Location: Quality & Safety Assessment ResearchTitle: A novel hyperspectral microscope imaging technology for rapid evaluation of particle size distribution in matcha
|OUYANG, QIN - Jiangsu University|
|YANG, YONGCUN - Jiangsu University|
|KANG, RUI - Nanjing Agricultural University|
|WU, JIZHONG - Jiangsu University|
|CHEN, QUANSHENG - Jiangsu University|
|GUO, ZHIMING - Jiangsu University|
|LI, HUANHUAN - Jiangsu University|
Submitted to: Journal of Food Engineering
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/24/2019
Publication Date: 10/31/2019
Citation: Ouyang, Q., Yang, Y., Park, B., Kang, R., Wu, J., Chen, Q., Guo, Z., Li, H. 2019. A novel hyperspectral microscope imaging technology for rapid evaluation of particle size distribution in matcha. Journal of Food Engineering. https://doi.org/10.1016/j.jfoodeng.2019.109782.
Interpretive Summary: Matcha is finely ground powder of green tea leaves with special planting and processing. High grade matcha depends on the quality of the raw leaves as well as fine particle size of the powder. Furthermore, bioaccessibility and antioxidant activity of tea powder can be improved by properly reducing the particle size. Thus, particle size is a critical parameter for estimating quality of matcha. Traditional and classical analysis methods for particle size are sieve and laser diffraction. They are accurate, but not suitable for on-site, real-time analysis of multiple physical/chemical parameters. Hyperspectral microscope imaging (HMI) technology can characterize samples at a micro-scale, which make it possible for predicting the particle size of matcha powder with the merits of rapid, nondestructive, accurate, objective and high-throughput. D-values are the most commonly used parameters to describe particle size distributions. In this study, D-values scaled from 10 to 90 in increments of 10 were investigated with HMI technology to predict the particle size distributions of matcha. The models with data fusion, which consisted of textural and spectral features, performed well for predicting particle size of matcha. In addition, artificial neural networks (ANN) enhanced the performance of models with spectral features. The accuracies were above 80% for data fusion with ANN models to predict all different particle size of matcha samples.
Technical Abstract: Hyperspectral microscope imaging (HMI) technology, as a novel approach, was proposed to evaluate physical characteristics of matcha. Particle size distribution is one of the significant physical characteristics investigated. Textural features from HMI images at 524 nm with the highest signal and spectral features from the regions of interest were extracted as the input for the models. The models, containing textural or spectral features only, and data fusion models, which consisted of both textural and spectral features, were established in sequence for predicting the particle size distributions between D10 and D90. Results showed that the predictive abilities can be improved with data fusion, and in the order of data fusion first (84%) followed by textural features only (77%) and spectral features only (70%) with D80 in particular. This work demonstrated that HMI technology is a rapid, accurate and highly effective protocol for predicting the particle size distribution in matcha.