Gait Analysis And Application In The Diagnosis Of KOA Based On Deterministic Learning Theory

Knee osteoarthritis(KOA)is a chronic degenerative bone and joint disease characterized by knee cartilage degeneration,destruction,and bone hyperplasia.It is clinically manifested as knee pain,stiffness,swelling,and limited mobility.In severe cases,it can cause joint deformity,and even the final loss of knee function,which greatly affects the quality of life of the elderly At present,X-ray are mainly used in the clinical diagnosis of KOA.This imaging technique can only evaluate KOA under static conditions,and cannot evaluate joint function under exercise This article attempts to find the relationship between the gait information of KOA patients and the Kellgren-Lawrence classification of X-ray assessment,and uses gait analysis technology to make an diagnosis of Kellgren-Lawrence classification of KOA patients.This gait analysis takes into account the dynamic information of the joints,making up for the lack of imaging technology that can only evaluate the condition of KOA patients under static conditions.In this paper,Kinect2.0 sensor is used to collect gait data of KOA patients,and the deterministic learning theory is applied to KOA Kellgren-Lawrence classification diagnosis.Finally,a KOA Kellgren-Lawrence classification auxiliary diagnosis system is designed and developed based on MATLAB and Python Deterministic learning theory is a locally accurate neural network modeling method thatcombines adaptive control,Lyapunov stability analysis,RBF neural network and persistence of excitation(PE)conditions,and can complete knowledge acquisition,expression,storage and reuse in unknown dynamic environments.In this paper,by confirming deterministic learning theory,the local accurate recognition of the gait dynamics of KOA patients is achieved,and the learned knowledge is stored with the constant RBF neural network weights.In this way,the time-varying gait dynamic pattern can be represented by a time-invariant neural network approximation.A set of dynamic estimators can be constructed using the stored constant RBF neural network weights.Comparing the test pattern with this set of estimators can generate a set of L1 norm errors.By selecting the smallest norm errors,The most similar training pat-tern is identified,and the Kellgren-Lawrence classification label of this training pattern is used as the KOA Kellgren-Lawrence classification diagnosis result of this test pattern.The overall diagnostic accuracy of the algorithm designed in this paper has reached 85.99%Based on deterministic learning theory,this paper studies and explores the practical appli-cation of KOA Kellgren-Lawrence classification algorithm on the basis of previous work and mainly designs and develops the KOA Kellgren-Lawrence classification diagnosis system.This system uses Kinect2.0 as a sensor,and realizes the functions of gait data acquisition module,data preprocessing module,gait pattern training module,KOA Kellgren-Lawrence classification di-agnosis module and pattern library management module,which can provide gait feature timing diagrams,neural network layout diagrams,gait three-dimensional dynamics diagrams,residual diagrams and other charts for researchers to analyze.The accuracy of KOA Kellgren-Lawrence classification diagnosis was verified on the self-built KOA patient gait database,which proved the engineering application value of the system.