Research On Key Technologies For Photoacoustic Glucose Concentration Measurement Based On Optical Power Balance Shift

Diabetes is one of the major life-threatening diseases,and people with diabetes need to continuously monitor their blood glucose to better cope with the risk of diabetes.Traditional blood sugar detection methods are usually invasive,and the detection process is not only painful but also accompanied by the risk of infection.Photoacoustic technology has been proved to be a potential non-invasive blood glucose detection scheme with great application potential.At present,some teams have completed the measurement of high-concentration glucose concentration based on photoacoustic technology,and some teams have achieved low-concentration glucose detection using multi-wavelength pulsed lasers.In thesis,a dual-wavelength differential photoacoustic experiment system was built based on the Optical Power Balance Shift to realize the concentration detection of low concentration glucose solution.Firstly,based on the 1560 nm single-wavelength photoacoustic experimental system,it has been confirmed that there is a significant linear relationship between glucose concentration and the amplitude and phase of photoacoustic signals and three factors that affect photoacoustic signals have been analyzed:temperature,water level,and water change fluctuations,with temperature having the greatest impact on the experimental system.In order to reduce the impact of temperature on experimental accuracy,a dual-wavelength differential photoacoustic experiment system consisting of a 1392 nm laser and a 1560 nm laser was selected in thesis.The experimental results showed that the influence of temperature on the photoacoustic signal was reduced by half,but the influence still existed.In this regard,based on the Optical Power Balance Shift,the correlation between photoacoustic signals and glucose concentration reached 0.94 within the range of human blood glucose concentration(0-400mg/d L).In order to further improve the prediction accuracy,thesis used traditional machine learning algorithms to collect 224 sets of experimental data as a dataset for model training.The final prediction result showed that the proportion of A+B regions reached over 82%.Finally,a small-scale human earlobe optoacoustic signal acquisition system based on self-designed earlobe fixtures was established in thesis,and an OGTT experiment was conducted on one subject The results show that based on the Optical Power Balance Shift,the photoacoustic signal has the same trend as the changes in human blood sugar levels,with a correlation of 0.82.However,the photoacoustic signal at the human earlobe is easily affected by interference factors,and the first set of data needs to be calibrated after wearing the detector.The calibrated human glucose predictions resulting in an RMSE,MAD,and MARD of 35.45 mg/dl,29.16 mg/dl,and 26.2%,respectively,with 54% of the estimates being within Zone A of the CEG,100% of the estimates being within Zone A+B of the CEG.