Intelligent analysis system and method for lumbar intervertebral disc protrusion rehabilitation treatment

Abstract

The invention discloses an intelligent analysis system and method for rehabilitation treatment of lumbar intervertebral disc herniation, and relates to the field of intelligent analysis. Abstract clinical information amounts such as lesion part features, pain perception description and movement function requirements are converted into high-dimensional vectors by adopting a semantic coding technology based on deep learning; the system can more accurately understand and simulate the condition of a patient in a multi-dimensional space, further, a dynamic mapping mechanism between a rehabilitation starting point and a target is established to capture non-linear correlation between the current state of the patient and an ideal rehabilitation target, and intelligent derivation of a rehabilitation path is achieved based on the non-linear correlation. Through the dynamic analysis of the multi-dimensional information of the patient, the system can better understand the unique demand of each patient, the matching precision of the rehabilitation treatment scheme is improved, and the treatment is more personalized and effective.

Description

Technical Field

[0001] The present application relates to the field of intelligent analysis, and more specifically, to an intelligent analysis system and method for rehabilitation treatment of lumbar disc herniation. Background Art

[0002] In the field of rehabilitation treatment for lumbar disc herniation, the formulation of traditional rehabilitation programs mainly relies on the doctor's clinical experience and static evaluation indicators, and lacks the ability to dynamically analyze and personalize the patient's multi-dimensional information. Existing technologies usually use standardized scales or imaging features to conduct single-dimensional evaluations, which makes it difficult to effectively integrate unstructured text descriptions (such as lesion site characteristics, individualized expressions of pain perception) and structured data (such as age, medical history, etc.) in patient medical documents, resulting in insufficient matching accuracy between rehabilitation goals and treatment methods.

[0003] In addition, due to the large differences in the specific conditions of each patient, such as the location of the lesion, degree of pain, motor function recovery needs, and personal basic information, it is difficult to meet personalized needs using a unified treatment plan, resulting in uneven treatment effects.

[0004] Therefore, an optimized intelligent analysis method for the rehabilitation treatment of lumbar disc herniation is expected. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. This application provides an intelligent analysis system and method for the rehabilitation treatment of lumbar disc herniation. By adopting semantic coding technology based on deep learning, abstract clinical information such as lesion site characteristics, pain perception description, and motor function requirements are converted into high-dimensional vectors, so that the system can more accurately understand and simulate the patient's condition in a multi-dimensional space. Furthermore, a dynamic mapping mechanism between the rehabilitation starting point and the target is established to capture the nonlinear relationship between the patient's current state and the ideal rehabilitation goal, and based on this, the intelligent derivation of the rehabilitation path is realized. Through dynamic analysis of the patient's multi-dimensional information, the system can better understand the unique needs of each patient, improve the matching accuracy of the rehabilitation treatment plan, and thus make the treatment more personalized and effective.

[0006] According to one aspect of the present application, there is provided an intelligent analysis method for rehabilitation treatment of lumbar disc herniation, comprising: Obtain the medical condition file of the target patient, which includes the current condition and the motor function recovery target; Extracting semantic information from the medical condition document to obtain a set of semantic coding vectors of motor function recovery requirements and semantic coding vectors of the current medical condition; The semantic coding vector of the motor function recovery demand is used as the rehabilitation target vector and the set of the semantic coding vectors of the current condition is used as the set of rehabilitation starting point vectors, which are input into the rehabilitation starting point-target semantic analysis network to obtain the rehabilitation starting point-target semantic dynamic response coding vector, including: extracting the deep implicit coding features of the rehabilitation target vector and the set of rehabilitation starting point vectors respectively, and constructing the local feature response matrix between the two based on the dynamic semantic matching mechanism to obtain the set of rehabilitation starting point-target local semantic feature response anchor coding matrices; realizing multi-matrix weighted fusion through an adaptive weight allocation mechanism to generate the rehabilitation starting point-target semantic dynamic response coding vector; Generate a personalized rehabilitation plan based on the rehabilitation starting point-goal semantic dynamic response encoding vector.

[0007] Furthermore, the current condition section includes basic information about the patient, descriptions of the site of the lesion, and descriptions of the degree of pain; the motor function recovery demand target section includes descriptions of the motor function recovery demand.

[0008] Furthermore, the semantic information of the medical condition document is extracted to obtain a set of semantic coding vectors of motor function recovery requirements and semantic coding vectors of the current medical condition, including: The descriptions related to the lesion site, the descriptions related to the pain level, the descriptions related to the motor function recovery requirements and the basic information of the patient are semantically embedded and encoded to obtain the semantic encoding vectors of the lesion site description, the pain level description, the motor function recovery requirements and the basic information of the patient; The semantic coding vector describing the lesion site, the semantic coding vector describing the pain level and the semantic coding vector of the patient's basic information are merged to obtain a set of semantic coding vectors for the current condition.

