Study Of Improved Adaptive Threshold Segmentation Of Ultrasound Medical Images

Ultrasound Medical Image(UMI)segmentation has has always been a challenging task in the field of computer vision.At present,the segmentation methods in the literature mainly include traditional image segmentation methods and image segmentation methods based on deep learning.Because of the small sample size of some ultrasonic medical images,it is difficult to segment the images using the deep learning method which needs a large number of samples;In the traditional image segmentation methods,some methods need to manually adjust the parameters to effectively segment the targets in ultrasonic medical images.Therefore,using automatic parameter selection method to realize ultrasonic medical image segmentation has attracted many researchers’ attention.This paper studies the method of automatically extracting regions of interest from ultrasonic medical images,and improves an adaptive threshold segmentation method for ultrasonic medical images.The main works are as follows:(1)An automatic extraction method of regions of interest in ultrasonic medical images based on target area detection is studied.On the basis of converting ultrasonic medical images into binary images,this method detects the connectivity of the target region,estimates the area parameter values of multiple target regions,and iteratively determines the maximum connected region according to the area parameter values,so as to extract the target region of interest.The experimental results of several ultrasonic medical images further verify the feasibility of this method.(2)An improved adaptive threshold based ultrasonic medical image segmentation algorithm is presented.Firstly,the region of interest extracted from ultrasonic medical image is transformed into integral image;Then,the mesh search method is used to find the best adaptive threshold segmentation parameters to segment the extracted region of interest.The effect of target segmentation is evaluated by calculating the overlapped region between the result extracted by medical experts and the segmentation result of this method.Compared with the segmentation methods in the literature,the improved method has better segmentation results.