Lai Ting-Yu, Chen Hsiao-I, Shih Cho-Chiang, Kuo Li-Chieh, Hsu Hsiu-Yun, Huang Chih-Chung
Department of Biomedical Engineering, National Cheng Kung University, Tainan 701, Taiwan.
Department of Occupational Therapy, National Cheng Kung University, Tainan 701, Taiwan.
Med Phys. 2016 Jan;43(1):148. doi: 10.1118/1.4937932.
Information about tendon displacement is important for allowing clinicians to not only quantify preoperative tendon injuries but also to identify any adhesive scaring between tendon and adjacent tissue. The Fisher-Tippett (FT) similarity measure has recently been shown to be more accurate than the Laplacian sum of absolute differences (SAD) and Gaussian sum of squared differences (SSD) similarity measures for tracking tendon displacement in ultrasound B-mode images. However, all of these similarity measures can easily be influenced by the quality of the ultrasound image, particularly its signal-to-noise ratio. Ultrasound images of injured hands are unfortunately often of poor quality due to the presence of adhesive scars. The present study investigated a novel Kalman-filter scheme for overcoming this problem.
Three state-of-the-art tracking methods (FT, SAD, and SSD) were used to track the displacements of phantom and cadaver tendons, while FT was used to track human tendons. These three tracking methods were combined individually with the proposed Kalman-filter (K1) scheme and another Kalman-filter scheme used in a previous study to optimize the displacement trajectories of the phantom and cadaver tendons. The motion of the human extensor digitorum communis tendon was measured in the present study using the FT-K1 scheme.
The experimental results indicated that SSD exhibited better accuracy in the phantom experiments, whereas FT exhibited better performance for tracking real tendon motion in the cadaver experiments. All three tracking methods were influenced by the signal-to-noise ratio of the images. On the other hand, the K1 scheme was able to optimize the tracking trajectory of displacement in all experiments, even from a location with a poor image quality. The human experimental data indicated that the normal tendons were displaced more than the injured tendons, and that the motion ability of the injured tendon was restored after appropriate rehabilitation sessions.
The obtained results show the potential for applying the proposed FT-K1 method in clinical applications for evaluating the tendon injury level after metacarpal fractures and assessing the recovery of an injured tendon during rehabilitation.
肌腱移位信息对于临床医生而言至关重要，这不仅有助于量化术前肌腱损伤情况，还能识别肌腱与相邻组织之间的任何粘连瘢痕。最近研究表明，在超声B模式图像中跟踪肌腱移位时，Fisher-Tippett（FT）相似性度量比拉普拉斯绝对差之和（SAD）以及高斯平方差之和（SSD）相似性度量更为准确。然而，所有这些相似性度量都很容易受到超声图像质量的影响，尤其是其信噪比。不幸的是，由于存在粘连瘢痕，受伤手部的超声图像质量往往较差。本研究探讨了一种新颖的卡尔曼滤波方案来克服这一问题。
使用三种先进的跟踪方法（FT、SAD和SSD）来跟踪模拟体和尸体肌腱的移位，同时使用FT来跟踪人体肌腱。这三种跟踪方法分别与所提出的卡尔曼滤波（K1）方案以及先前研究中使用的另一种卡尔曼滤波方案相结合，以优化模拟体和尸体肌腱的移位轨迹。在本研究中，使用FT-K1方案测量人体指总伸肌腱的运动。
实验结果表明，在模拟体实验中SSD表现出更好的准确性，而在尸体实验中FT在跟踪真实肌腱运动方面表现更佳。所有三种跟踪方法都受到图像信噪比的影响。另一方面，K1方案能够在所有实验中优化移位的跟踪轨迹，即使是在图像质量较差的位置。人体实验数据表明，正常肌腱的移位比受伤肌腱更多，并且经过适当的康复训练后，受伤肌腱的运动能力得到恢复。
所得结果表明，所提出的FT-K1方法在临床应用中具有潜力，可用于评估掌骨骨折后肌腱损伤程度以及评估康复过程中受伤肌腱的恢复情况。
