Intelligent Osteotomy System For Virtual Surgery In Orthopedics

In order to achieve more accurate deformity femoral osteotomy and pelvic tumor resection,a software system for deformity femoral osteotomy and pelvic tumor resection was designed to facilitate the clinical preoperative simulation.The system is divided into two major modules:the first module is the malformed femur osteotomy module.3D modeling was simulated using Mimics 17.0 and saved as STL file.Then,MFC and OpenGL was utilized to build the deformed femur osteotomy system,the accurate osteotome position and osteotomy angle were obtained by fitting the axis of the deformed femoral shaft and the geometric center of the greater trochanter of femoral head.After the calculation is completed,the osteotomy and splicing functions are simulated through the interactive operation of the mouse and keyboard.A case of abnormal femoral STL model was introduced into the osteotomy system and whose collodiaphyseal angle was measured as 57.7°,input osteotomy angle was 60°,and the postoperative collodiaphyseal angle was measured to be 102.3°after completing the osteotomy simulation.The second module was the pelvic bone tumor resection module.Deep convolutional neural network is a kind of neural network model in deep learning,which can recognize the position of the target in the image data and distinguish it from the background.In order to obtain accurate boundaries of pelvic tumors.The depth convolutional neural network was used to classify the input CT image data of the pelvis at the pixel level,predict the location of the bone tumor in the lamellar slice,and then carry out three-dimensional reconstruction and cutting.In order to obtain better characteristics,the roll base of the full convolution neural network(FCN)was replaced by the roll base of VGG19,and the model of vggl9-fcn deep neural network was built.The model was trained by 520 CT tomography images from 30 patients with pelvic bone tumor.Using an unknown case as test data,,by comparing the different activation functions of neurons and the number of training,finally selected Ruel as activation functions of neurons,the number of iterations for 12000 times.The model automatically calibrated the location of pelvic bone tumor in the layer slice,the prediction accuracy is 98.3%.The CT slice was input into the mimics,then the three-dimensional modeling was carried out.The generated model was converted into STL(ASCII)file output and imported into the system,the method of curvilinear is used to cutting tumor model.The experimental results show that the system can quickly provide doctors with a more accurate osteotomy program and resection of the tumor,which will save the doctor's clinical diagnosis time.