Autoencoder-Based Method for Enhancing and Manipulating Surface Plasmon Resonance Sensor Responses

An autoencoder-based approach for surface plasmon resonance (SPR) image compression and representation is presented. Integrated into a portable SPR sensor, the encoder/decoder tasks were performed across various latent vector dimensions to evaluate performance. The impact of the compressed layer on sensor functionality is analyzed, highlighting its influence on overall performance. Images of aqueous solutions were used to extract SPR reflectance curves and to determine the sensitivity figure of merit. This new image representation method reduces the overhead associated with processing, displaying, and transmitting sensor response data. Additionally, it supports the development of new intelligent tasks, particularly for field-deployment applications.