CN106650195A - Gait analysis method for assisting in screening meniscus injuries - Google Patents
Gait analysis method for assisting in screening meniscus injuries Download PDFInfo
- Publication number
- CN106650195A CN106650195A CN201610363760.8A CN201610363760A CN106650195A CN 106650195 A CN106650195 A CN 106650195A CN 201610363760 A CN201610363760 A CN 201610363760A CN 106650195 A CN106650195 A CN 106650195A
- Authority
- CN
- China
- Prior art keywords
- gait
- neural network
- knee joint
- dynamic
- meniscus
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000005021 gait Effects 0.000 title claims abstract description 160
- 206010072970 Meniscus injury Diseases 0.000 title claims abstract description 53
- 238000004458 analytical method Methods 0.000 title claims abstract description 21
- 238000012216 screening Methods 0.000 title claims abstract description 21
- 238000013528 artificial neural network Methods 0.000 claims abstract description 84
- 238000012549 training Methods 0.000 claims abstract description 34
- 238000012360 testing method Methods 0.000 claims abstract description 14
- 210000000629 knee joint Anatomy 0.000 claims description 56
- 230000006378 damage Effects 0.000 claims description 21
- 210000002569 neuron Anatomy 0.000 claims description 21
- 210000000689 upper leg Anatomy 0.000 claims description 20
- 238000006073 displacement reaction Methods 0.000 claims description 17
- 210000002303 tibia Anatomy 0.000 claims description 17
- 230000005499 meniscus Effects 0.000 claims description 16
- 208000027418 Wounds and injury Diseases 0.000 claims description 13
- 208000014674 injury Diseases 0.000 claims description 13
- 230000005284 excitation Effects 0.000 claims description 8
- 206010017577 Gait disturbance Diseases 0.000 claims description 6
- 238000000034 method Methods 0.000 claims description 6
- 238000013461 design Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000005312 nonlinear dynamic Methods 0.000 claims description 3
- 230000003287 optical effect Effects 0.000 claims description 3
- 230000033764 rhythmic process Effects 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 13
- 238000003745 diagnosis Methods 0.000 abstract description 3
- 238000010882 preoperative diagnosis Methods 0.000 abstract description 3
- 210000003127 knee Anatomy 0.000 description 5
- 238000001356 surgical procedure Methods 0.000 description 5
- 238000002595 magnetic resonance imaging Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000033001 locomotion Effects 0.000 description 3
- 230000035945 sensitivity Effects 0.000 description 3
- 210000000988 bone and bone Anatomy 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 210000002414 leg Anatomy 0.000 description 2
- 210000003141 lower extremity Anatomy 0.000 description 2
- 206010003694 Atrophy Diseases 0.000 description 1
- 208000012659 Joint disease Diseases 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 210000003484 anatomy Anatomy 0.000 description 1
- 230000037444 atrophy Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000000845 cartilage Anatomy 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000001727 in vivo Methods 0.000 description 1
- 210000001503 joint Anatomy 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 210000003041 ligament Anatomy 0.000 description 1
- 210000001699 lower leg Anatomy 0.000 description 1
- 210000003205 muscle Anatomy 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 210000003314 quadriceps muscle Anatomy 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035939 shock Effects 0.000 description 1
- 230000000087 stabilizing effect Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention discloses a gait analysis method for assisting in screening meniscus injuries. The gait analysis method includes the steps that gait feature data of meniscus injury patients and healthy able-bodied persons is collected and subjected to neural network modeling and training; proper gait features are selected through kinematic and dynamic analysis of body gait, and gait dynamic knowledge is obtained; then the gait feature data of the meniscus injury patients serves as a test set. In this way, screening is accurately and rapidly assisted, it is avoided that non-invasive diagnosis is carried out under MRI and an arthroscopy, the accuracy of preoperative diagnosis is greatly improved, and the detection cost and the detection time are saved. The gait analysis method for assisting in screening meniscus injuries can be widely applied to the field of medical treatment.
Description
Technical Field
The invention relates to the field of medical treatment, in particular to a gait analysis method for assisting in screening meniscus injury.
Background
The meniscus is a fibrous cartilage plate, has a meniscus shape at the inner and outer sides, and is located between the tibial plateau and the inner and outer side bones of the femur, and has an inner edge, an outer edge, and an anterior angle and a posterior angle. Meniscal injury is a knee joint disease mainly manifested by the phenomenon of soft leg or knee joint interlocking of partial patients, quadriceps femoris atrophy and limited tenderness of fixed knee joint gaps. Meniscal injuries are mostly caused by external torsional force, when one leg bears weight, and the lower leg is fixed at a semi-flexion and abduction position, the body and the thigh part rotate inwards suddenly, and the inner meniscus is subjected to rotary pressure between the femur and the tibia, so that the meniscus is torn. The meniscus is recessed above and flat below, is approximately wedge-shaped, is embedded in the joint space, and is an indispensable part in the complex structure for stabilizing the knee joint. The meniscus has elasticity, and can reduce the shock during the joint movement. When the knee joint bends and stretches, the meniscus concave surface and the femur move; when the knee joint rotates, movement occurs between the inferior meniscus and the tibial plateau, and thus damage to the meniscus occurs more below it. The current medical research results show that when meniscus injury occurs, gait parameters such as knee joint angle, displacement, contact force and moment can be obviously changed, so that gait abnormality of a patient is caused.
