Abstract
This thesis advances seafood processing by applying deep learning to Rapid Evaporative Ionization Mass Spectrometry (REIMS) data, enabling automated and accurate marine biomass analysis. It addresses critical industry challenges including species identification to combat mislabeling fraud, body part classification for by-product utilization, oil contamination detection, cross-species adulteration detection, and batch traceability. Key contributions include Transformer and Mixture of Experts (MoE) architectures achieving up to 100% accuracy in species identification, SpectroSim — a self-supervised contrastive learning framework for label-free batch traceability (70.8% accuracy) — and explainable AI integration via LIME and Grad-CAM for interpretable predictions.
Key Results
Deep learning methods consistently outperform the OPLS-DA baseline across all five analytical tasks:
| Task | Baseline (OPLS-DA) | Best Model | Gain |
|---|---|---|---|
| Fish Species | OPLS-DA: 96.39% | MoE Transformer: 100.00% | +3.61% |
| Fish Body Part | OPLS-DA: 51.17% | Ensemble Transformer: 74.13% | +22.96% |
| Oil Contamination | OPLS-DA: 26.43% | TL MoE Transformer: 49.10% | +22.67% |
| Cross-species Adulteration | OPLS-DA: 79.96% | Pre-trained Transformer: 91.97% | +12.01% |
| Batch Detection | OPLS-DA: 53.19% | SpectroSim (Transformer): 70.80% | +17.61% |