[0009] Furthermore, the patient's basic information includes age, gender, weight and medical history.

[0010] Furthermore, the deep implicit coding features of the sets of rehabilitation target vectors and rehabilitation starting point vectors are extracted respectively, and a local feature response matrix between the two is constructed based on a dynamic semantic matching mechanism to obtain a set of rehabilitation starting point-target local semantic feature response anchor coding matrices, including: Performing deep latent feature extraction based on fully connected coding on the rehabilitation target vector to obtain a deep latent coding vector of the rehabilitation target feature; Performing deep implicit feature extraction based on fully connected coding on each rehabilitation starting point vector in the set of rehabilitation starting point vectors to obtain a set of rehabilitation starting point feature deep implicit coding vectors; Rehabilitation starting point-target semantic dynamic response encoding is performed on each rehabilitation starting point feature deep implicit coding vector in the set of rehabilitation target feature deep implicit coding vector and rehabilitation starting point feature deep implicit coding vector to obtain a set of rehabilitation starting point-target local semantic feature response anchor coding matrices.

[0011] Furthermore, the weighted fusion of multiple matrices is realized through an adaptive weight allocation mechanism to generate a rehabilitation starting point-target semantic dynamic response encoding vector, including: Calculating the decision anchor adaptive splicing weight factor of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing weight factors; Based on a set of rehabilitation starting point-target decision anchor adaptive splicing weight factors, a set of rehabilitation starting point-target local semantic feature response anchor encoding matrices is fused to obtain a rehabilitation starting point-target semantic dynamic response encoding vector.

[0012] Further, the decision anchor adaptive splicing weight factor of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices is calculated to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing weight factors, including: Based on the feature distribution of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices, determining the rehabilitation starting point-target decision anchor adaptive splicing factor of each rehabilitation starting point-target local semantic feature response anchor coding matrix to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing factors; The set of rehabilitation starting point-target decision anchor adaptive splicing factors is weighted based on the Softmax function to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing weight factors.

[0013] Further, based on the feature distribution of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices, the rehabilitation starting point-target decision anchor adaptive splicing factor of each rehabilitation starting point-target local semantic feature response anchor coding matrix is ​​determined to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing factors, including: Based on the eigenvalues ​​of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices, a state transition relationship from the mean to the maximum value is constructed through the drift coefficient, and the intermediate transition weights are globally controlled with the eigenvalue as the important score to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing factors.

[0014] Furthermore, based on the rehabilitation starting point-target semantic dynamic response encoding vector, a personalized rehabilitation plan is generated, including: The rehabilitation starting point-target semantic dynamic response encoding vector is input into the rehabilitation plan generator based on the big model to obtain a personalized rehabilitation plan.

[0015] According to another aspect of the present application, there is provided an intelligent analysis system for rehabilitation treatment of lumbar disc herniation, comprising: A target patient object medical condition document acquisition module is used to acquire the target patient object's medical condition document, which includes a medical condition status part and a motor function recovery demand target part; A semantic information extraction module is used to extract the semantic information of the medical condition document to obtain a set of semantic coding vectors of motor function recovery requirements and semantic coding vectors of the current medical condition; A rehabilitation starting point-target semantic analysis module is used to use the motor function recovery demand semantic coding vector as the rehabilitation target vector and the set of the disease status partial semantic coding vector as the set of rehabilitation starting point vectors, and input them into the rehabilitation starting point-target semantic analysis network to obtain the rehabilitation starting point-target semantic dynamic response coding vector; The personalized rehabilitation program generation module is used to generate a personalized rehabilitation program based on the rehabilitation starting point-target semantic dynamic response encoding vector.

[0016] Compared with the prior art, the intelligent analysis system and method for rehabilitation treatment of lumbar disc herniation provided by this application converts abstract clinical information such as lesion site characteristics, pain perception description, and motor function requirements into high-dimensional vectors by adopting semantic coding technology based on deep learning, so that the system can more accurately understand and simulate the patient's condition in multi-dimensional space. Furthermore, a dynamic mapping mechanism between rehabilitation starting point and target is established to capture the nonlinear relationship between the patient's current state and the ideal rehabilitation target, and based on this, the intelligent derivation of the rehabilitation path is realized. Through dynamic analysis of the patient's multi-dimensional information, the system can better understand the unique needs of each patient, improve the matching accuracy of the rehabilitation treatment plan, and make the treatment more personalized and effective. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1It is a flow chart of an intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to an embodiment of the present application; Figure 2 A data flow diagram of an intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to an embodiment of the present application; Figure 3 It is a flowchart of sub-step S3 of the intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to an embodiment of the present application; Figure 4 It is a flowchart of sub-step S32 of the intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to an embodiment of the present application; Figure 5 It is a block diagram of an intelligent analysis system for rehabilitation treatment of lumbar disc herniation according to an embodiment of the present application. DETAILED DESCRIPTION

