Study On Ultrasound Image Processing Algorithm And Device Development For Real-time Quantitative Detection Of Musculoskeletal Dynamic Changes

Musculoskeletal plays an important role in human life activities and its functional changes are closely related to its own morphological structure.The study of the relationship between musculoskeletal structure and function can help us to understand the physiology basics of force to guide clinical practices.Ultrasonography has been widely used in the research of muscle properties since it is real-time,nonradiative,and inexpensive.However,most of the current research methods are to acquire ultrasound images or video signals,and then perform off-line analysis to extract relevant musculoskeletal morphology parameters.Off-line analysis is laborintensive,time-consuming,relies on the clinical experience of the sonographer and cannot display musculoskeletal function in real time.The goal of this thesis is to study the musculoskeletal ultrasound image processing algorithm and transplant the algorithm to the ultrasound imaging equipment to develop an ultrasound imaging device for real-time quantitative detection of musculoskeletal dynamic changes.The device can also display the musculoskeletal grayscale ultrasound image,as well as the musculoskeletal length,thickness,and pinnate angle parameters.The main work of this thesis focuses on three aspects:(1)The musculoskeletal morphology parameter detection algorithm based on traditional image processing method was studied,and the software for real-time image processing was designed.The data collected by the clinical experiment was processed,and the accuracy of the algorithm was analyzed by using the manual processing results as a reference.The results show that the method of this paper has better accuracy and detection speed,which can meet the requirements of real-time detection.And the method proposed in this thesis can extract three parameters of musculoskeletal length,thickness and pinnate angle.These three parameters can better reflect the changes of musculoskeletal morphology,while other researchers only studied the extraction algorithm of one of the parameters.(2)The musculoskeletal morphology parameter detection algorithm based on deep learning method is studied.Independently built a deep learning experimental platform for musculoskeletal morphology parameter extraction,training,testing and optimization of the network,and achieved good results.Few researchers have used deep learning methods to segment muscle texture structure,no researchers have used deep learning methods to extract musculoskeletal parameters.This thesis has tried in this area and reached a clinically acceptable error range.(3)Transplanting musculoskeletal morphology parameter detection algorithm based on traditional image processing method to ultrasound imaging device,and an ultrasound imaging device for real-time quantitative detection of musculoskeletal dynamic changes was developed.Compared with the products of major ultrasonic manufacturers at home and abroad,the device can display the curve of three musculoskeletal morphological parameters: length,thickness and feather angle in real time,which is not available in other products at present and can be provided for clinical diagnosis.