As an emerging technology, the gait analysis technology combines biomechanics, kinematics and anatomy, and can well complete the detection of the walking posture of a meniscus injury patient. The gait of a person can reflect the pathological characteristics of the person from one side, and particularly, the normality of bones, joints, muscles and ligaments of lower limbs can be objectively evaluated. By analyzing the gait and discussing the relevant variables of kinematics and dynamics such as joint angle, displacement, force, moment, power and the like in the gait motion, the gait rehabilitation robot can conveniently, simply and non-invasively help doctors to scientifically analyze the etiology and assist in screening, diagnose the state of an illness, evaluate the curative effect and guide the walking training of patients, and is widely applied to the treatment of the diseases of the lower limbs, the knees and the joints.
Magnetic Resonance Imaging (MRI) is currently the best imaging diagnostic method, and arthroscopic surgery is the "gold standard" for diagnosis and treatment of meniscal lesions. Both have certain drawbacks, for example, they are expensive; arthroscopic surgery belongs to invasive detection; patients with pacemakers or sites with certain metallic foreign bodies cannot be examined by MRI; most MRI equipment inspection space is comparatively closed, and scan time is relatively long, and some patients can not accomplish the inspection because of the fear cooperation. With the transformation of social life, labor, exercise level and the like, meniscus injuries are greatly changed in the aspects of onset age, occupational distribution and the like, more and more patients with meniscus injuries of knee joint gradually receive minimally invasive examination and treatment of arthroscopic knee joint surgery, and the expectation on the surgery is increased. Therefore, how to further improve the accuracy of preoperative diagnosis and screening and improve the curative effect of surgery for a patient clinically diagnosed with meniscus injury becomes an increasingly focused target for clinicians.
Disclosure of Invention
In order to solve the technical problems, the invention aims to: the gait analysis method for modeling the nonlinear gait system dynamics and distinguishing the two groups based on the difference in the gait system dynamics to realize the auxiliary screening and detection of meniscus injuries is provided.
The technical scheme adopted by the invention is as follows: a gait analysis method for assisting in screening meniscus injuries comprises the following steps:
A. respectively collecting knee joint gait feature data of a plurality of groups of meniscus injury patients and healthy normal people, extracting gait feature variables, and forming training sets by the collected knee joint gait feature data of the plurality of groups of meniscus injury patients and healthy normal people;
B. modeling unknown nonlinear gait system dynamics of meniscus injury patients and healthy normal persons in a training set according to the gait feature variables extracted in the step 1, and approximating the unknown nonlinear gait system dynamics locally by adopting a neural network identifier;
C. establishing a constant neural network by using the learning and training result of the RBF neural network identifier, and storing the learned nonlinear gait system dynamics knowledge in the form of constant neural network weight to form a training gait pattern library;
D. collecting knee joint gait feature data of a meniscus injury patient, extracting gait feature variables, and forming a test set by the collected knee joint gait feature data of the meniscus injury patient;
E. a group of dynamic estimators are constructed by utilizing a constant neural network, nonlinear gait system dynamics knowledge corresponding to meniscus injury patients and healthy normal people in a training gait pattern library is embedded into the dynamic estimators, the gait feature data of the meniscus injury patients are differentiated from the group of dynamic estimators to form a group of classification error results, and abnormal gaits of the meniscus injury patients are calculated according to minimum errors.
Further, knee joint gait characteristic data are collected through an optical sensor in the step A and the step D.
Furthermore, the gait characteristic data of the knee joint in the step A and the step B comprises knee joint angle characteristic data and knee joint displacement characteristic data, the knee joint angle characteristic data comprises an internal rotation angle, an external rotation angle and an internal and external rotation angle of the knee joint femur relative to the tibia, and the knee joint displacement characteristic data comprises front and back displacement characteristic data of the knee joint femur relative to the tibia.
Further, the unknown nonlinear gait system dynamic modeling in the step B is represented as:
wherein x is [ x ]1，…，xn]T∈RnB, extracting gait characteristic variables from the gait characteristic variables extracted in the step A, wherein p is a system constant parameter value, and n is the dimension of the gait characteristic variables; f (x; p) ═ F1(x；p)，…，fn(x；p)]TIs a smooth and unknown nonlinear dynamic variable representing the gait system dynamics of meniscus injured patients and healthy normal people, and v (x; p) ═ v1(x；p)，…，vn(x；p)]TFor modeling uncertainty, the two are merged intoDefined as the general nonlinear gait system dynamics.