[0019] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0020] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0021] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0022] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0023] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0024] In the technical solution of the present application, an intelligent analysis method for rehabilitation treatment of lumbar disc herniation is proposed. Figure 1 is a flow chart of an intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to an embodiment of the present application, Figure 2 Schematic diagram of data flow of the intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the intelligent analysis method for rehabilitation treatment of lumbar disc herniation includes the following steps: S1, obtaining the medical document of the target patient object, the medical document including a medical condition status part and a motor function recovery demand target part; S2, extracting the semantic information of the medical document to obtain a motor function recovery demand semantic coding vector and a set of semantic coding vectors of the medical condition status part; S3, taking the motor function recovery demand semantic coding vector as the rehabilitation target vector and the set of semantic coding vectors of the medical condition status part as the set of rehabilitation starting point vectors, and inputting them into the rehabilitation starting point-target semantic analysis network to obtain a rehabilitation starting point-target semantic dynamic response coding vector; S4, generating a personalized rehabilitation plan based on the rehabilitation starting point-target semantic dynamic response coding vector.

[0025] In particular, S1 obtains the medical condition file of the target patient object, and the medical condition file includes a current medical condition part and a motor function recovery demand target part. Among them, the current medical condition part includes the patient's basic information, a description of the lesion site, and a description of the pain level; the motor function recovery demand target part is a description of the motor function recovery demand; among them, the patient's basic information includes age, gender, weight and medical history. It should be understood that each patient's medical record document is unique, which contains rich unstructured text descriptions, such as lesion site characteristics, individualized expression of pain perception, etc., as well as structured data, such as age, gender, weight and medical history. In the technical solution of the present application, by obtaining the medical documents of the target patient and extracting the basic information of the patient, descriptions related to the lesion site, descriptions related to the pain level, and descriptions related to the motor function recovery needs from the medical documents of the target patient, the system can understand the patient's health status and specific needs more comprehensively and accurately; in addition, through in-depth analysis of the medical documents, the specific situation of each patient can be accurately captured, including special needs and restrictions. In this way, the system can dynamically adjust the rehabilitation strategy according to the patient's actual situation, which not only helps to improve the accuracy of diagnosis, but also provides a basis for the formulation of personalized treatment plans.

[0026] In particular, the S2 extracts the semantic information of the medical condition document to obtain a set of semantic coding vectors of motor function recovery requirements and semantic coding vectors of the current condition. In a specific example of the present application, first, the description of the lesion site, the description of the pain level, the description of the motor function recovery requirements and the basic information of the patient are semantically embedded and encoded to obtain the semantic coding vector of the lesion site description, the semantic coding vector of the pain level description, the semantic coding vector of the motor function recovery requirements and the semantic coding vector of the patient's basic information; considering that the patient's medical condition document contains a large amount of mixed information of unstructured medical descriptions (such as "L4-L5 segment annulus fibrosus tear with lateral posterior nucleus pulposus protrusion") and structured physiological indicators (such as body mass index, medical history), the traditional method uses manual annotation or simple keyword matching to extract features, which is difficult to accurately capture the complex semantics such as anatomical location and pathological evolution stage implied in the description of the condition. In the technical solution of the present application, descriptions related to the lesion site, descriptions related to the pain level, descriptions related to the motor function recovery requirements, and basic patient information are input into a semantic encoder based on the Bert model to obtain semantic encoding vectors for descriptions of the lesion site, semantic encoding vectors for descriptions of the pain level, semantic encoding vectors for motor function recovery requirements, and semantic encoding vectors for basic patient information. Among them, Bert is a natural language processing model based on deep learning, and its core design concept is to achieve a deep understanding of text semantics through bidirectional context modeling and large-scale pre-training. In this process, the medical semantic understanding ability obtained by the Bert model through pre-training can automatically parse the anatomical hierarchical relationship in the description of the lesion site, and capture the association pattern between subjective feelings and objective indicators in the description of the pain level; for abstract goals such as motor function recovery requirements (such as "restoring the ability to bend down to pick up objects"), the model extracts the implicit correspondence between the action function requirements and the biomechanical constraints of the spine through the contextual attention mechanism; the basic patient information includes age, gender, weight, and medical history, which is integrated into individual physiological characteristics through semantic encoding to avoid the problem of feature discretization caused by traditional numerical data processing. In this way, the system can not only automatically extract key information from unstructured descriptions of the condition, but also conduct a comprehensive analysis in combination with the patient's personal information, significantly improving the system's ability to parse the patient's condition and providing a solid foundation for more scientific and personalized rehabilitation treatment. Then, the semantic coding vectors describing the lesion site, the semantic coding vectors describing the pain level, and the semantic coding vectors of the patient's basic information are combined to obtain a set of semantic coding vectors for the current condition.

