Accelerated Spoilage Simulation of Sea Bream Using Odor Profiling, Thermal Imaging, and Machine Learning

This study aimed to simulate an accelerated spoilage scenario for sea bream ( Sparus aurata ) by exposing the samples to ambient thermal abuse conditions (28°C ± 3°C), representing uncontrolled summer storage after cold chain failure. A low‐cost electronic nose (e‐nose) with metal oxide sensors (MQ‐series, MG811, TGS813), combined with thermal imaging and machine learning (ResNet‐50/SVM), was utilized to monitor spoilage progression in real time. Odor changes detected by MQ131 (+176.6 units at 6 h) and thermal patterns revealed significant degradation, particularly in the tail region. Color alterations (ΔE76 > 3.5) and t‐SNE clustering supported visual quality loss. The model achieved classification accuracies of 86.67%–100%. This integrated system offers a practical, scalable solution for rapid quality control in seafood logistics under thermal abuse scenarios where refrigeration may be compromised.