Hardware Design Of Electrical Impedance Imaging Systems And Electrical Impedance Reconstruction Algorithm Based On Two-Dimensional Convolutional Neural Network

Electrical impedance tomography(EIT)is an emerging non-invasive visual imaging technique that is more suitable for prolonged bedside monitoring than traditional medical imaging methods,and has the ability to detect diseased tissue and monitor physiological activity,with a wide range of applications.EIT applies a voltage or current signal through an electrode array,acquires the voltage signal from the corresponding electrode,and solves the inverse problem according to a reconstruction algorithm to obtain tissue-related information.In this paper,an electrical impedance imaging data acquisition system and relative image reconstruction algorithm are designed and the main tasks accomplished are:Designing the excitation current source for the electrical impedance imaging system,using a direct frequency synthesis chip to generate a sinusoidal voltage signal,combined with a voltage-controlled current source to obtain an excitation current source with adjustable frequency and high output impedance.The EIT data acquisition circuit was designed and built,including a high-pass filter circuit,a pre-two stage amplifier circuit,a fourth order Butterworth bandpass filter circuit,an analogue demodulation circuit and a low-pass filter circuit.A microcontroller-based data acquisition control program was designed,and a host computer software was programmed for serial data transmission and image reconstruction.To avoid the electrode contact impedance affecting the imaging,the adjacent excitation and adjacent measurement method was used in the driving mode of the current source and voltage measurement mode of the EIT system.A physical model of a cylindrical brine tank was constructed and the EIT hardware and software system was tested.The test results show that the system is capable of identifying and locating target objects.A convolutional neural network model was constructed and the EIT database was built using the finite element method to solve the problem of few training samples for machine learning.The network model was trained and tested using the position,shape and number of objects as a priori information.The algorithm is an absolute EIT image reconstruction algorithm without the need to collect background measurements in advance.The experimental results show that the neural network model in this paper has certain generalization ability and better anti-interference ability.