these thermal imaging data through an advanced deep learning architecture to enable the monitoring of plant exposure to This enables non-invasive in situ monitoring of H 2 O 2 in plants under stress. Experimental validation shows that the average accuracy of multiple deep learning architectures reaches 98.8% and 99.6% under the challenges of drought and salinity stress test sets, respectively. In contrast to traditional methods, our focus is no longer on processing data and improving models in one go, but rather on the source of the data, which greatly improves the classification accuracy by acquiring thermal data with distinctive features. This integration of interdisciplinary techniques provides a non-destructive, rapid and accurate method for the early-stage stress monitoring of various plant stresses, and offers a new perspective for the study of plant stress characterization.

Code and data availability

The supplied blocks describe a custom thermal image dataset and CNN models, but contain no public deposit, repository, or availability statement for the data, code, or trained models. All URLs in the blocks are literature citations, not paper-specific asset locations.

No evidence-backed public reproduction asset is currently recorded.