Application of neural networks to estimate carotenoid content during ripening in tomato fruits (Solanum lycopersicum)

- We proposed ANN models as an alternative to estimate carotenoid content.
- ANN models showed a high correlation between dependent and independent variables.
- LAI showed a high impact on estimations of carotenoid content.
- Including LAI in ANN models provided reliable and accurate carotenoid estimations.

Commonly carotenoid determinations in tomato are performed in full ripening tomatoes. In this work six tomato ripening stages were established. The relationship among color parameters (L*, a*, b*, and hue), maturity stages, and leaf area with the lycopene and β-carotene concentration was analyzed with different regression models. The R2 values were low, showing that lycopene and β-carotene content was not well correlated with color during the ripening stages. The objective of this work was to provide an ANN model including leaf area index (LAI) and color readings as inputs to solve this lack of fit of the regression models for carotenoid estimations in tomatoes. Two multilayer perceptrons (MLPs) were trained and validated, with six input variables and one output variable, to estimate the concentration of both carotenoids in tomato samples at different ripening stages. Comparing the results of the MLPs with those obtained by regression models it was concluded that when the MLPs are used within the range studied, they are able to estimate lycopene and β-carotene concentrations of tomato with accuracy and reliability, solving the lack of fit by regression models.