Back to results list
Please use this identifier to cite or link to this item:
|Title:||Human motion analysis with sonomyography||Authors:||Zhou, Guangquan||Advisors:||Zheng, Yong-ping (BME)||Keywords:||Muscle contraction -- Regulation||Issue Date:||2015||Publisher:||The Hong Kong Polytechnic University||Abstract:||Skeletal muscles are essential for regulating force generation and controlling motions of human body as biological motors, and their pathological conditions can lead to abnormities in human motion. However, muscle activation is not involved in the analysis of conventional kinematics and kinetics.The currently available motion analysis systems mainly rely on surface electromyography (SEMG) to identify muscles with abnormalities,but it is challenging to differentiate activities of neighboring muscles using SEMG,particularly the deep muscles in the body. Moreover, it was reported that modulation of the interaction between the muscle and tendon tissue takes place in response to the effective utilization of the tendon elasticity in mechanical demands during walking. SEMG also cannot provide the information about the interaction between muscle and tendon during motion.Muscle imaging is a promising field of research to understand the biological and bioelectrical characteristics of muscles through the observation of muscle architectural change. Sonomyography (SMG) is a technique which can quantify the real-time architectural change of muscles under different contractions and motions with ultrasound imaging. SMG can provide complementary information to EMG for examining the static and dynamic muscle change and inherently differentiate activities of neighboring muscles. Many studies have been reported to use SMG for muscle assessment and rehabilitation outcome measurement in vivo. Among the reported muscle architecture parameters (MAPs) that can be identified in ultrasound images, the fascicle length (FL) and pennation angle (PA) are the most often used SMG parameters to quantify changes in fascicle geometry because the ability of muscle to generate force is mainly determined by its FL and PA. However, the FL and PA are mostly extracted from ultrasound images by manual operation, which is time-consuming and subjective, and impedes extensive application of SMG for dynamic muscle functional analysis. Moreover, although the orientation and placement of ultrasound probe can influence the imaging of muscle structure, the effects of probe orientation and position on continuous SMG measurements have not been well examined in previous studies.In this study, a human motion analysis system with ultrasound imaging and automatic SMG extraction was advanced, including an ultrasound scanner with custom-designed probe and software, a vision-based motion capture system, and a custom-designed synchronization device. The automatic SMG extraction was achieved by applying a number of newly developed image processing techniques, including steps of ultrasound image enhancement using Gabor wavelet filter, muscle fascicle orientation detection using variance Hough transform (VHT), and a novel feature tracking method with the orientation-sensitive segmentation algorithm (FT-OSS). It is beneficial to enhance fascicle structures before the detection of fascicles in the musculoskeletal ultrasound images. However, the previous used enhancement methods were mainly dependent on the gradient or Hessian matrix, which is sensitive to the speckle noise in ultrasound images. Motivated by the fact that Gabor wavelet is close to human visual system, Gabor wavelet enhancement filter and Total Variation with Gabor Wavelet (GWVTV) were designed to enhance the ultrasound image as a preprocessing step through detecting and weighting fascicle structures with a set of Gabor filters. With the consideration of the typical characteristics of fascicles in ultrasound images, an automatic fascicle measurement method was developed through applying VHT on the image weighted with the Gabor wavelet enhancement filter. Moreover, based on the assumption of the homogeneous affine transformations in the muscle region, a new feature tracking method, FT-OSS, was also proposed to quantitatively identify the change of fascicles using motion estimation with fascicle structures. The method started with the manual marking points on the fascicular path on the first frame of image. Then the cohesiveness fascicle orientation was used as a feature for segmentation. LucasKanade optical flow algorithm was applied on the segmented image to estimate the global affine transform parameter for tracking the fascicle. After the fascicles were identified with the above methods, FL and PA were derived with the position of the superficial and deep aponeuroses that were recognized using the normalized Radon transform with the consideration of the distribution of aponeuroses in ultrasound images.
