Study On Percutaneous Bilirubin Detection Tochnology Based On Spectral Analysis

Bilirubin is an important metabolite in the human body.Clinically,it is one of the important indicators for liver function testing.Neonatal jaundice is also known as neonatal hyperbilirubinemia and it may seriously harm the health of newborns.The incidence of neonatal jaundice is higher than 50%.Therefore,it’s necessary to closely monitor the neonatal bilirubin level.At present,the accurate detection of bilirubin relies on the blood sample testing.This method can cause pain,trauma and a risk of infection,so it is not suitable for continuous monitoring.There is an urgent need for a percutaneous bilirubin detection technology that can accurately detection bilirubin levels.Most current transcutaneous bilirubinometers are based on the reflection spectrum of skin.But this method is affected by other interfering substances such as melanin in the skin and its performance of detection and screening are limited.In view of the above situation,this paper proposes a bilirubin detection technology based on skin fluorescence spectrum analysis.The fluorescence of the molecules is the emit light after the molecules absorbing light of a specific wavelength and it contains the inherent properties of the substance.Bilirubin has a significant absorption peak on light around 460 nm and generates fluorescence in the range of 500-600 nm.Based on this feature,this paper proposes a method for detecting bilirubin levels using skin fluorescence spectrum and sets up a bilirubin fluorescence detection system.The system uses a light source with a wavelength of460 nm as the excitation light to illuminate the skin surface.The skin fluorescence is collected by a fiber probe and detected by a spectrometer for modeling analysis.In order to analyze the correlation between the fluorescence spectrum and bilirubin concentration,we established a data set of the fluorescence spectra of the newborns and their serum bilirubin concentration.In the spectral analysis,the principal component analysis method is used for dimension reduction and feature extraction,and the mixed model which contains support vector regression model and artificial neural network model are used to predict the bilirubin concentration.The model obtained good performance in 71 test data.The correlation coefficient between the predicted value of bilirubin concentration and the actual serum bilirubin concentration achieves 0.92,and the average absolute error reaches31.76 μmol/L,which is comparable to the performance of current transcutaneous bilirubinometers.Besides,the proposed method achieves a sensitivity of 1.00 at a specificity of 0.904 when it is used for screening total serum bilirubin level above 205μmol/L,which is better than the performance of current transcutaneous bilirubinometers.Concerning that some babies with jaundice have symptoms of high direct bilirubin levels,support vector machine classification model is used in this paper to classify the fluorescence spectra of samples and all high direct bilirubin samples in test data can be screened.These show that the methods proposed in this paper have great potential in use of neonatal jaundice detecting.In non-invasive detection of bilirubin based on reflection spectrum,the overlapping of spectra of different skin depths causes serious interference.To solve this problem,spectrum-depth analysis method based on optical coherence tomography is proposed in this paper.This method can obtain the scattering spectra of samples at different depths,avoiding the overlapping and interference of spectral at different depths of the sample.