Research On Modeling Of Puncture Force And Localization Of The Needle During Puncture Surgery

Compared with traditional open surgery, puncture surgery has the advantage of small wound, light pain, quick recovery and fewer complications, so it is finding wider and wider application in the medical treatment process. Currently, two-dimensional ultrasound imaging is mainly used for navigation during puncture. The imperfect reflection of lesion area and inflexible operation will reduce puncture success rate. Robot system is completely controllable, dependable and programmable, which makes it perform better than free-hand biopsy. So combine three-dimensional ultrasound(3D US) imaging and robot system together to perform a puncture operation has become a research focus.Modeling of the interaction between soft tissue and needle and real-time localization of the needle during operation are mainly discussed in this paper. Details are described below:This paper mainly focuses on kidney puncture surgery. Interaction during puncture is modeling and simulated using Abaqus to obtain the displacement curve of tissue and the force curve of needle during the whole progress.Analysis is performed about the curves to direct the real surgery and make it be more accurate.In order to locate the needle accurately and timely, real-time localization of the needle during the surgery is needed. Due to the complexity of real US data, a new location algorithm named multi-scale random sample consensus(MS-RANSAC) which is improved by traditional RANSAC is proposed in this paper. MS-RANSAC can locate the puncture needle correctly in real US dataset.The large data quantity of 3D US data give rise to a long computing time. In order to improve efficiency of the localization algorithm, parallel strategy is designed and implemented using CUDA, making it usable in practical application. There datasets are chosen to test the algorithm result and efficiency. The results show that, MS-RANSAC can locate the needle correctly in complex background. The speed of the algorithm is greatly improved after implemented by CUDA.