Research On Motion Estimation Methods For Skeletal Muscle Contraction Based On Ultrasonography And Their Applications

Skeletal muscle is an important tissue of the human body’s composition. Tounderstand the role of the skeletal muscle in body motions will be helpful for re-solving some potential problems about musculoskeletal system in diferent humangroups. As a non-invasive, real-time and convenient imaging technology, ultra-sonography has been increasingly used in the research on skeletal muscle. Sincethe1990s, researchers attempted to gain muscle structural parameters from theultrasound images, such as muscle cross-sectional area, muscle thickness, penna-tion angle and muscle fber length, etc., so as to refect the state changes of themuscle. However, apart from the physiological structure, quantitative estimateson local motions will be helpful for providing more information about the changesof muscle tissue in the process of muscle contraction. In this dissertation, we adoptmainly the motion estimation method to describe the movement of muscles.Variational optical fow computation has obvious advantages in many aspectssuch as the universality of the model and the attainment of dense optical fow feld,etc. At the same time, due to some characteristics such as its stronger adaptability,close to the image features, etc., it has become one of the mainstream methodsabout the motion estimation. In this dissertation, according to assumptions suchas invariant brightness, invariant gradient, etc., diferent data terms and smoothterms are constructed, eight diferent optical fow models are derived, and thesplit Bregman (SB) method is adopted to solve the typical H-S model, L1-L2model and L1-High order model, respectively. Due to the characteristics such asits fast computational speed, easy implementation of numerical experiments, etc.,the split Bregman method has been widely used in image denoising, deblurringand compressed sensing, etc. For the solution of the system of linear equations inthe process of solving the model, the preconditioned conjugate gradient method isadopted to improve the computation speed, and the theoretical analysis that itscoefcient matrix is symmetric positive defnite is also given.In the numerical calculation part, real scene images and ultrasound simulationimages are respectively used to compare performance of diferent motion estima-tion algorithms. Under the multi-scale pyramid optical fow computing framework,those algorithms mentioned in the above are implemented. The results demon-strate that the L1-High order model synthesizing the advantages of the L1termand the higher order term has certain advantages in improving the precision of optical fow computation.In the application part, experiments are frst designed to collect the ultrasoundimages in the process of the contraction of gastrocnemius muscle and quadricepsfemoris muscle. Then the split Bregman method for the L1-High order model isapplied to calculate the related motion fled. Respectively according to their d-iferent physiological structure characteristics, the diferent characteristics of themovement are extracted to characterize the functional status of the muscle. Atthe same time, a comparative analysis with synchronous acquisition of EMG sig-nals and Torque signals is made. The results show that features of the musclemovement extracted from ultrasonic image sequences by adopting the variation-al optical fow algorithm can quantitatively express many important informationsabout the morphology and the function of the skeletal muscle movement and havethe potential value of practical applications.