Scientific trends in spectroscopy and regression chemometric modelling for the estimation of whole fruit quality: A systematic review

Spectroscopic techniques, supported by chemometrics, provide rapid, non-destructive, and sustainable solutions for assessing the quality of whole fruits. This study systematically reviews and analyzes research on spectroscopy and regression chemometric modelling for the estimation of whole fruits from 1997 to 2025. A total of 389 English-language articles were retrieved from Scopus using a hybrid strategy that combined an initial search and snowballing. Geographical analysis identified China as the leading country in scientific output, followed by Spain, Italy, and the United States, while Africa and Oceania showed limited participation. Apples, grapes, pears, and mangoes were the most frequently studied fruits, and commonly modeled attributes included SSC, physicochemical properties, and bioactive compounds. NIR and Vis-NIR were the predominant techniques, complemented by emerging methods such as HSI and Raman. Among chemometric approaches, preprocessing relied mainly on hybrid strategies followed by SNV and derivatives. For dimensionality reduction, CARS, SPA, and hybrid methods were the most relevant. PLSR remained dominant for modeling, although there was an increasing use of advanced algorithms such as SVMR and deep neural networks. The review also examined current trends and future directions, highlighting progress in robust modeling algorithms, portable and online detection systems, and multimodal spectroscopy with data fusion. Key priorities include methodological harmonization, open data practices, and large-scale field validation. Overall, the findings highlight a transition toward more precise and adaptable systems while underscoring persistent challenges in standardization, real-world validation, and equitable access. This review provides a strategic foundation for advancing non-destructive technologies across the fruit value chain.