中文摘要：

超声成像是一种无创、实时和便捷的成像方法,已被广泛用于人体肌肉运动分析。临床上目前主要采用手工处理的方法来提取超声图像中的形态结构,不仅依赖于操作者的主观经验,且耗时严重、重复性低。提出一种自动计算与跟踪超声图像序列中肌束方向的新方法,利用基于约束互信息的自由变换算法自动提取感兴趣区域,然后基于局部Radon变换对超声图像进行投影,并利用相邻帧间肌束的连贯性,采用卡尔曼滤波的图像处理方法。对6名试验对象进行步行实验,共采集小腿内侧腓肠肌的动态超声影像1 080帧,并用所提出的方法计算肌束方向。实验结果表明,所提出的自动检测方法有较好的鲁棒性。与人工检测结果相比,两者的平均相关系数为0.92±0.02,平均误差为0.30°±0.62°。该方法可以一定程度地替代人工测量,在处理大量超声数据时具有较大的优势。

英文摘要：

Ultrasound imaging is a noninvasive,real-time and convenient imaging method,which has been widely applied in analysis of human muscle contraction. So far manual operation is commonly used in clinical applications to extract morphological parameters from ultrasound image,however,may bring outcomes such as subjective,time-consuming and low repetitive. In this paper,an automatic method is proposed to calculate and track the fascicle orientation in sequences of ultrasound images. We constrained mutual information-based freeform deformation algorithm to automatically detect reading of interest. Then the detected results were projected based on local Radon transform. Then the Kalman filter image processing technique was applied to predict and track the next frame change of fascicle orientation. Six subjects participated in the walking experiment. We collected the dynamic ultrasound images from the medial gastrocnemius muscle during the process and calculated the muscle fascicle orientation using the proposed method. Experimental results showed that the proposed automatic detection method achieved better robustness. Compared with the manual measurement,the average correlation coefficient was 0. 92 ± 0. 02,and the average error was 0. 30° ± 0. 62°. To some extent,theproposed method can replace the manual measurement,presenting a significant advantage especially when dealing with a large number of ultrasound data.