Pathogen Recognition via Serface-Enhanced Raman Scattering: Self-Supervised Learning and Transfer Learning approaches

Surface-enhanced Raman Scattering (SERS) has emerged as a promising analytical technique owing to its ability to provide molecular fingerprints while minimizing water interference. However, despite the increasing availability of public SERS datasets, the scarcity of labeled data remains a significant obstacle to the adoption of deep learning approaches. In this study, we proposed a self-supervised representation learning to improve data efficiency in label-scarce scenarios. We present a novel application of the Bootstrap Your Own Latent (BYOL) algorithm algorithm to SERS data, enabling the extraction of meaningful spectral representations and enhancing performance in downstream classification tasks. The proposed framework establishes a foundation for more robust and scalable SERS-based applications, particularly in the development of intelligent biosensing systems.