Neural network model and ultrasonic beam forming method based on neural network model

Abstract

The invention provides an ultrasonic beam forming method based on a neural network model. The method comprises: receiving, by the processing device, the ultrasonic echo radio frequency signal, and performing a preliminary processing of the ultrasonic echo radio frequency signal, so that the ultrasonic echo radio frequency signal is applicable to the neural network model; inputting the ultrasonic echo radio frequency signal into the neural network model; performing a convolution processing on the ultrasonic echo radio frequency signal for multiple times, and performing pooling processing after each convolution processing; flattening the ultrasonic echo radio frequency signal after the last pooling process is completed, so as to convert the ultrasonic echo radio frequency signal from three-dimensional data into one-dimensional data; classifying the one-dimensional data for multiple times; and performing grayscale mapping on the classified ultrasonic echo radio frequency signal to form an ultrasonic beam.