摘要
目的:基于二维超声图像提出一种肝脏运动跟踪方法,以减小手术过程中由于呼吸运动带来的误差。方法:先选用中值流算法对二维超声图像感兴趣点进行跟踪,在跟踪过程中,如果中值流算法成功跟踪到目标,则在目标附近进行模板匹配,通过模板匹配来校正中值流算法跟踪过程中出现的漂移;如果中值流算法未能跟踪到目标,则需使用ORB(oriented fast and rotated brief)特征点进行匹配判断感兴趣点的位移,重新选定感兴趣点继续进行跟踪;如果连续跟踪失败,则说明中值流算法不适用该情况下的跟踪,则需使用KCF(kernelized correlation filters)跟踪算法代替中值流算法进行跟踪,并使用模板匹配校正长时目标跟踪误差。采用CLUST(Challenge on Liver Ultrasound Tracking)2015二维数据集进行测试和验证,并将算法校正前后的跟踪误差进行对比。结果:该方法对测试集血管等感兴趣点的平均跟踪误差为1.29 mm、平均跟踪帧率大于22帧/s,且校正后的算法跟踪误差明显减小,基本满足实时要求。结论:该方法跟踪精度较高、运行速度较快,可以满足肝肿瘤热消融手术中对肿瘤跟踪的临床需求。
Objective To propose a liver motion tracking method based on two-dimensional ultrasound images to reduce the error due to respiratory motion during surgery.Methods The points of interest in 2D ultrasound images were tracked with the median flow algorithm,and template matching was performed near the target if the target was caught successfully,and template matching was used to correct the drift in the tracking process of the median flow algorithm;if the target was not tracked,oriented fast and rotated brief(ORB)feature point matching had to be applied to determining the displacement of the point of interest.The median flow algorithm was not suitable for tracking in this case if continuous tracking failed,then it had to be replaced by kernelized correlation filters(KCF)tracking algorithm and the long-time target tracking error should be corrected using template matching.Testing and validation were carried out with CLUST(Challenge on Liver Ultrasound Tracking)20152D datasets,and the tracking errors before and after the algorithm correction were compared.Results The method had the average tracking error being 1.29 mm and the average tracking frame rate being higher than 22 frames/s,which decreased the tracking error of the corrected algorithm significantly and met real-time requirements generally.Conclusion The method has high tracking accuracy and operation speed,and satisfies the clinical demands of tumor tracking in thermal ablation procedures for liver tumors.
作者
庞杰
周著黄
吴水才
PANG Jie;ZHOU Zhu-huang;WU Shui-cai(Department of Biomedical Engineering,Faculty of Environment and Life of Beijing University of Technology,Beijing 100124,China)
出处
《医疗卫生装备》
CAS
2022年第12期8-14,共7页
Chinese Medical Equipment Journal
关键词
肝肿瘤
呼吸运动
肝脏运动
中值流算法
KCF算法
超声图像
liver tumor
respiratory motion
liver motion
median flow algorithm
kernelized correlation filters algorithm
ultrasound image
