基于高斯过程回归高强度聚焦超声仪器采集图像的层间测距算法

Interlayer fault location algorithm based on Gauss process regression for high intensity focused ultrasound image acquisition

提出基于高斯过程回归的层间测距算法,通过对已知样本的训练,求解模型的参数,得到回归器,用于层间测距。通过对各实体组织进行超声扫描,得到相应的序列图像。运用到由高强度聚焦超声(high intensity focused ultrasound,HIFU)仪器采集的3组超声图像样本集中进行实验,验证方法的有效性和稳定性。测量绝对误差均值相对较小,在1～3 mm的范围内。绝对误差均值都低于0. 5 mm,而且相对误差的均值保持在10%以内,标准差比较稳定,浮动不大。而在4 mm和5 mm处测量误差相对较大,绝对误差均值在1mm左右,相对误差也会出现变大的趋势。但总体的相对误差在25%内。本研究提出的基于高斯过程回归的方法在1～3 mm的范围内,具有良好的测距结果,当距离超过3 mm时,测距误差相对较大。

To propose an algorithm for inter-layer ranging based on Gauss process regression.By training known samples,the parameters of the model were solved,and the regression was obtained for inter-layer ranging.Through the ultrasonic scanning of the solid tissues,the corresponding sequence images were obtained.Three groups of ultrasonic image samples collected by HIFU instrument were used to verify the effectiveness and robustness of the method.The mean absolute error measured was relatively small,ranging from 1 mm to 3 mm.The mean absolute error was less than 0.5 mm,and the mean relative error was less than 10%.The standard deviation was stable and fluctuates little.The measurement error was relatively large at 4 mm and 5 mm,and the mean absolute error was about 1 mm.The relative error would become larger.But the overall relative error was within 25%.The proposed method based on Gauss process regression has good ranging results in the range of 1 mm to 3 mm.When the distance exceeds 3 mm,the ranging error is relatively large.