Relationship between sensory and NIR spectroscopy in consumer preference of table grape (cv Italia)

tA combination of near infrared spectroscopy (NIR) instrumental measurements and sensory analysiswas investigated to predict solids soluble content (SSC, assessed as Brix) and to classify preference intable grape cv Italia. SSC was monitored in each berry of whole bunches in order to evaluate intra-bunch distribution and variability. NIR spectra were recorded in the spectral region 12,000–4000 cm−1(833–2500 nm) using a set of 682 berries. The Partial Least Square (PLS) model based on cross-validationprovided acceptable value for the main statistical parameters (coefficient of determination of cross-validation, r2: 0.85; standard error of cross-validation, SECV: 1.08; residual predictive deviation, RPD:2.6) and was confirmed by external validation performed with 115 independent berries (coefficient ofdetermination of prediction, rp2: 0.82; standard error of prediction, SEP: 0.83). For consumer testing, theselected PLS model was used to predict the Brix value in 400 berries and Discriminant Analysis (DA) wasthen carried out to classify berries in terms of preference by relating NIR data to consumer judgment.The three defined preference clusters of berries were fully classified obtaining 100% membership. Incross-validation the value decreased especially for class 1 (78.5%) and 3 (75%) whereas class 2 obtainedcomparable values (98.7%). According to our results, NIR technology appears to be a promising techniquefor predicting SSC and obtaining information with regard to consumer preference in ‘Italia’ table grapefor application of efficient and low cost on-line instruments in the fruit industry.