多神经网络在超声图像感兴趣区域检测中的应用

Application of multiple neural networks in the detection of ultrasonic image interest region

研究超声图像感兴趣区域准确检测问题,针对传统检测算法检测耗时、精度低等缺陷,提出了多神经网络的超声图像检测算法。首先采用离散余弦变换对超声图像感兴趣区域进行特征信息提取,然后分别利用两种神经网络对图像感兴趣区域进行检测,最后利用证据理论对检测结果进行融合,得到最终检测结果。在Matlab环境下进行仿真,仿真结果表明,多神经网络检测算法较传统方法的检测速度更高,定位更加准确。研究超声图像感兴趣区域准确检测问题,针对传统检测算法检测耗时、精度低等缺陷,提出了多神经网络的超声图像检测算法。首先采用离散余弦变换对超声图像感兴趣区域进行特征信息提取,然后分别利用两种神经网络对图像感兴趣区域进行检测,最后利用证据理论对检测结果进行融合,得到最终检测结果。在Matlab环境下进行仿真,仿真结果表明,多神经网络检测算法较传统方法的检测速度更高,定位更加准确。

Study on the accurate detection of the interest region in ultrasound image,aiming at the defects of traditional detection algorithm such as time consuming and low precision,an ultrasonic image detection algorithm based on multi neural network is proposed. Firstly,extract the feature information of ultrasonic image interest region by using discrete cosine transform. And then detect the interest region based on neural network. Finally fuse the detection results according to the DS theory and get the final result. The simulation under the Matlb environment results show that the new method is more accurate than the traditional method,and the positioning is more precise.