Research And Design Of Exhaled Biomarker Detection Method For Patients With Acute Kidney Injury

As an important metabolic organ in the body,the kidney plays a role in maintaining acid-base balance in the body.However,with the increase in the variety and frequency of drugs used in medical practice,the risk of kidney damage has increased,leading to a significant increase in the incidence of acute kidney injury.The metabolic markers in human exhaled breath can reflect the health status to a certain extent.According to relevant studies,breath testing can be used as a non-invasive,convenient and low-cost test for the initial screening and early diagnosis of acute kidney injury diseases.In this paper,we study and design a breath detection device based on FPGA technology for ammonia and acetone gas in the breath markers of patients with acute kidney injury,and realize the detection of patients with acute kidney injury by constructing a metal oxide semiconductor sensor array with pattern recognition algorithm.The main work of this paper is divided into three parts as follows:1)The selection of sensor arrays and the pre-processing methods of experimental data as well as the sensor fault identification algorithms are investigated.First,the static gas distribution system required for the experiments was built according to the experimental objectives,the stability,sensitivity and correlation of different types of Metal-OxideSemiconductor(MOS)sensors were tested,the sensor characteristics were further analyzed,and based on the results of the comparative analysis,the optimal four sensors were selected from the alternative sensors to form a 2 *2 sensor array.Next,the acquired sensor data were pre-processed using the pre-processing method of data.Finally,during the testing of the sensor array some of the sensors showed faults such as aging of the heating end,aging of the sensitive end and poor contact,which caused large fluctuations in the acquired data.In order to reduce the influence of the abnormal sensor state on the pattern recognition results,this paper also investigates the sensor fault identification method,based on the traditional Support Vector Machine(SVM)The results show that the SVM model optimized by using the Sparrow search optimization algorithm(SSA)can not only effectively identify common faults in the data acquisition process It also reduces data redundancy and improves the reliability of the detection system at the same time.2)A collection system for breath markers was designed.In order to realize the detection of ammonia and acetone,the breath markers in patients with acute kidney injury,a breath marker detection device was designed based on FPGA in this paper.Firstly,the FPGA development environment and chip selection were compared and studied,and finally Altera’s field-programmable gate array EP4CE6F17C8 was selected,and the design and synthesis of the field-programmable gate array was carried out under Quartus 18.1 software.Secondly,the design of the breath marker acquisition system was carried out,and the hardware system was decomposed into data cache module,power supply module,interface circuit module,USB to serial module,AD conversion module,communication module,display module,and clock module,etc.The purpose of the modular design approach is to improve the development efficiency and reduce the difficulty of subsequent hardware upgrades.Finally,the circuit schematic of the whole system is drawn using Altium Designer 21 software.3)Qualitative and quantitative algorithms for breath markers were investigated.Due to the cross-sensitivity of MOS sensors,it is necessary to select an appropriate pattern recognition algorithm for the analysis of data collected by the breath detection device.During the qualitative identification of breath markers in patients with acute kidney injury,the BP neural network model(BPNN)optimized by using the Bat Algorithm(BA)achieved an accuracy of 98.24% and 98.17%,an increase of 4.61% and 4.68%,respectively,compared with the unimproved neural network.In the quantitative analysis,the BP neural network model optimized by Particle Swarm Optimization(PSO)was chosen to predict the exhaled breath markers ammonia and acetone in patients with acute kidney injury,and the experimental results showed that the mean absolute error(MAE)of the two gases were In summary,through the optimization of the intelligent population algorithm,the prediction effect of the BP neural network model has been greatly improved with higher prediction accuracy and generalization ability.