Research On Automatic Target Location And Tracking Method For Transcranial Magnetic Stimulation Robot

Transcranial magnetic stimulation technology is a non-invasive brain nerve regulation technology that has been widely used in clinical practice due to its advantages of non-invasive,safe,reversible,and painless.However,in the actual treatment process,the random movement of the patient’s head can cause the treatment to miss the target and greatly affect the treatment effect.At present,transcranial magnetic navigation systems that address this issue typically use magnetic resonance imaging(MRI)image navigation.However,due to the cumbersome preparation work before and during treatment,the clinical applicability needs to be improved.This article studies a method for automatic target localization and tracking using depth cameras and two-dimensional markers to address the above issues.This method not only meets clinical usage needs but also better conforms to the operating habits of doctors;At the same time,this article also proposes a fast repetitive target localization method based on facial feature points,which can effectively reduce the preparation time for patients before re treatment.The specific research content is as follows:(1)In response to the requirements of the target automatic positioning and tracking system,system performance indicators were proposed,and the treatment process and overall plan including hardware platform construction and algorithm modules were designed.(2)A target localization algorithm based on 2D labeled corner spatial coordinates has been proposed.Firstly,the type of two-dimensional markers was determined and the size parameters and placement positions of the markers were determined through testing.Multiple image processing methods were used to achieve the recognition and positioning of the markers;Secondly,in response to the noise and missing issues in depth images captured by depth cameras,this paper uses methods such as lowpass,spatial edge preservation filtering,and hole filling to repair depth images.By aligning the color image with the depth image and converting the pixel coordinate system to the world coordinate system,the twodimensional labeled corner spatial coordinates are calculated;Finally,a target localization algorithm using a depth camera and two-dimensional markers was proposed,achieving an automatic target localization function that is more in line with doctors’ operating habits.(3)A target tracking algorithm based on multiple two-dimensional markers has been proposed.Firstly,the calculation principle of the relative position relationship between the depth camera coordinate systems on both sides of the positioning and tracking sides was studied;Then,by solving the matrix equation to obtain the coordinate conversion matrix from the depth camera to the robot,the target spatial coordinate information was transformed into the robot base coordinate system;Finally,a target tracking algorithm using two depth cameras and two-dimensional markers was proposed,achieving the function of target tracking.(4)A fast and repetitive target localization method has been proposed.This article implements facial region detection and key point position detection for patients based on multitasking convolutional neural networks,as well as facial recognition for facial regions based on Face Net networks;A method is proposed to establish a facial coordinate system based on the key points of the patient’s face collected by a depth camera.By combining the facial coordinate system with target positioning information in the patient information database,the rapid and repetitive target positioning function of the transcranial magnetic stimulation robot is achieved.(5)We designed and built a hardware experimental platform for the transcranial magnetic stimulation robot system,designed an experimental process to verify the system performance,and verified the system accuracy through experiments.The experimental results show that the initial positioning accuracy of the proposed transcranial magnetic stimulation robot target automatic positioning and tracking method is 4.32 mm,with a tracking accuracy of 7.28 mm based on the initial positioning,6.47 mm based on repeated target positioning,and 10.21 mm based on repeated target positioning;The response time of the system includes a response time of 0.4 seconds for target tracking and 2.24 seconds for repeated target localization,which can meet the clinical treatment needs of transcranial magnetic therapy.