부산대학교 도서관

Many diseases can be diagnosed by blood sample tests using automated analyzers. Novel imaging techniques, like Hyperspectral (HS) Imaging (HSI), are arising as potential tools for this purpose, offering quick and cost-effective approaches. In this paper, a methodology to use HS microscopic data of blood plasma samples to identify potential spectral biomarkers (SBs) and then quantify them in HS macroscopic data is presented. This quantification was performed using spectral unmixing algorithms. Finally, the abundance of each potential SB was correlated with each one of the blood test variable levels with and without removing outliers. Results found several statistically significant correlations (p - value < 0.05) and a highly significant correlation (p - value < 0.01) with the creatinine variable. These preliminary results show the potential of HSI sensors for performing quick and cost-effective blood sample tests.