Toward smart and in-situ mycotoxin detection in food via vibrational spectroscopy and machine learning

Recent advances in vibrational spectroscopy combined with machine learning are enabling smart and in-situ detection of mycotoxins in complex food matrices. Infrared and spontaneous Raman spectroscopy detect molecular vibrations or compositional changes in host matrices, capturing direct or indirect mycotoxin fingerprints, while surface-enhanced Raman spectroscopy (SERs) amplifies characteristic mycotoxins molecular vibrations via plasmonic nanostructures, enabling ultra-sensitive detection. Machine learning further enhances analysis by extracting subtle and unique mycotoxin spectral features from information-rich spectra, suppressing noise, and enabling robust predictions across heterogeneous samples. This review critically examines recent sensing strategies, model development, application performance, non-destructive screening, and potential application challenges, highlighting strengths and limitations relative to conventional methods. Innovations in portable, miniaturized spectrometers integrated with cloud computation are also discussed, supporting scalable, rapid, and on-site mycotoxin monitoring. By integrating state-of-art vibrational fingerprints with computational analysis, these approaches provide a pathway toward sensitive, smart, and field-deployable mycotoxin detection in food.