Low Field-NMR-based Prediction Of Chicken Egg Freshness

Low field nuclear magnetic resonance（LF-NMR）and magnetic resonance imaging（MRI）are becoming an viable approach to the measurement of agricultural and forestry products such as produce,grains,meat,poultry and aquatic products.Taking fresh chicken egg as the research subject,following an investigation of the mechanism of its freshness change,LF-NMR models of chicken fresheness were built with the aid of machine learnings to explore a new way to the nondestructive and rapid measurement of chicken egg freshness.For reference,freshness-indicating indices were collected using conventional physical and chemical measurements.Considering the biological structure,nutritents,and egg change in storage,the 3 freshness-indices including thick albumen height,Haugh unit（HU）,and yolk index were recorded and their correlations regarding the number of days in storage and between each other were investigated with regression analysis.The feasibility and mechanism were examined for chicken egg freshness measurement via LF-NMR and MRI.T₂ inversion spectra and MRI images were collected,and nuclear magnetic resonance attributes of chicken eggs were compared with freshness-indices.It was observed that with the increase of storage time,peak times T_2iand peak ratios S_2i changed,as both T₂₁ for bound water and T₂₂ for semi-bound water increased gradually,so did S₂₁ and S₂₂ slightly,as to free-water,T₂₃ decreased gradually,but S₂₃ dropped more gradually.These changes of LF-NMR attributes reflect the innernal migrations of water and the change of large molecular substances within chicken eggs during storage so that statistics were able to show that thick albumen height,HU,and egg yolk index all having good correlations withT₂₁,T₂₂,A₂₁ andA₂₂,supporting the feasibility of egg freshness measurement according toT₂ inversion spectra.The visual changes during storage of a subject’s contour of air chamber as well as the shape of yolk were also manifested in the MRI images acquired and pseudo-colored,showing stale eggs’biological traits.The prediction of egg freshness on LF-NMR was finally investigated using machine learnings.All 5 regression modelling algorithms,including univariate linear regression,multivariate linear regression,partial least squares regression,BP network regression,and support vector machine regression（SVR）,achieved good freshness-prediction results on LF-NMR attributes,among them,non-linear model is superior to linear model,SVR model has the best prediction,moreover,it has the best prediction effect on the internationally most authoritative physicochemical index of freshness-Haugh unit.As to LF-NMR-based automatic grading,SVR also excelled in grading for HU,with an accuracy of 80.6%.This work shows that LF-NMR can be used for the nondestructive measurement of chicken egg freshness,showing the technological feasibility of developing LF-NMR equipment for egg freshness measuring.