Three-dimensional ultrasonic image segmentation method based on machine learning

Abstract

The invention discloses a three-dimensional ultrasonic image segmentation method based on machine learning. The method comprises following steps of sampling different regions in multiple groups of marked two-dimensional images, taking an image segment which takes a certain pixel point as a center, and the category represented by the pixel point as a group of samples, establishing many data samplesthereby and dividing the samples into a training set and a test set; constructing a convolution neural network as a classifier, and training the convolution neural network by taking the image segmentas input and the category of the pixel point as a tag; using the trained classifier to process data of the test set and calculating accuracy, estimating the performance of the classifier; adjusting parameters of the convolution neural network, repeating training and estimating processes for many times, and finally, taking an optimal result as the parameter of the classifier; and using the trainedclassifier to process the whole picture, and dividing pixel points in the image into different categories, thereby achieving segmentation of the image.

Description

technical field [0001] The invention belongs to the field of ultrasonic image processing, in particular to an optimization method for three-dimensional ultrasonic image segmentation. Background technique [0002] Ultrasound image segmentation is an important field in ultrasound image processing, that is, to segment regions of interest in ultrasound images. In the traditional ultrasonic image processing, most of them are based on some traditional image processing techniques. The information in the ultrasonic image is complex, the gray level distribution is uneven, the noise is large, and the organs and tissues are prone to deformation. These factors increase the quality of feature selection and image quality. The difficulty of segmentation makes it almost impossible to achieve segmentation for all ultrasound images using a general method. Deep learning has multiple hidden layers and the ability of autonomous learning, and is widely used in the field of image processing. As ...

AI Technical Summary

Problems solved by technology

[0003] Purpose of the invention: The technical problem to be solved by the present invention is mainly aimed at the problems that the existing ultrasonic image segmentation is mainly realized by traditional image processing technology, requires manual intervention, has low efficiency and poor versatility, and provides a machine learning-based The method improves the segmentation efficiency and ensures the generality of the method

Images