TY - JOUR
T1 - Fast quantitative analysis and chemical visualization of amylopectin and amylose in sweet potatoes via merging 1D spectra and 2D image
AU - He, Hong Ju
AU - Liu, Hongjie
AU - Wang, Yuling
AU - Chew, Kit Wayne
AU - Ou, Xingqi
AU - Zhang, Mian
AU - Bi, Jicai
N1 - Publisher Copyright:
© 2024 Elsevier B.V.
PY - 2024/3
Y1 - 2024/3
N2 - The quantitative analysis and spatial chemical visualization of amylopectin and amylose in different varieties of sweet potatoes were studied by merging spectral and image information. Three-dimensional (3D) hyperspectral images carrying 1D spectra and 2D images of hundreds of the samples (amylopectin, n = 644; amylose, n = 665) in near-infrared (NIR) range of 950–1650 nm (426 wavelengths) were acquired. The NIR spectra were mined to correlate with the values of the two indexes using a linear algorithm, generating a best performance with correlation coefficients and root mean square error of prediction (rP and RMSEP) of 0.983 and 0.847 g/100 mg for amylopectin, and 0.975 and 0.500 g/100 mg for amylose, respectively. Then, 14 % of the wavelengths (60 for amylopectin, 61 for amylopectin) were selected to simplify the prediction with rP and RMSEP of 0.970 and 1.103 g/100 mg for amylopectin, and 0.952 and 0.684 g/100 mg for amylose, respectively, comparable to those of full-wavelength models. By transferring the simplified model to original images, the color chemical maps were created and the differences of the two indexes in spatial distribution were visualized. The integration of NIR spectra and 2D image could be used for the more comprehensive evaluation of amylopectin and amylose concentrations in sweet potatoes.
AB - The quantitative analysis and spatial chemical visualization of amylopectin and amylose in different varieties of sweet potatoes were studied by merging spectral and image information. Three-dimensional (3D) hyperspectral images carrying 1D spectra and 2D images of hundreds of the samples (amylopectin, n = 644; amylose, n = 665) in near-infrared (NIR) range of 950–1650 nm (426 wavelengths) were acquired. The NIR spectra were mined to correlate with the values of the two indexes using a linear algorithm, generating a best performance with correlation coefficients and root mean square error of prediction (rP and RMSEP) of 0.983 and 0.847 g/100 mg for amylopectin, and 0.975 and 0.500 g/100 mg for amylose, respectively. Then, 14 % of the wavelengths (60 for amylopectin, 61 for amylopectin) were selected to simplify the prediction with rP and RMSEP of 0.970 and 1.103 g/100 mg for amylopectin, and 0.952 and 0.684 g/100 mg for amylose, respectively, comparable to those of full-wavelength models. By transferring the simplified model to original images, the color chemical maps were created and the differences of the two indexes in spatial distribution were visualized. The integration of NIR spectra and 2D image could be used for the more comprehensive evaluation of amylopectin and amylose concentrations in sweet potatoes.
KW - Amylopectin
KW - Amylose
KW - Image
KW - Sweet potato
KW - Visualization
UR - http://www.scopus.com/inward/record.url?scp=85182739639&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85182739639&partnerID=8YFLogxK
U2 - 10.1016/j.ijbiomac.2024.129421
DO - 10.1016/j.ijbiomac.2024.129421
M3 - Article
C2 - 38228206
AN - SCOPUS:85182739639
SN - 0141-8130
VL - 260
JO - International Journal of Biological Macromolecules
JF - International Journal of Biological Macromolecules
M1 - 129421
ER -