[0027] In particular, the S3 uses the semantic coding vector of the motor function recovery demand as the rehabilitation target vector and the set of the semantic coding vectors of the current condition as the set of rehabilitation starting point vectors, which are input into the rehabilitation starting point-target semantic analysis network to obtain the rehabilitation starting point-target semantic dynamic response coding vector. It should be understood that there is a complex nonlinear mapping relationship between the patient's current pathological state (such as L5-S1 segment protrusion with sciatica) and the desired rehabilitation goal (such as restoring lumbar flexion range of motion). The traditional method uses expert experience rules or linear regression models for path planning, which is difficult to effectively model the dynamic coupling effect between the anatomical characteristics of the lesion site, the change of pain threshold, the constraints of physiological characteristics and the demand for motor function recovery. Therefore, in the technical solution of the present application, the semantic coding vector of the motor function recovery demand is used as the rehabilitation target vector and the semantic coding vector of the lesion site description, the semantic coding vector of the pain level description and the semantic coding vector of the patient's basic information are used as the set of rehabilitation starting point vectors, which are input into the rehabilitation starting point-target semantic analysis network to obtain the rehabilitation starting point-target semantic dynamic response coding vector. Specifically, the network extracts the deep implicit features of the rehabilitation target vector (i.e., the need for motor function recovery) and the rehabilitation starting point vector (including the lesion site, pain level, and personal information), and constructs a local feature response matrix between the two based on a dynamic semantic matching mechanism, thereby generating a set of rehabilitation starting point-target local semantic feature response anchor coding matrices. Then, the weighted fusion of multiple matrices is realized through an adaptive weight allocation mechanism, and finally a rehabilitation starting point-target semantic dynamic response coding vector is generated. That is, a dynamic response model with medical interpretability is constructed by deeply mining the potential association pattern between the starting point vector set and the target vector. In this way, not only the semantic fragmentation problem caused by the discrete distribution of multi-source feature vectors in the vector space is solved, but also a dynamic evolution path from the current state to the target state is established, overcoming the defect that the static scheme cannot adapt to the stage changes such as pain relief and muscle function reconstruction in the rehabilitation process; in addition, through the dynamic response coding mechanism, the influence weight of different rehabilitation starting points on the achievement of the goal is quantified, avoiding the homogenization of key decision factors in traditional methods. As a result, the system can identify and emphasize those factors that are most important to specific patients or at a specific rehabilitation stage, making the generated rehabilitation scheme more in line with actual needs. In particular, in a specific example of the present application, Figure 3 As shown, the S3 includes: S31, respectively extracting the deep implicit coding features of the set of rehabilitation target vectors and rehabilitation starting point vectors, and constructing a local feature response matrix between the two based on a dynamic semantic matching mechanism to obtain a set of rehabilitation starting point-target local semantic feature response anchor coding matrices; S32, realizing multi-matrix weighted fusion through an adaptive weight allocation mechanism to generate a rehabilitation starting point-target semantic dynamic response coding vector.

[0028] Specifically, the S31 extracts the deep implicit coding features of the set of rehabilitation target vectors and rehabilitation starting point vectors respectively, and constructs a local feature response matrix between the two based on a dynamic semantic matching mechanism to obtain a set of rehabilitation starting point-target local semantic feature response anchor coding matrices. In an embodiment of the present application, first, the rehabilitation target vector is subjected to deep implicit feature extraction based on fully connected coding to obtain a deep implicit coding vector of rehabilitation target features; that is, by taking the semantic coding vector of motor function recovery requirements as the rehabilitation target vector and performing deep implicit feature extraction on it, it can be converted into a higher dimensional representation with rich semantic information. In this way, the system can understand the patient's rehabilitation needs at a higher dimension, thereby providing more accurate and personalized suggestions. In a specific example, the rehabilitation target vector is subjected to deep implicit feature extraction based on fully connected coding to obtain a deep implicit coding vector of rehabilitation target features using the following feature extraction formula; wherein the feature extraction formula is: ; in, represents the rehabilitation target vector, is matrix multiplication, and are the rehabilitation target weight matrix and the rehabilitation target bias vector, respectively. is the activation function, Represents the deep latent encoding vector of the rehabilitation target features.