Further, in the step B, a dynamic RBF neural network is adopted to construct the neural network identifier.
Further, the dynamic RBF neural network identifier is of the form:
whereinIs the state of the neural network identifier; a ═ diag [ a ═ d1，…，an]Is a diagonal matrix, aiSatisfying 0 < | a as a constant of designi|＜1，Is a dynamic RBF neural network for approximating unknown general non-linear gait system dynamicsS(x)＝[S1(||X-ξ1||，…，SN(||X-ξn||]TIs a Gaussian radial basis function, N > 1 is the number of neural network nodes, ξiIs the central point of the neuron, RBF neural network weightThe regulation law of (2) is as follows:
wherein i represents the ith dimension variable in the n dimension gait characteristic variables,is the error in the state of the device,σimore than 0 is the adjusting parameter of the rhythm regulation, and the weight of the dynamic RBF neural networkInitial value of (2)
Further, the locally accurate modeling of the general nonlinear gait system dynamics may be formulated as follows:
wherein,i1is an approximation error; the local accurate modeling refers to the approximation of the internal system dynamic track along the gait feature data through the RBF neural network, and the internal dynamic far away from the track is not approximated.
Further, the step C specifically includes: according to the determined learning theory, the neurons of the RBF neural network along the characteristic track of the gait system meet the continuous excitation condition, the weight of the RBF neural network converges to an optimal value, the average value of the weight within a period of time after the weight converges is taken as the learning training result, the constant neural network is established by using the result, and the learned gait system dynamics knowledge is stored in the form of the constant neural network weight to form a training gait pattern library.
Further, the RBF neural network weight is adjusted according to the Lyapunov stability theorem and the deterministic learning theory, so that both the state error and the weight estimation are bounded and exponentially converged, wherein the weight convergence of the RBF neural network has two conditions:
in the first case: the neurons of the RBF neural network along the gait characteristic regression trajectory meet continuous excitation conditions, and the weight of the neurons converges into a small neighborhood of an optimal value;
in the second case: neurons of the RBF neural network far away from the gait feature regression trajectory are not excited and are not adjusted, and the weight value is approximate to zero.
The invention has the beneficial effects that: the method collects the gait feature data of the meniscus injury patient and healthy normal people, carries out modeling and training, selects proper gait features through the kinematics and dynamics analysis of human gait, acquires gait dynamics knowledge, and then takes the gait feature data of the meniscus injury patient as a test set, thereby accurately and quickly assisting screening, avoiding non-invasive diagnosis under MRI and arthroscopy, greatly improving the accuracy of preoperative diagnosis, and saving detection cost and time.
Drawings
FIG. 1 is a flow chart of the steps of the present invention;
FIG. 2 is a schematic representation of the medial-lateral rotation angle characteristics of a femur of a knee joint relative to a tibia of a patient having meniscal damage as used in the examples of the present invention;
FIG. 3 is a schematic representation of the flexion and extension angle characteristics of the femur of the knee joint relative to the tibia of a patient with meniscal damage in an embodiment of the present invention;
FIG. 4 is a schematic representation of the varus-valgus angulation feature of a femur of a knee joint relative to a tibia of a patient having meniscal injury as used in embodiments of the present invention;
FIG. 5 is a graphical representation of the anterior-posterior displacement characteristics of a femur of a knee joint relative to a tibia of a patient having meniscal damage as used in the examples of the present invention;
FIG. 6 is a schematic illustration of the internal and external rotation angle characteristics of a healthy normal human knee femur relative to a tibia as used in an embodiment of the present invention;
FIG. 7 is a schematic view of a flexion-extension angle characteristic of a femur of a knee joint of a healthy normal person relative to a tibia in an embodiment of the present invention;
FIG. 8 is a schematic illustration of the varus-valgus angulation characteristics of a healthy normal human knee femur relative to a tibia as used in an embodiment of the present invention;
FIG. 9 is a schematic representation of the anteroposterior displacement characteristics of a healthy normal human knee femur relative to a tibia as used in an embodiment of the present invention;
FIG. 10 is a simplified schematic diagram of the topology of an RBF neural network employed in an embodiment of the present invention;
FIG. 11 is a diagram illustrating the convergence of RBF neural network weights in an embodiment of the present invention.
Detailed Description
The following further describes embodiments of the present invention with reference to the accompanying drawings:
a gait analysis method for assisting in screening meniscus injuries comprises the following steps:
A. respectively collecting knee joint gait feature data of a plurality of groups of meniscus injury patients and healthy normal people, extracting gait feature variables, and forming training sets by the collected knee joint gait feature data of the plurality of groups of meniscus injury patients and healthy normal people;
B. modeling unknown nonlinear gait system dynamics of meniscus injury patients and healthy normal persons in a training set according to the gait feature variables extracted in the step 1, and approximating the unknown nonlinear gait system dynamics locally by adopting a neural network identifier;
C. establishing a constant neural network by using the learning and training result of the RBF neural network identifier, and storing the learned nonlinear gait system dynamics knowledge in the form of constant neural network weight to form a training gait pattern library;
D. collecting knee joint gait feature data of a meniscus injury patient, extracting gait feature variables, and forming a test set by the collected knee joint gait feature data of the meniscus injury patient;
E. a group of dynamic estimators are constructed by utilizing a constant neural network, nonlinear gait system dynamics knowledge corresponding to meniscus injury patients and healthy normal people in a training gait pattern library is embedded into the dynamic estimators, the gait feature data of the meniscus injury patients are differentiated from the group of dynamic estimators to form a group of classification error results, and abnormal gaits of the meniscus injury patients are calculated according to minimum errors.
