CONVOLUTIONAL NEURAL NETWORKS OF RECOVERING ACOUSTIC ABSORPTION IN ULTRASOUND TOMOGRAPHY FROM GRADIENT DATA
摘要
In this work, a method for reconstructing acoustic absorption from the measured pressure in sensors around the body under study is implemented. We develop a neural network architecture based on CNN that uses not only pressure data, but also the gradient of functional which is used to solve the inverse problem. We use the distribution of the gradient with respect to space variables as an input of the neural network and obtain the target image of the absorption as an output.