[0029] Next, deep implicit feature extraction based on fully connected coding is performed on each rehabilitation starting point vector in the set of rehabilitation starting point vectors to obtain a set of deep implicit coding vectors of rehabilitation starting point features; by using a fully connected neural network to extract deep implicit features from the rehabilitation starting point vectors (i.e., semantic coding vectors describing the lesion site, semantic coding vectors describing the pain level, and semantic coding vectors describing the patient's basic information), these complex descriptions of the condition can be converted into a higher-dimensional representation with rich semantic information, thereby better reflecting the actual situation of the patient. In this way, the system's ability to understand the patient's current health status can be enhanced to ensure that the generated rehabilitation plan can accurately match the patient's needs. Specifically, by performing deep implicit feature extraction on each rehabilitation starting point vector, the system can learn a global representation of the patient's health status and capture the complex interactive relationship between the original features through a nonlinear activation function. This deep information processing method helps to reveal the potential pattern behind the patient's condition and lays the foundation for subsequent semantic alignment with the rehabilitation target vector. In a specific example, the following feature extraction formula is used to perform deep implicit feature extraction based on fully connected coding on each rehabilitation starting point vector in the set of rehabilitation starting point vectors to obtain a set of deep implicit coding vectors of rehabilitation starting point features; wherein, the feature extraction formula is: ; ; in, represents the set of rehabilitation starting point vectors, , , and Respectively represent the first, second, and third rehabilitation starting point vectors in the set and A rehabilitation starting point vector, and are the rehabilitation starting point weight matrix and the rehabilitation starting point bias vector, respectively. The first one in the set of deep implicit encoding vectors representing the starting point of rehabilitation The deep latent encoding vector of the rehabilitation starting point features.

[0030] Then, each rehabilitation starting point feature deep implicit coding vector in the set of rehabilitation target feature deep implicit coding vector and rehabilitation starting point feature deep implicit coding vector is subjected to rehabilitation starting point-target semantic dynamic response coding to obtain a set of rehabilitation starting point-target local semantic feature response anchor coding matrices. Considering that traditional rehabilitation assessment methods often rely on doctors' experience and static indicators, this method is difficult to fully consider the individual differences and complex conditions of patients. For example, different patients may show different symptoms, pain perception and recovery needs at the same stage of the disease, and these factors will affect the final rehabilitation effect. By performing rehabilitation starting point-target semantic dynamic response coding on each rehabilitation starting point feature deep implicit coding vector in the set of rehabilitation target feature deep implicit coding vector and rehabilitation starting point feature deep implicit coding vector, the system can identify the complex interactions between these features and their impact on the rehabilitation process. Specifically, the generated set of rehabilitation starting point-target local semantic feature response anchor coding matrices not only represents the interactive information between rehabilitation target and starting point features, but also serves as a bridge to capture the global and local semantic alignment relationship. By constructing a local feature response anchor coding matrix, the system can understand the patient's health status at a higher dimension and identify those details that are crucial to the rehabilitation process. In a specific example, each rehabilitation starting point feature depth implicit coding vector in the set of rehabilitation target feature depth implicit coding vector and rehabilitation starting point feature depth implicit coding vector is subjected to rehabilitation starting point-target semantic dynamic response coding to obtain a set of rehabilitation starting point-target local semantic feature response anchor coding matrices; wherein the dynamic response coding formula is: ; in, yes The transposed vector of yes Length, represents vector multiplication, Represents the first in the set of rehabilitation starting point-target local semantic feature response anchor encoding matrices A rehabilitation starting point-target local semantic feature response anchor encoding matrix.

[0031] Specifically, the S32 implements multi-matrix weighted fusion through an adaptive weight allocation mechanism to generate a rehabilitation starting point-target semantic dynamic response encoding vector. Figure 4 As shown, the S32 includes: S321, calculating the decision anchor adaptive splicing weight factor of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing weight factors; S322, fusing the set of rehabilitation starting point-target local semantic feature response anchor coding matrices based on the set of rehabilitation starting point-target decision anchor adaptive splicing weight factors to obtain a rehabilitation starting point-target semantic dynamic response coding vector.

