An Automatic Classification Of Proximal Tibia Morphological Characterization And Optimal Design Of Steel Plates Based On Machine Learning

With the development of the Chinese economy and society,we are entering an aging society,accompanied by some middle-aged and elderly diseases that increased year by year,such as a tibial fracture.Plate treatment is the standard treatment of tibia fracture and the current tibial plates do not meet the Chinese population’s standards.Therefore,it is particularly important to conduct a professional classification study on the morphological characteristics of the proximal tibia of patients,and further develop an individualized treatment for patients according to the morphological types and fracture degree of the proximal tibia.Morphological characteristics of the tibia can help orthopedic surgeons design internal fixation plates matched accurately.At present,the measurement of tibial morphology mainly depends on manual measurement by doctors,which requires a lot of manual interaction.By cooperating with the General Hospital of the People’s Liberation Army of China and relying on the 3D tibial CT data provided by the hospital and relevant medical expertise,this project carried out the research on the classification of proximal tibial morphological characteristics and the optimization design of steel plate based on machine learning.The main work of this paper is as follows:(1)During the construction of the tibial statistical model,a tibial point cloud registration method based on feature points was proposed.Firstly,the Fast Point Feature Histograms algorithm(FPFH)and Balanced Clustering Reducing and Clustering using Hierarchies(BIRCH)were used to achieve the fast registration.Then,Coherent Point Drift(CPD)and Iterative Closest Point(ICP)were used to finish the accurate registration.Finally,the corresponding relationship of the tibial point cloud was obtained and the statistical model was constructed.The high efficiency and accuracy of this method were verified by the comparison experiment of the point cloud registration algorithm.(2)Aiming at the time-consuming and laborious problem of manual measurement of morphological parameters,an automatic measurement module of tibial morphological characteristics based on 3D Slicer software was implemented.Firstly,the orthopedic surgeon marked the required markers on the statistical model.Next,we established the point correspondence between the statistical model of the tibia and the tibial model to be measured by using the deformable model matching method.Finally,the morphological parameters of the proximal tibia of each patient were automatically measured according to the position relationship of corresponding points.By comparing the points marked manually with those matched automatically,the performance of automatic matching was evaluated,and the accuracy of automatic measurement was further verified.(3)A classification method based on morphological characteristics of the proximal tibia was proposed.Firstly,morphological features of the proximal tibia were extracted by a morphological fitting algorithm.Next,the features were clustered by the K-means clustering algorithm based on density peak,realizing automatic classification of morphological features of the proximal tibia.Finally,according to the classification results,parameters were provided for the optimal design of different types of tibia plates.The results show that the clustering method proposed in this paper works well,and the sample difference between different categories was large and the sample difference within the category was small.The clustering results provided an effective basis for optimizing the design of different classes of tibial plates.(4)Based on the module extension function of the 3D Slicer software platform,the morphological feature classification and steel plate-aided optimization design system of the proximal tibia were realized.The system improves the efficiency and accuracy of point cloud registration,realizes the automatic measurement of tibial morphological parameters,and effectively solves the problem of tibial plate mismatch in Chinese people.Through the module expansion function of the system,the automatic measurement technology is promoted,which has important application value to the clinical diagnosis of various orthopedic diseases.