Dynamic characterization of volatile profiles and intelligent discrimination of raw Milk under frozen storage using...

During frozen storage, the volatile profile of raw milk undergoes great changes, leading to quality deterioration, highlighting the need to identify key markers during this process. In this study, gas chromatography-time-of-flight mass spectrometry (GC-ToF/MS) combined with electronic nose (E-nose) and machine learning (ML) was applied to investigate the evolution of the volatile profile in raw milk during frozen storage. The results showed that acids dominated milk volatile organic compounds (VOCs) during storage, reaching 75-78%. Multivariate analysis revealed clear differences among storage stages and identified key VOCs. Furthermore, ML models based on optimized sensor features achieved high classification accuracy, with SVM and MLP models showing superior performance. Correlation analysis further revealed strong associations between selected E-nose sensors and key VOCs, indicating hexanoic acid, octanoic acid, 2-butanone, and dimethyl disulfide as key markers. This integrated strategy provides a reliable basis for understanding volatile changes and assessing the rapid freshness of frozen milk.