SERS based determination of ceftriaxone, ampicillin, and vancomycin in serum using WS2/Au@Ag nanocomposites and a 2D-CNN regression model

Abstract

Accurate therapeutic drug monitoring (TDM) of antibiotics including ceftriaxone, ampicillin, and vancomycin plays an important role in the treatment of neonatal sepsis, a common and life-threatening disease in neonates. A highly sensitive surface-enhanced Raman spectroscopy (SERS) method using tungsten disulfide/gold and silver core–shell (WS₂/Au@Ag) nanocomposites was developed for the rapid detection of the three antibiotics, with a wide response range (0.5–1000 μg/mL). A two-dimensional convolutional neural network (2D-CNN) regression model was proposed to predict antibiotic concentrations in complex mixed serum solutions, simulating various drug use scenarios. The model achieved excellent regression results for ceftriaxone and ampicillin simultaneously, with R-squared (R²) values of 0.9993 and 0.9997. The integration of ultra-sensitive SERS with the 2D-CNN based deep learning model provides a promising approach for rapid TDM and personalized patient treatment.