Effects of multiple factors on photoacoustic detection of glucose based on back propagation neural network combined with intelligent optimization algorithms

In the work, a set of photoacoustic detection system of glucose based on different influence factors was established. In the system, a kind of Nd: YAG pumped 532nm OPO pulsed laser was used as the excitation source, and a focusing ultrasonic transducer with central frequency of 9.5MHz was used to capture the photoacoustic signals of glucose. In the experiments, the time-resolved photoacoustic signals and peak-to-peak values of glucoses with different concentrations and other three factors (temperature, excitation energy, and flow velocity) were obtained by using the orthogonal experiment method. To precisely predict the photoacoustic peak-to-peak values and glucose concentrations, the back propagation neural network (BPNN) was used to build a nonlinear models between the multiple factors and the photoacoustic peak-to-peak values and glucose concentrations. In BPNN, 108 train samples and 36 test samples were used. To further improve the prediction accuracy, BPNN combined with intelligent optimization algorithms, i.e., genetic algorithm (GA) and particle swarm optimization (PSO) algorithms were used. At the same time, the parameters adjusting for all algorithms were performed. Results show that the prediction root-mean-square error (RMSE) values of peak-to-peak values and glucose concentration were all improved, the RMSE of peak-to-peak values improve 19.9%, and the RMSE of glucose concentrations improve 34.62%. Therefore, it is found that the BPNN combined with intelligent optimization algorithms has an efficient effect on the prediction of photoacoustic detection for glucose concentration.