Non-contact Blood Species Identification Based On Visible And Near-infrared Spectroscopy

Non-contact identifying of blood species is of great importance in the field of medical biology research,national health technical supervision,import and export sample detection in customs and many other fields,and is significantly important for safeguarding national security and protecting people's lives.Current methods for the identification of blood species are usually destructive and open sampling,which may pollute and destroy the precious blood samples.Therefore,noncontact methods for blood species identification was urgently needed.In view of the fact that molecular absorption spectrometry can realize accurate,fast,stable and non-destructive measurement,the method was highly potential for the supervision of blood products.However,there were many problems in the non-contact measurement of blood samples,such as the detection of weak signal,the variantion of sample cell,and so on.The performance of spectroscopy methods can be easily affected by many other factors during measurement and analysis.And the influence on the discrimination performance of the varience of blood property would be uncertain.We fistly explored the feasibility of spectroscopy combined with chemometric methods for non-contact blood species identification,and theoretical and experimental work has been done in important parts,such as signal acquisition,data processing and modeling algorithm,to improve the performance for identification of blood origin.This paper also done some work in reducing the impact of interfering factors,including environmental factors and also blood itself.These works laid solid foundation for the application of spectroscopy method on noncontact blood species identification.The main research work involves:1.We firstly applied diffuse reflectance spectroscopy combined with PLSDA methods for identifying the origin(human or animal)of whole blood.PLSDA model successfully realized the discrimination of human and animal whole blood.We have firstly realized noncontact discriminating whole blood with human origin.2.From the perspective of fully obtaining and taking the advantage of blood information,spatially resolved transmission spectroscopy method was applied for improving prediction accuracy in noncontact blood species identification.The research demonstrated that spatially resolved diffuse transmission spectroscopy can improve the predictive performance for noncontact identifying blood species.We further studied the influence on the prediction effect of the calibration model built with the spectra from single point to all ten detection points.3.From the perspective of signal processing and model establishing,theoretical and experimental work has been done in some important links,such as wavelength selection,calibration samples selection,and modeling algorithm improvement,to improve the performance for identification of blood origin.We applied Kennard-Stone combined with LSSVM method to improve the predictive performance of calibration model for blood species identification.4.To improve the robustness of spectroscopy method for identifying blood species,this work studied the influence of storage time of whole blood in vitro,the clotted blood and hemolytic blood on the spectra and on the discriminating blood species.We tried calibration model optimization method,making the variation range of the analyzed element in the calibration set cover that in the prediction set,to improve the performance of the calibration model.The results demonstrated the effectiveness of the optimization method.