The temporal offset between the ultrasound scanner and the vision-based motion capture system was firstly identified for improving the performance of the newly developed system. Preliminary results also demonstrated that the relative change of probe orientation was not the determinant factor in the developed system after the representative ultrasound plane was identified for the measurements of FL and PA. Moreover, the results on simulated images generated with Matlab demonstrated that the designed image enhancement methods could preserve edges well (FOM > 0.8). The visual examinations on results of real ultrasound images also suggested that the line features enhanced with Gabor wavelet were consistent with the characteristics of human visual system. The proposed enhancement methods could therefore be applied as a preprocessing step for fascicle estimation when the quality of ultrasound images was poor. Furthermore, tests on synthetic images with simulation program Field II showed that the VHT method could identify the fascicle orientation with an average error less than 0.1°. The performances of both VHT and FT-OSS were tested on ultrasound images of Gastrocnemius medialis muscle (GM) obtained from four normal subjects (32 ± 4.0 years). A good agreement was found between the measurements obtained by the manual and automatic methods (all r > 0.9; p < 0.001).The values of FL and PA were also consistent with the previous reports. These results supported that both the methods could accurately identify FL and PA of muscles.In summary, the developed system with newly developed image processing techniques can provide us an effective way for the assessment of muscle function in research and clinical practices.The behaviors and performance of the muscles are different between males and females, which can be ascribed to the geometric and architectural differences in muscles. Previous studies have reported differences in FL and PA between genders. However, there is a lack of report about the gender differences in muscle and tendon when performing dynamic task. Cyclic plantar flexion experiments were therefore conducted in this study to examine the gender dependence of the muscle and tendon.Seven male and seven female subjects (mean age of 31.9 ± 6.1 years) without history of musculoskeletal injury were recruited from the Hong Kong Polytechnic University for the study. The subject was instructed to stand straight at the beginning of the examination and then repeatedly conduct the two-legged plantar-flexion exercises while keeping the knee angle unchanged. All 3-dimensional (3D) spatial information of human body was obtained from the vision-based motion capture system. A portable ultrasound instrument was also implemented to obtain the ultrasound images of GM with a linear probe simultaneously. The probe was secured steadfastly on the mid-belly of GM muscle with bandage, and carefully adjusted to visualize the fascicles from superficial to deep aponeuroses. The FL and PA of GM were automatically extracted from the ultrasound images. The tendon length was determined as the difference between FLcos(PA) and muscle-tendon unit (MTU) length calculated from the average angle of joint. Significant correlations between the ankle angle and PA (r = 0.92 ± 0.06; all p < 0.001) and FL (r = -0.93 ± 0.03; all p < 0.001) were found for the ankle plantar flexion. There was no significant difference in the tendon and muscle architecture between the male and female subjects at relaxed condition (All p> 0.05). However, when comparing the muscle and tendon in the same plantar flexion range, the changes of PA (p = 0.06) for the men were larger than those of the women which could be ascribed to pennation effect under the similar FL change. Moreover, the lack of FL change at the initial part of plantar flexion phase and significantly larger area of force-length relationship for the men inferred the higher mechanical power of MTU during plantar flexion. In addition, the larger regression coefficients between the fascicle length and muscle force (female: 0.96 ± 0.37; male: 1.28 ± 0.34) implied that higher muscle stiffness for the men was recruited in the demands of maintaining biomechanical stability for the heavier body. The body weight might be the main contributor to the gender difference in muscle and tendon behaviors when performing dynamic tasks. Further studies with a large group of subjects and other types of motion including walking should be conducted for full understanding of muscle activity differences using the newly developed system.In conclusion, we have demonstratedThe human motion analysis system with sononmyography was successfully developed with considerations of probe orientation influence and temporal offset between the ultrasound scanner and the vision-based motion capture system.The innovative image processing techniques were advanced to facilitate the measurements of FL and PA with multi-resolution analysis and prior knowledge about fascicles.For the first time, the gender effects on the changes of the muscle and tendon in performing dynamic movement were investigated.
|Description:||PolyU Library Call No.: [THS] LG51 .H577P BME 2015 Zhou
xxii, 189 pages :illustrations ;30 cm
|URI:||http://hdl.handle.net/10397/35215||Rights:||All rights reserved.|
|Appears in Collections:||Thesis|
Show full item record
Files in This Item:
|b28164210_link.htm||For PolyU Users||203 B||HTML||View/Open|
|b28164210_ir.pdf||For All Users (Non-printable)||9.2 MB||Adobe PDF||View/Open|
Citations as of Dec 9, 2018
Citations as of Dec 9, 2018
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.