Selection of industrial tomatoes using TD-NMR data and computational classification methods

Tomato processing chain has a world economic relevance for the food industry and the agribusiness, providing ready-to-eat products and raw material for other production chains. The product quality is depending on control of some fruit attributes, such as color, soluble solids content (SSC), and defects. The aim of this study was to develop accurate and nondestructive classification models according to the tomato maturation stage, SSC, and presence of defects using Time-Domain Nuclear Magnetic Resonance (TD-NMR) associated with computational classification methods. Each class showed different decay times. Green tomatoes showed a shorter decay signal than red tomatoes, mainly due to the relaxation signal being related to the water mobility in different vegetable tissue compartments. Classification models resulted in great accuracy performances, the best accuracy for each classification were: maturity index: 97% (SVM); SSC: 100% (SVM and kNN); presence of defects: 90% (PLS-DA). These results show that CPMG decays associated with computational methods can be used in the tomato processing industry to classify tomato samples. These classification models showed the potential of TD-NMR technique in a high-throughput screening application before the processing.

Automated summary

This study successfully developed accurate and nondestructive classification models based on tomato maturation stage, SSC, and presence of defects using Time-Domain Nuclear Magnetic Resonance (TD-NMR) associated with computational classification methods. The results demonstrate the potential of TD-NMR technique in high-throughput screening applications in the tomato processing industry.