Further as a preferred embodiment, the gait feature data of the knee joint is collected by an optical sensor in the steps a and D.
Further as a preferred embodiment, the gait feature data of the knee joint in the step a and the step b includes knee joint angle feature data and knee joint displacement feature data, the knee joint angle feature data includes an internal rotation angle, an external rotation angle, a flexion-extension angle and an internal inversion angle of the knee joint femur relative to the tibia, and the knee joint displacement feature data includes anterior-posterior displacement feature data of the knee joint femur relative to the tibia.
In the embodiment of the invention, an infrared light navigation Knee joint in-vivo detection system Opti _ Knee can be adopted to acquire gait feature data, referring to fig. 2, wherein the unit of the joint angle is radian, and the unit of the displacement is millimeter, so as to form a group of Knee joint angle and displacement feature variables: x is [ inward-outward rotation angle, flexion-extension angle, inward-outward turning angle, front-back displacement]TTo reduce feature dimensions and computational complexity. In the test process of the specific embodiment of the invention, a total of 22 meniscus injury patients comprise 11 males and 11 females, the ages of which are between 11 and 76 years old, 11 of the patients are randomly selected as training set data acquisition objects, and the remaining 11 patients are used as test set data acquisition objects to be detected; and 28 healthy persons with normal gait, including 14 males and 14 females, with the age ranging from 20 to 30 years, wherein 14 persons are randomly selected as training set data collection objects. As shown in the figureFig. 2-9 show the difference between the meniscus injury patient and the healthy normal patient in the internal and external rotation angle, flexion and extension angle, internal and external rotation angle of the knee joint femur relative to the tibia, and the anterior and posterior displacement characteristics of the knee joint femur relative to the tibia.
Further as a preferred embodiment, the unknown nonlinear gait system dynamic modeling in the step B is represented as:
wherein x is [ x ]1，…，xn]T∈RnB, extracting gait characteristic variables from the gait characteristic variables extracted in the step A, wherein p is a system constant parameter value, and n is the dimension of the gait characteristic variables; f (x; p) ═ F1(x；p)，…，fn(x；p)]TIs a smooth and unknown nonlinear dynamic variable representing the gait system dynamics of meniscus injured patients and healthy normal people, and v (x; p) ═ v1(x；p)，…，vn(x；p)]TFor modeling uncertainty terms, the modeling uncertainty term v (x; p) and the gait system dynamics F (x; p) are combined intoDefined as the general nonlinear gait system dynamics.
Further as a preferred embodiment, in the step B, a dynamic RBF neural network is used to construct a neural network identifier, which is used to identify general nonlinear gait system dynamics:
a schematic diagram of a neural network topology for learning the dynamics of a nonlinear gait system is shown in fig. 4.
Further as a preferred embodiment, the dynamic RBF neural network identifier is in the form of:
whereinIs the state of the neural network identifier; a ═ diag [ a ═ d1，…，an]Is a diagonal matrix, aiSatisfying 0 < | a as a constant of designi|＜1，Is a dynamic RBF neural network for approximating unknown general non-linear gait system dynamicsS(x)＝[S1(||X-ξ1||，…，SN(||X-ξn||]TIs a Gaussian radial basis function, N > 1 is the number of neural network nodes, ζiIs the central point of neuron, and the neuron is uniformly distributed in the region [ -1, 1 [)]×[-1，1]×[-1，1]×[-1，1]Within, and the width is 0.15; normalizing all gait feature data to [ -1, 1 [ -1 [ ]]An interval; RBF neural network weightThe regulation law of (2) is as follows:
wherein i represents the ith dimension variable in the n dimension gait characteristic variables,in order to be a state error,σigreater than 0 is the regulating parameter of the rhythm regulationIn the examplesi＝35，σi0.4, weight of dynamic RBF neural networkInitial value of (2)
Constant value neural networkIs time-invariant and spatially distributed, i.e. the effective information is stored only in neurons close to the intrinsic system dynamics trajectories of the gait feature data, whereas neurons far from the trajectories have no information stored, constant neural networksOnly the internal dynamic along the gait characteristic data space track is approached, and the internal dynamic far away from the track is not approached; therefore, according to the determined learning theory, the neurons of the RBF neural network along the characteristic track of the gait system meet the continuous excitation condition, the weight value of the RBF neural network converges to the optimal value, the average value of the weight value within a period of time after the weight value converges is taken as the learning training result, and the result is utilized to establish the constant neural networkThe learned gait system dynamics knowledge is stored in the form of constant neural network weight to form a training gait mode library; the constant value neural network weightCharacterized by the following formula:wherein, [ t ]a，tb]Representing a time period after the constant neural network weights have completed the transition towards their optimum values, such thatCan be composed of constant value neural networkAnd carrying out local accurate approximation.
