Prediction at harvest and detection at the ready‐to‐eat stage of internal black flesh severity in mangoes using portable Vis–NIR spectroscopy

BACKGROUND: Although previous studies have demonstrated that visible-near infrared (Vis-NIR) spectroscopy can accurately predict black flesh incidence at harvest and detect the disorder in ready-to-eat mango fruit, its ability to discriminate among severity levels has not yet been established. 'Tommy Atkins' mangoes were harvested and spectra (310-1100 nm) were acquired at harvest (n = 458) and after storage at 9 ± 0.5 °C, when the fruit reached ready-to-eat quality with flesh firmness of ≤15 N (n = 329). After reaching ready-to-eat ripeness, the fruit cheeks were cut longitudinally close to the seed and severity level categorized based on visual assessment of the percentage of the mesocarp area exhibiting visible black flesh symptoms. Fruit were classified as healthy (0%), low severity (1-30%), moderate severity (31-50%), or severe (51-100%). Spectral data (402-1100 nm) were preprocessed using standard normal variate and second derivative (Savitzky-Golay). Classification algorithms, including partial least squares discriminant analysis, simple logistic, sequential minimal optimization, multilayer perceptron, and random forest, were evaluated using accuracy, F-measure, and Kappa. RESULTS: Chlorophyll-related absorbance increased with severity, indicating that affected fruit were greener than non-affected fruit. The simple logistic algorithm showed the best performance on the test datasets, with accuracies of 77.4% for prediction at harvest and 81.8% for detection at the ready-to-eat stage. The models performed well on the extreme classes (healthy and severe), with classification accuracy >80%, whereas low and moderate severity levels were poorly discriminated. CONCLUSION: Vis-NIR spectroscopy reliably predicted, at harvest, and discriminated, at the ready-to-eat stage, symptom-free fruit from those developing severe black flesh, but showed limited accuracy for fruit developing mild or moderate symptoms during cold storage. © 2026 The Author(s). Journal of the Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.