Quantitative analysis and early detection of postharvest soft rot in kiwifruit using E-nose and chemometrics

Kiwifruit is susceptible to postharvest decay caused by a variety of fungal pathogens. The fungal fruit rot disease continues to occur during storage, transportation, and sales, caused major economic losses to major producing nations. The electronic nose (E-nose) combined with chemometrics methods were applied for investigating the changes in volatile compounds of the kiwifruit experiencing soft rot during storage. Three pattern recognition methods, including Fisher function discriminant model, multilayer perceptron neural network (MLPNN), and radial basis function neural network (RBFNN), were applied to analyze the data obtained from the characteristic sensors, aiming at establishing the prediction model of the kiwifruit experiencing soft rot. The prediction accuracy of the model for identifying soft rot in kiwifruit was ranked as MLPNN > Fisher > RBFNN. The results showed that the model with the optimum discrimination accuracy was MLPNN, with prediction accuracies of 100% (healthy fruits) and 100% (end diseased fruits). MLPNN could better distinguish the early diseased kiwifruits. The accuracy rates of MLPNN model for the training set and the verification set is 90.91% and 87.50%, respectively. This study provides an effective method to detect and identify the healthy and diseased fruits.