Further as a preferred embodiment, the locally accurate modeling of the general nonlinear gait system dynamics may be formulated as follows:
wherein,i1is an approximation error; the local accurate modeling refers to the approximation of the internal system dynamic track along the gait feature data through the RBF neural network, and the internal dynamic far away from the track is not approximated.
Further as a preferred embodiment, the step C specifically includes: according to the determined learning theory, the neurons of the RBF neural network along the characteristic track of the gait system meet the continuous excitation condition, the weight of the RBF neural network converges to an optimal value, the average value of the weight within a period of time after the weight converges is taken as the learning training result, the constant neural network is established by using the result, and the learned gait system dynamics knowledge is stored in the form of the constant neural network weight to form a training gait pattern library.
Further as a preferred embodiment, the RBF neural network weight is adjusted according to lyapuloff's stability theorem and deterministic learning theory, so that both the state error and the weight estimation are bounded, and the state error and the weight estimation are exponentially converged, wherein there are two cases for the weight convergence of the RBF neural network:
in the first case: the neurons of the RBF neural network along the gait characteristic regression trajectory meet continuous excitation conditions, and the weight of the neurons converges into a small neighborhood of an optimal value;
in the second case: neurons of the RBF neural network far away from the gait feature regression trajectory are not excited and are not adjusted, and the weight value is approximate to zero.
For example, the weight value converges to a constant value (optimal value) in a period of time, the convergence of the neural network weight value in the learning stage is as shown in fig. 5, and the weight value of the neuron close to the system trajectory satisfies a part of continuous excitation conditions, thereby converging to the optimal value; whereas neurons far from the system trajectory are stimulated to a small extent and are hardly modulated, remaining substantially in a small neighborhood of zero.
For 11 persons serving as test set data acquisition objects to be detected in the embodiment of the invention, knee joint gait feature data of a meniscus injury patient are acquired, gait feature variables are extracted, and the acquired knee joint gait feature data of the meniscus injury patient form a test set; then using a constant neural networkBuilding a group of dynamic estimators, embedding the dynamic knowledge of the nonlinear gait systems corresponding to the meniscus injury patients and healthy normal persons in a training gait mode library into the dynamic estimators, subtracting the gait feature data of the meniscus injury patients from the group of dynamic estimators to form a group of classification error results, and calculating the abnormal gait of the meniscus injury patients according to the minimum error. The specific implementation of the above operation method is as follows:
firstly, identifying results, namely constant neural network weight, of RBF neural network according to the dynamic state of a general nonlinear gait system of meniscus injury patients and healthy normal persons in a training gait pattern libraryConstructing a group of dynamic estimators, and embedding the gait system dynamics knowledge of the meniscus damage patients and healthy normal persons learned in the step B and the step C into the dynamic estimators, wherein the gait system dynamics knowledge is expressed as follows:
wherein,as the state of the dynamic estimator, biFor dynamic estimator parameters, take b in this embodimenti＝-55，xtiThe gait feature data of the meniscal injury patient to be detected in a centralized mode are tested, k represents the kth training mode in M training modes, M is the mode total amount in a training gait mode library, the gait feature data sequences extracted in each walking process of the meniscal injury patient and a healthy normal person form a mode, and the test object walks for how many times in the test process, so that the corresponding extracted gait feature data sequences form how many modes;
secondly, tests are concentrated on gait feature data x of meniscus injury patientstiAnd performing subtraction with the group of dynamic estimators to obtain a classification detection error system as follows:
wherein,is the state estimation error, calculatingAverage L of1The norm is as follows:
wherein, TcRepresenting the gait cycle;
finally, if the gait pattern of the test concentration meniscal injury patient is similar to the training gait pattern s (s ∈ {1, …, k }), dynamic embedding is performedConstant RBF neural network in estimator sThe learned knowledge can be quickly recalled and an accurate approximation to gait dynamics is provided; thus, the corresponding errorIn all errorsThe detection method is characterized in that the detection method comprises the following steps of (1) minimizing the defect, and based on the principle of minimum error, the abnormal gait of the meniscus injury patient can be rapidly classified and detected:
if there is a finite time tsS ∈ {1, …, k } and some i ∈ {1, …, n }, such thatFor all t > tsIf the detection result is positive, the abnormal gait pattern of the meniscus damage patient can be detected in a classified mode, and therefore auxiliary detection of meniscus damage is achieved. The classification detection result is evaluated by using performance indexes such as Sensitivity (Sensitivity), Specificity (Specificity) and Accuracy (Accuracy), and the indexes are calculated as follows:
where TP represents true positive samples, TN represents true negative samples, FP represents false positive samples, and FN represents false negative samples.