[0032] More specifically, the S321 calculates the decision anchor adaptive splicing weight factor of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing weight factors. In the technical solution of the present application, firstly, based on the eigenvalues ​​of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices, a state transition relationship from the mean to the maximum value is constructed through the drift coefficient, and the intermediate transition weight is globally controlled with the eigenvalue as the important score to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing factors; that is, by performing feature distribution analysis on the rehabilitation starting point-target local semantic feature response anchor coding matrix, the rehabilitation starting point-target decision anchor adaptive splicing factor is determined, so that the system can identify the dominance of each local feature in the rehabilitation task, and based on this, the system can dynamically adjust the importance of the feature at the global and local levels, thereby ensuring that the generated rehabilitation plan can accurately match the actual situation of the patient. For example, if a patient's main problem is severe pain rather than limited motor function, the system will increase the weight of pain management accordingly and reduce the proportion of exercise training. In particular, it should be understood that the importance of features in different stages of rehabilitation of patients presents a non-uniform distribution characteristic. For example, the pain level feature in the acute inflammatory response stage may dominate the treatment priority, while the weight of the motor function demand feature in the chronic recovery period increases significantly. In the technical solution of the present application, based on the eigenvalues ​​of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices, the state transition relationship from the mean to the maximum value is constructed through the drift coefficient, and the intermediate transition weight is globally controlled with the eigenvalue as the important score to obtain a set of adaptive splicing factors of the rehabilitation starting point-target decision anchor. Specifically, by introducing a state transition mechanism driven by a drift coefficient, it is intended to establish a dynamic interpretable generalization path from mean statistics (weakly interpretable overall distribution) to maximum focus (strongly interpretable significant features). That is, a dynamic balance mechanism is established between the statistical characteristics of the feature distribution (the overall trend represented by the mean) and the pathological significance (the emergency state pointed to by the maximum value), simulating the comprehensive weighing process of "the most urgent problem at the moment" and "long-term risk control" by clinical experts when formulating the plan. Specifically, in this process, the feature value participates in the global control of the intermediate transition weight as an importance score. When formulating a rehabilitation plan, it is necessary to integrate the multi-dimensional information of the patient (such as the averaged background of age, medical history, etc.), and focus on strengthening the decision weight of key pathological indications (such as severe pain or local strong features such as specific intervertebral disc compression).For example, when strong semantic features such as "radiating pain" appear frequently in the patient's pain description, the system enhances the weight of the corresponding local response matrix through the eigenvalue importance score, so that the rehabilitation plan gives priority to analgesic intervention measures rather than conventional exercise training. This global control of weights based on state transition enables the system to simulate the progressive adjustment logic of the clinical rehabilitation path and achieve truly personalized dynamic program optimization while ensuring medical standardization. In a specific example, based on the feature distribution of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices, the rehabilitation starting point-target decision anchor adaptive splicing factor of each rehabilitation starting point-target local semantic feature response anchor coding matrix is ​​determined by the following decision anchor adaptive splicing factor calculation formula to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing factors; wherein, the decision anchor adaptive splicing factor calculation formula is:. ; ; ; ; in, express The variance of express The mean of express The number of eigenvalues ​​is used as the difference amplification factor, For calculation The number of eigenvalues ​​of express The maximum value of ， represents the drift coefficient, It is an intermediate state transition. yes No. eigenvalues, represents an exponential function with a natural constant as base, Represents the first in the set of rehabilitation starting point-target decision anchor adaptive splicing factors A rehabilitation starting point-target decision anchor adaptive splicing factor.

[0033] Next, the set of rehabilitation starting point-target decision anchor adaptive splicing factors is weighted based on the Softmax function to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing weight factors. It should be understood that the dynamic association between the patient's rehabilitation starting point features (such as lesion site description, pain perception, individual basic information) and the rehabilitation goal (motor function recovery requirements) has significant nonlinearity and uncertainty. The traditional fixed weight allocation method is difficult to adapt to the complexity implicit in the distribution of different patient data. If the original splicing factors are directly used for feature fusion, the contribution of each local response matrix lacks a normalized quantitative standard, resulting in the weakening of key features or the excessive amplification of redundant features. In the technical solution of the present application, the set of rehabilitation starting point-target decision anchor adaptive splicing factors is weighted based on the Softmax function to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing weight factors. By using the Softmax function, the system can convert the original decision anchor adaptive splicing factors into an easily interpretable weight distribution, so that the system can dynamically adjust the weight of each local feature according to its specific importance, so that those features that are critical to the rehabilitation process get a higher priority. For example, in a patient's rehabilitation plan, if pain management is more critical than motor function recovery, then the corresponding pain-related features will account for a larger proportion in the final rehabilitation plan. In a specific example, the set of rehabilitation starting point-target decision anchor adaptive splicing factors is weighted based on the Softmax function using the following weight calculation formula to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing weight factors; wherein the weight calculation formula is: ; in, is the normalization function, represents the first in the set of adaptive splicing weight factors of the rehabilitation starting point-target decision anchor The adaptive splicing weight factor of the rehabilitation starting point-target decision anchor.

[0034] More specifically, the S322, based on the set of adaptive splicing weight factors of the rehabilitation starting point-target decision anchor, fuses the set of local semantic feature response anchor coding matrices of the rehabilitation starting point-target to obtain the rehabilitation starting point-target semantic dynamic response coding vector. That is, by fusing each local semantic feature response anchor coding matrix according to its corresponding adaptive splicing weight factor, the system can comprehensively consider all relevant factors at the global level to form a dynamic response coding vector that fully reflects the patient's health status. This approach allows the system to flexibly adjust the rehabilitation strategy according to the specific situation of each patient. For example, if the main problem of a patient is severe pain rather than limited motor function, the system will increase the weight of pain management accordingly and reduce the proportion of exercise training, thereby generating a more personalized rehabilitation plan for the patient. In a specific example, based on the set of adaptive splicing weight factors of the rehabilitation starting point-target decision anchor, the set of local semantic feature response anchor coding matrices of the rehabilitation starting point-target is fused with the following fusion formula to obtain the rehabilitation starting point-target semantic dynamic response coding vector; wherein, the fusion formula is: ; in, is the rehabilitation starting point-target dynamic response encoding matrix, is the reshaping operation, Represents the rehabilitation starting point-target semantic dynamic response encoding vector.

