A hybrid RSM-PSO-ANN approach for developing cleaner-label emulsified sausages with roasted Lyophyllum decastes...

The development of cleaner-label meat products with reduced fat and enhanced nutritional value is a key trend in the food industry. This study developed a hybrid response surface methodology-particle swarm optimization-artificial neural networks (RSM-PSO-ANN) modeling approach to optimize emulsified sausages formulated with roasted Lyophyllum decastes mushroom powder (rLMP) for cleaner-label production. The model effectively captured the non-linear effects of fat (15-35%), rLMP (8-16%), and ice-water (30-50%) on sensory quality, with PSO-ANN showing superior prediction (R² = 0.9921) over RSM. The optimal formulation (25% fat, 12% rLMP, 40% ice water) yielded the highest sensory score. The optimized sausages exhibited improved emulsion stability (42.4% reduction in cooking loss) and enhanced texture. Furthermore, rLMP incorporation significantly increased antioxidant activity (DPPH: 37.1% to 77.2%; ABTS: 21.0% to 40.6%) and reduced lipid oxidation (65.6% lower TBARS). These findings validate rLMP as a multifunctional ingredient for developing healthier meat products, and the integrated modeling strategy provides an efficient tool for intelligent food formulation.