Unified Classification of Bacterial Colonies on Different Agar Media Based on Hyperspectral Imaging and Machine Learning.

A universal method by considering different types of culture media can enable convenient classification of bacterial species. The study combined hyperspectral technology and versatile chemometric algorithms to achieve the rapid and non-destructive classification of three kinds of bacterial colonies (Escherichia coli, Staphylococcus aureus and Salmonella) cultured on three kinds of agar media (Luria-Bertani agar (LA), plate count agar (PA) and tryptone soy agar (TSA)). Based on the extracted spectral data, partial least squares discriminant analysis (PLS-DA) and support vector machine (SVM) were employed to established classification models. The parameters of SVM models were optimized by comparing genetic algorithm (GA), particle swarm optimization (PSO) and grasshopper optimization algorithm (GOA). The best classification model was GOA-SVM, where the overall correct classification rates (OCCRs) for calibration and prediction of the full-wavelength GOA-SVM model were 99.45% and 98.82%, respectively, and the Kappa coefficient for prediction was 0.98. For further investigation, the CARS, SPA and GA wavelength selection methods were used to establish GOA-SVM simplified model, where CARS-GOA-SVM was optimal in model accuracy and stability with the corresponding OCCRs for calibration and prediction and the Kappa coefficients of 99.45%, 98.73% and 0.98, respectively. The above results demonstrated that it was feasible to classify bacterial colonies on different agar media and the unified model provided a continent and accurate way for bacterial classification.

Establishment of validated models for non-invasive prediction of rectal temperature of sows using infrared thermography and chemometrics.

Rectal temperature is an important physiological indicator used to characterize the reproductive and health status of sows. Infrared thermography, a surface temperature measurement technology, was investigated in this study to explore its feasibility in non-invasive detection of rectal temperature in sows. A total of 124 records of rectal temperature and surface temperature in various body regions of 99 Landrace × Yorkshire crossbred sows were collected. These surface temperatures together with ambient temperature, ambient humidity, and wind speed in pig pens were correlated with the real rectal temperature of sows to establish rectal temperature prediction models by introducing chemometrics algorithms. Two types of models, i.e., full feature models and selected feature models, were established by applying the partial least squares regression (PLSR) method. The optimal model was attained when 7 important features were selected by LARS-Lasso, where correlation coefficients and root mean squared errors of calibration were 0.80 and 0.30 °C, respectively. Particularly, the validity and stability of established simplified models were further evaluated by applying the model to an independent prediction set, where correlation coefficients and root mean squared errors for prediction were 0.80 and 0.35 °C, respectively. The validation of established models is scarce in previous similar studies. Above all, this study demonstrated that introduction of chemometrics methodologies would lead to more reliable and accurate model for predicting sow rectal temperature, thus the potential for ensuring animal welfare in a broader view if extended to more applications.

Application of invasive weed optimization and least square support vector machine for prediction of beef adulteration with spoiled beef based on visible near-infrared (Vis-NIR) hyperspectral imaging.

Different multivariate data analysis methods were investigated and compared to optimize rapid and non-destructive quantitative detection of beef adulteration with spoiled beef based on visible near-infrared hyperspectral imaging. Four multivariate statistical analysis methods including partial least squares regression (PLSR), support vector machine (SVM), least squares support vector machine (LS-SVM) and extreme learning machine (ELM) were carried out in developing full wavelength models. Good prediction was obtained by applying LS-SVM in the spectral range of 496-1000 nm with coefficients of determination (R2) of 0.94 and 0.94 as well as root-mean-squared errors (RMSEs) of 5.39% and 6.29% for calibration and prediction, respectively. To reduce the high dimensionality of hyperspectral data and to establish simplified models, a novel method named invasive weed optimization (IWO) was developed to select key wavelengths and it was compared with competitive adaptive reweighted sampling (CARS) and genetic algorithm (GA). Among the four multivariate analysis models based on important wavelengths determined by IWO, the LS-SVM simplified model performed best where R2 of 0.97 and 0.95 as well as RMSEs of 4.74% and 5.67% were attained for calibration and prediction, respectively. The optimum simplified model was applied to hyperspectral images in pixel-wise to visualize the distribution of spoiled beef adulterant in fresh minced beef. The current study demonstrated that it was feasible to use Vis-NIR hyperspectral imaging to detect homologous adulterant in beef.