TY - JOUR
T1 - Spectral markers and machine learning
T2 - Revolutionizing Rice evaluation with near infrared spectroscopy
AU - Sampaio, Pedro Sousa
AU - Carbas, Bruna
AU - Soares, Andreia
AU - Sousa, Inês
AU - Brites, Carla
N1 - Funding Information:
Funding for this research has been received from TRACE-RICE\u2014Tracing rice and valorizing side streams along with Mediterranean blockchain, grant no. 1934 (call 2019, Section 1 Agrofood) of the PRIMA Program supported under Horizon 2020, the European Union's Framework Program for Research and Innovation, and Research Unit, GREEN-IT Bioresources for Sustainability Base Funding https://doi.org/10.54499/UIDB/04551/2020 .
Publisher Copyright:
© 2025 The Authors
PY - 2025/11/15
Y1 - 2025/11/15
N2 - The evaluation of rice varieties is a complex, time-consuming process requiring advanced equipment. This study aimed to discriminate 22 commercial rice varieties from six types by analyzing biochemical, physicochemical, and cooking properties. Near-infrared (NIR) spectroscopy, combined with machine learning, linked molecular properties with quality traits, offering a high-throughput solution. Partial Least Squares (PLS) models accurately predicted parameters such as whiteness (R2 = 0.94), width (R2 = 0.94), resilience (R2 = 0.96), and springiness (R2 = 0.98), highlighting key wavelength regions. Principal Component Analysis (PCA) revealed distinct clustering patterns, while Partial Least Squares Discriminant Analysis (PLS-DA) achieved a 17 % error rate in external predictions. Spectral markers at A6032/4457 cm−1, A7004/5241 cm−1, and A7004/4749 cm−1 reflected biomolecular differences among varieties. This innovative approach enables precise quantification, classification, and differentiation of rice types, enhancing quality control, improving consumer satisfaction, and optimizing breeding selection processes efficiently.
AB - The evaluation of rice varieties is a complex, time-consuming process requiring advanced equipment. This study aimed to discriminate 22 commercial rice varieties from six types by analyzing biochemical, physicochemical, and cooking properties. Near-infrared (NIR) spectroscopy, combined with machine learning, linked molecular properties with quality traits, offering a high-throughput solution. Partial Least Squares (PLS) models accurately predicted parameters such as whiteness (R2 = 0.94), width (R2 = 0.94), resilience (R2 = 0.96), and springiness (R2 = 0.98), highlighting key wavelength regions. Principal Component Analysis (PCA) revealed distinct clustering patterns, while Partial Least Squares Discriminant Analysis (PLS-DA) achieved a 17 % error rate in external predictions. Spectral markers at A6032/4457 cm−1, A7004/5241 cm−1, and A7004/4749 cm−1 reflected biomolecular differences among varieties. This innovative approach enables precise quantification, classification, and differentiation of rice types, enhancing quality control, improving consumer satisfaction, and optimizing breeding selection processes efficiently.
KW - Classification models
KW - Machine learning techniques
KW - NIR spectroscopy
KW - PCA
KW - PLS-DA
KW - Rice
KW - Spectral markers
UR - https://www.scopus.com/pages/publications/105010891204
U2 - 10.1016/j.foodchem.2025.145569
DO - 10.1016/j.foodchem.2025.145569
M3 - Article
C2 - 40684570
AN - SCOPUS:105010891204
SN - 0308-8146
VL - 492
JO - Food Chemistry
JF - Food Chemistry
M1 - 145569
ER -