Automatic Positioning Of Epidural Anesthesia Guided By Ultrasound

Epidural anesthesia(EA)is a central nerve blockade technique,which is used for painless delivery of maternal and postoperative pain relief.Anesthetists are required to identify the needle entry site for epidural anesthesia.However,in the practice,there still remains a high failure rate.In recent years,ultrasound is widely used to assist the anesthesia,but operators often can not accurately interpret the ultrasound image of the lumbar spine.To solve this problem,it is necessary to design an ultrasound-guided automatic epidural anesthesia positioning system,which is of great significance to promote medical automation and social harmony.In this thesis,we developed a system of automatic interpretation of ultrasound images,and presented the overall framework for the system,according to the analysis of the system functions,with focus on the lumbar vertebral longitudinal view processing system and lateral view processing.The method can achieve the real-time processing of ultrasound image and identify the needle insertion position accurately.The main contribution of this thesis includes five parts as follows:1.In view of the current characteristics of epidural anesthesia and the research of ultrasound-guided EA,the advantages and disadvantages of the EA.are analyzed,the demands and overall design of the system studied in this paper was given.And the main innovations of our system were listed.2.The characteristics of lumbar ultrasound images were analyzed.In light of the problem that the ultrasound images are obscure and have low resolution,a robust anisotropic diffusion filter was proposed,which can smooth the speckle noise.Meanwhile,a method of contrast enhancement based on morphological transformation was used to enhance the brightness and highlight the main anatomical structures.Then,an adaptive threshold binarization algorithm was used for the image preprocessing.3.An automatic lumbar level recognition system was proposed by analyzing the characteristics of longitudinal view of lumbar spine.The image processing procedure starts with sacral Identification,using support vector machine(SVM)classification.After the sacrum is identified,the panoramic image was spliced to create a complete spinous processes structure.In the meantime,spinous levels were recognized and counted in the panoramic image,to achieve the rough positioning for EA.4.In the transverse view of the lumbar spine,the interspinous images are the best needle site,so they need to be separated from the bone images.Therefore,we proposed a cascade classifier composed of four weak classifiers,with feature extraction based on template matching and midline detection,which can successfully identify interspinous images and have high accuracy,so as to realize precise positioning of EA.5.A user-friendly GUI is designed to show the results in real time,guide the anesthetist intuitively,and further conceal the complexity of the algorithm.Simulation results show that the proposed ultrasound-guided automatic epidural anesthesia positioning system is feasible and effective,which not only can improve the success rate of puncture,but also has great significance for the development of medical equipment,and has a promising market.