The Algorithm Of Hip Contact Pressure Estimation During Dynamic Process

The hip joint plays an important role in the body weight in the daily activities,and the hip joint is often affected by various diseases due to the fact that the hip joint is often affected by the contact pressure.The abnormal contact pressure is the main cause of the hip arthritis.Therefore,it is an effective way to understand the stress of the hip joint,which is far from the hip joint injury and degeneration.Under the condition of no damage to the experimenter,the method of quantifying the contact pressure of the hip joint is of great significance in the clinical treatment and the improvement of our understanding of the hip joint.This thesis investigates the change of the contact pressure of the acetabulum in a series of continuous dynamic postures.The main research work and innovations are as follows:1.This work proposes the acetabulum contact pressure estimation based on the improved iterative closest point algorithm and the finite element method.Due to the fact that there exists little work to investigate the contact pressure of the hip joint in a series of continuous dynamic postures of the daily activities,in this work,the motion of the five daily life of 20 subjects is simulated,including the slow walking,the normal walking,the fast walking,the upper and the lower stairs,which serves as the motion model.Moreover,in the complex dynamic postures,the relationship of the anatomical model and the motion model is analyzed.A two-step modified method of iterative closest points is proposed to calculate the transformation matrixes that are applied on three parts of the hip joint.So the relative positions of the anatomical model at any time are obtained in the motion processing,which can implement that the motion model and the anatomical model are registered.In the processing of registration,the proposed modified method of iterative closest points can avoid markers shift due to muscle movement.Finally,the finite element method is utilized to predict the contact pressure distribution of the acetabulum.Experimental results demonstrate that: 1)the region of the maximum value of the contact pressure is in the front half of the middle region and the front half of the rear region;2)the peak posture of the contact pressure is mostly the heel;3)the area of the contact pressure in one cycle is shifted from the front region to the intermediate region,and then transition from the intermediate region to the rear region,and then back to the front region.2.This work proposes deep neural networks based contact pressure estimation of the acetabulum.The model utilizes the architecture of encoder-decoder to implement the fast prediction of the acetabulum contact pressure,which avoid the long-time calculation of the finite element-based method.First,the fully-connected neural networks(FNN)encode four acetabulum components(the acetabulum shape,the femur shape,the force bearing on the central of femur,and the distance between acetabulum and femur)into the contact pressure code.Then,the contact pressure code is decoded into the contact pressure distribution by deconvolution neural networks.Different from previous methods that utilized principal component analysis(PCA)to reduce the dimension,our proposed FNN combines the dimension reduction with the encoding procedure in an end-to-end manner.Depending on this synergistic effect,FNN can preserve effective information for acetabulum shape and femur shape,and enable the predicted acetabulum contact pressure code to contain representative features.Experimental results show that the dimension of shape and the number of layers for FNN has an influence on the performance of the model.In addition,our proposed method can achieve more lower error rates in comparisons to PCA-based methods.In the range of certain error,the acetabular contact pressure estimation from the model can replace the results of the finite element method,and the time cost can reduce mostly in comparison to finite element method.Therefore,the estimation result of the model can be fed back to the doctor in time.