A multilevel cooperative attention network of precise quantitative analysis for predicting ractopamine concentration via adaptive weighted feature selection and multichannel feature fusion

• Published in: Food chemistry 2025-02, Vol.464 (Pt 3), p.141884, Article 141884
• Main Authors: Yin, Tianzhen, Peng, Yankun, Li, Yongyu, Chao, Kuanglin, Nie, Sen, Tao, Feifei, Zuo, Jiewen
• Format: Article
• Language: English
• Subjects: Hierarchical gradient aware spectral network; Multi-channel peak attention mechanism; Ractopamine; SERS; Spectral gradient weighted attention
• ISSN: 0308-8146; 1873-7072; 1873-7072
• DOI: 10.1016/j.foodchem.2024.141884

Surface enhanced Raman spectroscopy (SERS) holds great potential due to its rapid detection and high sensitivity. However, issues such as signal noise, fluctuations, and spectral shifts can negatively impact its performance in detecting ractopamine in pork. Hierarchical Gradient Aware Spectral Network (HGASNet) was proposed to address these issues. The key innovations are the Spectral Gradient Weighted Attention (SGWA) and Multi-Channel Peak Attention Mechanism (MCPAM) modules within HGASNet. The SGWA module dynamically adjusts feature weights, enhancing sensitivity to critical Raman spectral shifts. Meanwhile, MCPAM leverages multi-channel attention to better capture long range dependencies and fuse global and local information. Additionally, HGASNet's hierarchical structure incrementally extracts and integrates features at various levels, enabling the model to focus on global features in the higher layers while preserving fine-grained local details in the lower layers. Experimental results show HGASNet outperforming existing approaches, achieving R2 of 0.9972, RMSRE of 0.0413, RMSLE of 0.0548, sMAPE of 6.47 %, and MAPE of 6.72 %. •Gradient-based dynamic weighting calculates Raman shift contributions.•Multi-channel attention improves long-range dependency capture via shared weight.•Hierarchical structure enhances focus on global and local spectral details.•Progressive feature selection increases model predictive accuracy and robustness.