In the embodiment of the invention, TP is 9, TN is 13, FN is 2, and FP is 1.
The following table is a classification test result table of meniscus injury patients and healthy people:
| performance index | Results (%) |
| Sensitivity | 81.82 |
| Specificity | 92.86 |
| Accuracy | 88 |
While the invention has been described with reference to a preferred embodiment, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (9)
1. A gait analysis method for assisting in screening meniscus injuries is characterized in that: the method comprises the following steps:
A. respectively collecting knee joint gait feature data of a plurality of groups of meniscus injury patients and healthy normal people, extracting gait feature variables, and forming training sets by the collected knee joint gait feature data of the plurality of groups of meniscus injury patients and healthy normal people;
B. modeling unknown nonlinear gait system dynamics of meniscus injury patients and healthy normal persons in a training set according to the gait feature variables extracted in the step 1, and approximating the unknown nonlinear gait system dynamics locally by adopting a neural network identifier;
C. establishing a constant neural network by using the learning and training result of the RBF neural network identifier, and storing the learned nonlinear gait system dynamics knowledge in the form of constant neural network weight to form a training gait pattern library;
D. collecting knee joint gait feature data of a meniscus injury patient, extracting gait feature variables, and forming a test set by the collected knee joint gait feature data of the meniscus injury patient;
E. a group of dynamic estimators are constructed by utilizing a constant neural network, nonlinear gait system dynamics knowledge corresponding to meniscus injury patients and healthy normal people in a training gait pattern library is embedded into the dynamic estimators, the gait feature data of the meniscus injury patients are differentiated from the group of dynamic estimators to form a group of classification error results, and abnormal gaits of the meniscus injury patients are calculated according to minimum errors.
2. A gait analysis method for assisting in screening meniscal injuries according to claim 1, wherein: and B, collecting gait characteristic data of the knee joint through an optical sensor in the step A and the step D.
3. A gait analysis method for assisting in screening meniscal injuries according to claim 1, wherein: and step A and step B, the gait characteristic data of the knee joint comprise knee joint angle characteristic data and knee joint displacement characteristic data, the knee joint angle characteristic data comprise an internal rotation angle, an external rotation angle and an internal and external rotation angle of the knee joint femur relative to the tibia, and the knee joint displacement characteristic data comprise front and back displacement characteristic data of the knee joint femur relative to the tibia.
4. A gait analysis method for assisting in screening meniscal injuries according to claim 1, wherein: the unknown nonlinear gait system dynamic modeling in the step B is expressed as:
wherein x is [ x ]1，…，xn]T∈RnB, extracting gait characteristic variables from the gait characteristic variables extracted in the step A, wherein p is a system constant parameter value, and n is the dimension of the gait characteristic variables; f (x; p) ═ F1(x；p)，…，fn(x；p)]TIs a smooth and unknown nonlinear dynamic variable representing the gait system dynamics of meniscus injured patients and healthy normal people, and v (x; p) ═ v1(x；p)，…，vn(x；p)]TFor modeling uncertainty, the two are merged intoDefined as the general nonlinear gait system dynamics.
5. The gait analysis method of assisting in screening for meniscus damage according to claim 4, wherein: and B, constructing a neural network identifier by adopting a dynamic RBF neural network.
6. A gait analysis method for assisting in screening meniscal injuries according to claim 5, wherein: the dynamic RBF neural network identifier is in the form of:
whereinIs the state of the neural network identifier; a ═ diag [ a ═ d1，…，an]Is a diagonal matrix, aiSatisfying 0 < | a as a constant of designi|＜1，Is a dynamic RBF neural network for approximating unknown general non-linear gait system dynamicsS(x)＝[S1(||X-ξ1||，…，SN(||X-ξn||]TIs a Gaussian radial basis function, N > 1 is the number of neural network nodes, ξiIs the central point of the neuron, RBF neural network weightThe regulation law of (2) is as follows:
wherein i represents the ith dimension variable in the n dimension gait characteristic variables,is the error in the state of the device,i＝i T＞0，σimore than 0 is the adjusting parameter of the rhythm regulation, and the weight of the dynamic RBF neural networkInitial value of (2)
7. A gait analysis method for assisting in screening meniscal injuries according to claim 6, wherein: the locally accurate modeling of the general nonlinear gait system dynamics can be formulated as follows:
wherein,i1is an approximation error; the local accurate modeling refers to the approximation of the internal system dynamic track along the gait feature data through the RBF neural network, and the internal dynamic far away from the track is not approximated.
