Intelligent sensory technologies, NIR spectroscopy and chemometrics combined with machine learning based on multi-source data fusion for comprehensive evaluation of Sinapis Semen in different processing degrees.

Authors

Weiting Liang, Rongxiao Zhong, Zhiguo Ma, Menghua Wu, Ying Zhang, Wei Zhang, Wenting Zhong, Yangfei Ding, Xinyuan Zhang, Hui Cao

Published in

Journal of pharmaceutical and biomedical analysis. Volume 268. Pages 117212. Oct 22, 2025. Epub Oct 22, 2025.

Abstract

Sinapis Semen, as a traditional Chinese medicine, has an unclear relationship between its stir-frying degrees and sensory characteristics. Therefore, it is essential to develop a multi-index evaluation method to classify the processing degree of Sinapis Semen. Based on diverse intelligent sensory technologies and chemical analysis, the features of "color-aroma-taste-quality" of raw and stir-frying Sinapis Semen were systematically collected and objectively characterized, establishing discriminative models by integrating with machine learning. The results indicated that as the stir-frying increased, the overall color brightness diminished, the volatile constituents of sulfides and aromatic compounds exhibited a significant increase, and the taste discrepancies were primarily concentrated in saltiness, astringency, and sourness, which were related to the alkaloids and polyphenols contained in Sinapis Semen. Three machine learning models were employed to evaluate and compare their performance. TabTransformer achieved an accuracy of 96.92 % in single-source modeling using the NIRS data; on the fused dataset, TabTransformer and MLP attained accuracies of 100 % and 98.33 %, respectively, demonstrating the effective integration in handling multidimensional information from diverse data sources. This research successfully developed discrimination models for Sinapis Semen at varying processing degrees, providing a valuable reference for its standardized production, and offering a novel approach for process optimization of others.

PMID:
41145066
Bibliographic data and abstract were imported from PubMed on 28 Oct 2025.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.