[0035] In particular, the S4 generates a personalized rehabilitation plan based on the rehabilitation starting point-target semantic dynamic response coding vector. In the technical solution of the present application, the rehabilitation starting point-target semantic dynamic response coding vector is input into a rehabilitation plan generator based on a large model to obtain a personalized rehabilitation plan. It should be understood that the multidimensional feature vector (such as lesion site characteristics, pain dynamic evolution pattern, individualized rehabilitation goal) after semantic analysis and dynamic response coding is essentially an abstract representation of a high-dimensional latent space, and the mapping relationship between it and specific treatment methods (such as physical therapy intensity, exercise training type, and drug intervention strategy) has highly nonlinear and cross-modal association characteristics. Existing rule engines or simple classification models are difficult to capture this implicit association from complex semantic features to multi-dimensional rehabilitation actions. In the technical solution of this application, by inputting the rehabilitation starting point-target semantic dynamic response encoding vector into the rehabilitation program generator based on the large model, end-to-end mapping reconstruction from the semantic feature space to the clinical executable program is achieved. That is, through the pre-training understanding of massive cross-modal medical knowledge (such as clinical guidelines, rehabilitation medicine literature, and patient treatment trajectory data) by the large model, the individualized rehabilitation needs implicit in the dynamic response encoding vector (such as analgesia priority in the acute phase and functional remodeling in the chronic phase) are converted into specific treatment steps with timing logic and dosage control. In this way, the system can more deeply understand the complex interactions between the local feature response anchor encoding matrices, thereby generating more accurate rehabilitation suggestions.

[0036] In summary, the intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to the embodiment of the present application is explained, and the abstract clinical information such as lesion site characteristics, pain perception description, motor function requirements, etc. is converted into high-dimensional vectors by adopting semantic coding technology based on deep learning, so that the system can more accurately understand and simulate the patient's situation in multi-dimensional space. Furthermore, a dynamic mapping mechanism between rehabilitation starting point and target is established to capture the nonlinear relationship between the patient's current state and the ideal rehabilitation target, and based on this, the intelligent derivation of the rehabilitation path is realized. Through dynamic analysis of the patient's multi-dimensional information, the system can better understand the unique needs of each patient, improve the matching accuracy of the rehabilitation treatment plan, and thus make the treatment more personalized and effective.

[0037] Furthermore, an intelligent analysis system for rehabilitation treatment of lumbar disc herniation is also provided.

[0038] Figure 5 FIG. 1 is a block diagram of an intelligent analysis system for rehabilitation treatment of lumbar disc herniation according to an embodiment of the present application. Figure 5As shown, according to the embodiment of the present application, the intelligent analysis system 300 for rehabilitation treatment of lumbar disc herniation includes: a target patient object medical document acquisition module 310, which is used to acquire the medical document of the target patient object, and the medical document includes a current medical condition part and a motor function recovery demand target part; a semantic information extraction module 320, which is used to extract the semantic information of the medical document to obtain a motor function recovery demand semantic coding vector and a set of semantic coding vectors of the current medical condition part; a rehabilitation starting point-target semantic analysis module 330, which is used to use the motor function recovery demand semantic coding vector as the rehabilitation target vector and the set of semantic coding vectors of the current medical condition part as the set of rehabilitation starting point vectors, and input them into a rehabilitation starting point-target semantic analysis network to obtain a rehabilitation starting point-target semantic dynamic response coding vector; a personalized rehabilitation plan generation module 340, which is used to generate a personalized rehabilitation plan based on the rehabilitation starting point-target semantic dynamic response coding vector.

[0039] As described above, the intelligent analysis system 300 for rehabilitation treatment of lumbar disc herniation according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with an intelligent analysis algorithm for rehabilitation treatment of lumbar disc herniation. In one possible implementation, the intelligent analysis system 300 for rehabilitation treatment of lumbar disc herniation according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the intelligent analysis system 300 for rehabilitation treatment of lumbar disc herniation can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the intelligent analysis system 300 for rehabilitation treatment of lumbar disc herniation can also be one of the many hardware modules of the wireless terminal.

