Integration of Mid and Near-Infrared Spectroscopy for Determining TPC and TFC in Wheat Varieties Via Multivariate Analysis

The current research has been focused on the quantification of total phenolic content (TPC) and total flavonoid content (TFC) based on spectral data collected from mid-infrared (MIR) and near-infrared (NIR) spectroscopy. Low-level fusion (NIR-MIRLL) and mid-level fusion (NIR-MIRML) were employed to enhance the statistical performance parameters with model partial least squares (PLS). In measuring the performance of the PLS regression models for TPC and TFC, the measures used were the coefficient of prediction (Rpred), root mean square error of prediction (RMSEP), and residual predictive deviation (RPD). The best results were noted for TPC concerning the NIR-MIRML fusion model, which yielded Rpred = 0.8998, RMSEP = 8.82 mg GAAC/100 g, and RPD = 2.20. Best results for TFC were noted with the NIR-MIRLL fusion model at Rpred = 0.9086, RMSEP = 2.28 mg/100g, and RPD = 2.35. The NIR-MIRLL and NIR-MIRML fusion models surpassed the single sensor models' results regarding TPC and TFC content in wheat flour. The outcome indicates that the proposed fusion strategy has been successful with enhanced prediction accuracy.