8. A gait analysis method for assisting in screening meniscal injuries according to claim 1, wherein: the step C is specifically as follows: according to the determined learning theory, the neurons of the RBF neural network along the characteristic track of the gait system meet the continuous excitation condition, the weight of the RBF neural network converges to an optimal value, the average value of the weight within a period of time after the weight converges is taken as the learning training result, the constant neural network is established by using the result, and the learned gait system dynamics knowledge is stored in the form of the constant neural network weight to form a training gait pattern library.
9. A gait analysis method for assisting in screening meniscal injuries according to claim 8, wherein: the RBF neural network weight is adjusted according to the Lyapunov stability theorem and the definite learning theory, so that both the state error and the weight estimation are bounded, and the index is converged, wherein the weight convergence of the RBF neural network has two conditions:
in the first case: the neurons of the RBF neural network along the gait characteristic regression trajectory meet continuous excitation conditions, and the weight of the neurons converges into a small neighborhood of an optimal value;
in the second case: neurons of the RBF neural network far away from the gait feature regression trajectory are not excited and are not adjusted, and the weight value is approximate to zero.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610363760.8A CN106650195A (en) | 2016-05-26 | 2016-05-26 | Gait analysis method for assisting in screening meniscus injuries |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610363760.8A CN106650195A (en) | 2016-05-26 | 2016-05-26 | Gait analysis method for assisting in screening meniscus injuries |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106650195A true CN106650195A (en) | 2017-05-10 |
Family
ID=58848706
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610363760.8A Pending CN106650195A (en) | 2016-05-26 | 2016-05-26 | Gait analysis method for assisting in screening meniscus injuries |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106650195A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108305683A (en) * | 2018-02-09 | 2018-07-20 | 马立敏 | A kind of diseases of knee joint prediction meanss and system |
| CN108538388A (en) * | 2018-03-06 | 2018-09-14 | 上海逸动医学科技有限公司 | A kind of motion of knee joint obstacle function evaluation method |
| CN109157286A (en) * | 2018-10-25 | 2019-01-08 | 北京爱康宜诚医疗器材有限公司 | Data predication method and device |
| CN109498025A (en) * | 2018-09-13 | 2019-03-22 | 龙岩学院 | Knee osteoarthritis diagnostic system based on phase space reconfiguration, Euclidean distance and neural network |
| JP2021033622A (en) * | 2019-08-23 | 2021-03-01 | 富士通株式会社 | Obstacle determination method, obstacle determination program and information processing apparatus |
| CN113706491A (en) * | 2021-08-20 | 2021-11-26 | 西安电子科技大学 | Meniscus injury grading method based on mixed attention weak supervision transfer learning |
| CN114641234A (en) * | 2019-12-06 | 2022-06-17 | 波音公司 | Model of cumulative damage to soft tissue material to reduce repetitive stress damage in performing a treatment |
| CN114743664A (en) * | 2022-03-04 | 2022-07-12 | 山东大学 | Gait analysis and determination learning-based spinal cervical spondylosis auxiliary diagnosis system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103519819A (en) * | 2012-07-06 | 2014-01-22 | 王振兴 | Gait analysis method and gait analysis system |
| CN104091177A (en) * | 2014-06-30 | 2014-10-08 | 华南理工大学 | Abnormal gait detection method based on determined learning theory |
| CN104834888A (en) * | 2014-12-04 | 2015-08-12 | 龙岩学院 | Abnormal gait identification method capable of facilitating screening Parkinsonism |
| CN105286875A (en) * | 2015-11-24 | 2016-02-03 | 龙岩学院 | Gait analysis method for auxiliary screening of injury of anterior cruciate ligament |
| CN105468908A (en) * | 2015-11-24 | 2016-04-06 | 龙岩学院 | Gait analysis method capable of carrying out auxiliary screening on knee osteoarthritis |
| US20160095538A1 (en) * | 2014-10-07 | 2016-04-07 | Samsung Electronics Co., Ltd. | Method and apparatus for recognizing gait motion |
-
2016
- 2016-05-26 CN CN201610363760.8A patent/CN106650195A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103519819A (en) * | 2012-07-06 | 2014-01-22 | 王振兴 | Gait analysis method and gait analysis system |
| CN104091177A (en) * | 2014-06-30 | 2014-10-08 | 华南理工大学 | Abnormal gait detection method based on determined learning theory |
| US20160095538A1 (en) * | 2014-10-07 | 2016-04-07 | Samsung Electronics Co., Ltd. | Method and apparatus for recognizing gait motion |
| CN104834888A (en) * | 2014-12-04 | 2015-08-12 | 龙岩学院 | Abnormal gait identification method capable of facilitating screening Parkinsonism |
| CN105286875A (en) * | 2015-11-24 | 2016-02-03 | 龙岩学院 | Gait analysis method for auxiliary screening of injury of anterior cruciate ligament |
