Dynamic surface-enhanced Raman spectroscopy and positively charged probes for rapid detection and accurate identification of fungal spores in infected apples via deep learning methods

Abstract

Fungal infections pose a significant threat to apples; therefore, the detection of fungal spores is imperative for controlling infection spread and ensuring food safety. In this study, dynamic surface-enhanced Raman spectroscopy (D-SERS) and positively charged probes were developed to detect and identify the fungal spores via deep learning methods. Firstly, the gold nanorods were modified with cysteamine to develop the positively charged SERS probes, enhancing the capture of fungal spores by promoting interactions with the negatively charged cell wall. Then, the probes and D-SERS were combined to measure the SERS spectra of fungal spores, and the optimal spectral signals were obtained under the metastable state of D-SERS from wet to dry. This was due to capillary forces inducing nanoparticles to form a large number of 3D hot spots, resulting in significant enhancement. Spores of Aspergillus flavus, Rhizopus stolonifer, and Botrytis cinerea can be easily detected with excellent SERS signals from infected apples after simple separation through filtration and centrifugation. Furthermore, the best recognition model was developed by ZFNet, a powerful deep learning method, with the accuracies in the training set, validation set, and prediction set of 100%, 99.44%, and 99.44%, respectively. The proposed method provides a simple, rapid, and accurate approach for the detection and identification of fungal infections in apples, and can be extended to other agricultural products.