Scholarlink

<p>The accurate quantification of lipids, proteins, and carbohydrates is essential for evaluating the nutritional quality of food materials. In this study, transmission terahertz time-domain spectroscopy imaging (THz-TDS) was evaluated for predicting the chemical composition of structurally heterogeneous foods, using sunflower kernels and biscuit tablets as representative samples. Partial least squares regression (PLSR), support vector regression (SVR), and random forest regression (RF) models were developed for THz-TDS data under a consistent validation protocol. The results showed strong prediction performance for sunflower kernels, where SVR achieved high prediction accuracy for carbohydrates, protein, and lipid with <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\:{R}_{cv}^{2}\)</EquationSource> </InlineEquation> of 0.963, 0.943, and 0.909, respectively. For biscuit, SVR and RF yielded comparable results with <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\:{R}_{cv}^{2}\)</EquationSource> </InlineEquation> values ranging from 0.705 to 0.780 across the three constituents. These findings highlight the potential of THz-TDS imaging for nutrient profiling in. dense food matrices and indicate that porous and strongly scattering matrices remain challenging. under the present transmission configuration due to sample structural characteristics.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Terahertz time-domain spectroscopy imaging for predicting lipid, protein, and carbohydrate content of structurally heterogeneous food materials

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月22日

查看全文
  • Qingxia Li,
  • Yuqiao Ren,
  • Wenrui Dong,
  • Brijesh K. Tiwari,
  • Da-Wen Sun

摘要

The accurate quantification of lipids, proteins, and carbohydrates is essential for evaluating the nutritional quality of food materials. In this study, transmission terahertz time-domain spectroscopy imaging (THz-TDS) was evaluated for predicting the chemical composition of structurally heterogeneous foods, using sunflower kernels and biscuit tablets as representative samples. Partial least squares regression (PLSR), support vector regression (SVR), and random forest regression (RF) models were developed for THz-TDS data under a consistent validation protocol. The results showed strong prediction performance for sunflower kernels, where SVR achieved high prediction accuracy for carbohydrates, protein, and lipid with \(\:{R}_{cv}^{2}\) of 0.963, 0.943, and 0.909, respectively. For biscuit, SVR and RF yielded comparable results with \(\:{R}_{cv}^{2}\) values ranging from 0.705 to 0.780 across the three constituents. These findings highlight the potential of THz-TDS imaging for nutrient profiling in. dense food matrices and indicate that porous and strongly scattering matrices remain challenging. under the present transmission configuration due to sample structural characteristics.