| CN105468908A (en) * | 2015-11-24 | 2016-04-06 | 龙岩学院 | Gait analysis method capable of carrying out auxiliary screening on knee osteoarthritis |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108305683A (en) * | 2018-02-09 | 2018-07-20 | 马立敏 | A kind of diseases of knee joint prediction meanss and system |
| CN108538388A (en) * | 2018-03-06 | 2018-09-14 | 上海逸动医学科技有限公司 | A kind of motion of knee joint obstacle function evaluation method |
| CN108538388B (en) * | 2018-03-06 | 2022-09-06 | 上海逸动医学科技有限公司 | Knee joint dyskinesia function judging method |
| CN109498025A (en) * | 2018-09-13 | 2019-03-22 | 龙岩学院 | Knee osteoarthritis diagnostic system based on phase space reconfiguration, Euclidean distance and neural network |
| CN109157286A (en) * | 2018-10-25 | 2019-01-08 | 北京爱康宜诚医疗器材有限公司 | Data predication method and device |
| JP2021033622A (en) * | 2019-08-23 | 2021-03-01 | 富士通株式会社 | Obstacle determination method, obstacle determination program and information processing apparatus |
| JP7358842B2 (en) | 2019-08-23 | 2023-10-11 | 富士通株式会社 | Failure determination method, failure determination program, and information processing device |
| CN114641234A (en) * | 2019-12-06 | 2022-06-17 | 波音公司 | Model of cumulative damage to soft tissue material to reduce repetitive stress damage in performing a treatment |
| CN114641234B (en) * | 2019-12-06 | 2025-03-14 | 波音公司 | Methods and systems for reducing the likelihood of repetitive stress damage to soft tissue |
| CN113706491A (en) * | 2021-08-20 | 2021-11-26 | 西安电子科技大学 | Meniscus injury grading method based on mixed attention weak supervision transfer learning |
| CN113706491B (en) * | 2021-08-20 | 2024-02-13 | 西安电子科技大学 | Meniscus injury grading method based on mixed attention weak supervision migration learning |
| CN114743664A (en) * | 2022-03-04 | 2022-07-12 | 山东大学 | Gait analysis and determination learning-based spinal cervical spondylosis auxiliary diagnosis system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106650195A (en) | Gait analysis method for assisting in screening meniscus injuries | |
| CN105286875A (en) | Gait analysis method for auxiliary screening of injury of anterior cruciate ligament | |
| CN105468908A (en) | Gait analysis method capable of carrying out auxiliary screening on knee osteoarthritis | |
| CN109498025A (en) | Knee osteoarthritis diagnostic system based on phase space reconfiguration, Euclidean distance and neural network | |
| Hatamzadeh et al. | A new method of diagnosing athlete's anterior cruciate ligament health status using surface electromyography and deep convolutional neural network | |
| Winiarski et al. | A new method of evaluating the symmetry of movement used to assess the gait of patients after unilateral total hip replacement | |
| Wang et al. | Efficient subject-independent detection of anterior cruciate ligament deficiency based on marine predator algorithm and support vector machine | |
| Glaros et al. | A wearable intelligent system for monitoring health condition and rehabilitation of running athletes | |
| CN110097967A (en) | Knee osteoarthritis diagnostic system based on intrinsic time Scale Decomposition, phase space reconfiguration and neural network | |
| CN119400344A (en) | Construction method and system based on anorectal surgery postoperative nursing rehabilitation evaluation model | |
| Rahman et al. | Muscle forces and power are significantly reduced during walking in patients with peripheral artery disease | |
| Zeng et al. | Classification of gait patterns in patients with unilateral anterior cruciate ligament deficiency based on phase space reconstruction, Euclidean distance and neural networks | |
| Salim et al. | An automated method for analysis of gait data to aid clinical interpretation | |
| Cedeno-Moreno et al. | A stacked autoencoder-based aid system for severity degree classification of knee ligament rupture | |
| CN118078312A (en) | A muscle function assessment system for residual limbs of forearm amputees based on electromyographic signals | |
| Senanayake et al. | Artificial neural network based gait patterns identification using neuromuscular signals and soft tissue deformation analysis of lower limbs muscles | |
| Byfield et al. | Machine learning full 3-D lower-body kinematics and kinetics on patients with osteoarthritis from electromyography | |
| Sartori et al. | Electromyography‐driven modeling for simulating subject‐specific movement at the neuromusculoskeletal level | |
| Mezghani et al. | Biomechanical signal classification of surgical and non-surgical candidates for knee arthroplasty | |
| Koktas et al. | Ensemble classifiers for medical diagnosis of knee osteoarthritis using gait data | |
| Patil et al. | An osteoarthritis classifier using back propagation neural network | |
| Agarwal et al. | Precision measurements in healthcare systems and devices | |
| Samadi et al. | Development of Machine learning algorithms to identify the Cobb angle in adolescents with idiopathic scoliosis based on lumbosacral joint efforts during gait (Case study) | |
| Pauk et al. | Parametric identification of lower limbs during walking of a man | |
| CN115736955B (en) | Evaluation system of disuse muscle atrophy after bone and joint injury based on surface electromyography |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170510 |
|
| WD01 | Invention patent application deemed withdrawn after publication |