Texture Feature Extraction from 1 H NMR Spectra for the Geographical Origin Traceability of Chinese Yam.
Zhongyi HuZhenzhen LuoYanli WangQiuju ZhouShuangyan LiuQiang WangPublished in: Foods (Basel, Switzerland) (2023)
Adulteration is widespread in the herbal and food industry and seriously restricts traditional Chinese medicine development. Accurate identification of geo-authentic herbs ensures drug safety and effectiveness. In this study, 1 H NMR combined intelligent "rotation-invariant uniform local binary pattern" identification was implemented for the geographical origin confirmation of geo-authentic Chinese yam (grown in Jiaozuo, Henan province) from Chinese yams grown in other locations. Our results showed that the texture feature of 1 H NMR image extracted with rotation-invariant uniform local binary pattern for identification is far superior compared to the original NMR data. Furthermore, data preprocessing is necessary. Moreover, the model combining a feature extraction algorithm and support vector machine (SVM) classifier demonstrated good robustness. This approach is advantageous, as it is accurate, rapid, simple, and inexpensive. It is also suitable for the geographical origin traceability of other geographical indication agricultural products.
Keyphrases
- deep learning
- high resolution
- magnetic resonance
- machine learning
- solid state
- big data
- electronic health record
- artificial intelligence
- randomized controlled trial
- contrast enhanced
- neural network
- risk assessment
- ionic liquid
- magnetic resonance imaging
- climate change
- south africa
- human health
- mass spectrometry
- heavy metals
- emergency department
- data analysis
- quantum dots