A 2D-digital spectral sensing method for rapid antibiotic detection in water

Antibiotics (ABs) have emerged as a global emerging pollutant, and the efficient detection of ABs is of paramount importance for ecological and health risk management. Traditional detection instruments, operating under the principle of "separation first, analysis later," are characterized by intricate procedures and high costs. As a result, they struggle to meet the demands of rapid monitoring in practical scenarios. In this study, we present a novel method for the rapid detection of ABs, integrating two-dimensional digital spectral images (2D-DS images) with deep learning techniques. This approach utilizes an innovative spectral imaging system and combinatorial chemical probes to transform complex sample information into 2D-DS images with 2D holographic spectra. Concurrently, a Digital Spectral Convolutional Neural Network (DSCNN) model is developed. This model enables end-to-end quantitative analysis, directly inferring antibiotic concentrations from 2D-DS images. The model demonstrates a high predictive accuracy, with an R² value ranging from 0.85 to 0.93, and a detection limit as low as 1.94 mg L^-1. Building upon these findings, we further establish a rapid detection platform for ABs in water. Leveraging the fourth research paradigm, which is driven by spectral big data, this method effectively deciphers complex samples by exploiting the correlations in the optical information of 2D-DS images. The resulting monitoring solution is characterized by high throughput and low cost, with a single analysis taking approximately 3 min. This study thus provides a solid technical foundation for advancing the development of intelligent environmental monitoring networks.