[0040] Alternatively, in another example, the intelligent analysis system 300 for rehabilitation treatment of lumbar disc herniation and the wireless terminal may also be separate devices, and the intelligent analysis system 300 for rehabilitation treatment of lumbar disc herniation may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0041] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. An intelligent analysis method for rehabilitation treatment of lumbar disc herniation, characterized in that: include: Obtain the medical condition file of the target patient, which includes the current condition and the motor function recovery target; Extracting semantic information from the medical condition document to obtain a set of semantic coding vectors of motor function recovery requirements and semantic coding vectors of the current medical condition; The semantic coding vector of the motor function recovery demand is used as the rehabilitation target vector and the set of the semantic coding vectors of the current condition is used as the set of rehabilitation starting point vectors, which are input into the rehabilitation starting point-target semantic analysis network to obtain the rehabilitation starting point-target semantic dynamic response coding vector, including: extracting the deep implicit coding features of the rehabilitation target vector and the set of rehabilitation starting point vectors respectively, and constructing the local feature response matrix between the two based on the dynamic semantic matching mechanism to obtain the set of rehabilitation starting point-target local semantic feature response anchor coding matrices; realizing multi-matrix weighted fusion through an adaptive weight allocation mechanism to generate the rehabilitation starting point-target semantic dynamic response coding vector; Generate a personalized rehabilitation plan based on the rehabilitation starting point-goal semantic dynamic response encoding vector.

2. The intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to claim 1, characterized in that: The current condition includes basic information about the patient, description of the lesion site, and description of the pain level; The motor function recovery demand target section contains descriptions of the motor function recovery demand.

3. The intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to claim 2, characterized in that: Extract the semantic information of the medical condition document to obtain a set of semantic coding vectors of motor function recovery requirements and semantic coding vectors of the current medical condition, including: The descriptions related to the lesion site, the descriptions related to the pain level, the descriptions related to the motor function recovery requirements and the basic information of the patient are semantically embedded and encoded to obtain the semantic encoding vectors of the lesion site description, the pain level description, the motor function recovery requirements and the basic information of the patient; The semantic coding vector describing the lesion site, the semantic coding vector describing the pain level and the semantic coding vector of the patient's basic information are merged to obtain a set of semantic coding vectors for the current condition.

4. The intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to claim 2, characterized in that: Basic patient information included age, gender, weight, and medical history.

5. The intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to claim 1, characterized in that: The deep implicit coding features of the sets of rehabilitation target vectors and rehabilitation starting point vectors are extracted respectively, and a local feature response matrix between the two is constructed based on a dynamic semantic matching mechanism to obtain a set of rehabilitation starting point-target local semantic feature response anchor coding matrices, including: Performing deep latent feature extraction based on fully connected coding on the rehabilitation target vector to obtain a deep latent coding vector of the rehabilitation target feature; Performing deep implicit feature extraction based on fully connected coding on each rehabilitation starting point vector in the set of rehabilitation starting point vectors to obtain a set of rehabilitation starting point feature deep implicit coding vectors; Rehabilitation starting point-target semantic dynamic response encoding is performed on each rehabilitation starting point feature deep implicit coding vector in the set of rehabilitation target feature deep implicit coding vector and rehabilitation starting point feature deep implicit coding vector to obtain a set of rehabilitation starting point-target local semantic feature response anchor coding matrices.

6. The intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to claim 5, characterized in that: The adaptive weight allocation mechanism is used to achieve multi-matrix weighted fusion and generate a rehabilitation starting point-target semantic dynamic response encoding vector, including: Calculating the decision anchor adaptive splicing weight factor of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing weight factors; Based on a set of rehabilitation starting point-target decision anchor adaptive splicing weight factors, a set of rehabilitation starting point-target local semantic feature response anchor encoding matrices is fused to obtain a rehabilitation starting point-target semantic dynamic response encoding vector.

7. The intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to claim 6, characterized in that: The decision anchor adaptive splicing weight factor of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices is calculated to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing weight factors, including: Based on the feature distribution of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices, determining the rehabilitation starting point-target decision anchor adaptive splicing factor of each rehabilitation starting point-target local semantic feature response anchor coding matrix to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing factors; The set of rehabilitation starting point-target decision anchor adaptive splicing factors is weighted based on the Softmax function to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing weight factors.

8. The intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to claim 7, characterized in that: Based on the feature distribution of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices, the rehabilitation starting point-target decision anchor adaptive splicing factor of each rehabilitation starting point-target local semantic feature response anchor coding matrix is ​​determined to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing factors, including: Based on the eigenvalues ​​of each rehabilitation starting point-target local semantic feature response anchor coding matrix in the set of rehabilitation starting point-target local semantic feature response anchor coding matrices, a state transition relationship from the mean to the maximum value is constructed through the drift coefficient, and the intermediate transition weights are globally controlled with the eigenvalue as the important score to obtain a set of rehabilitation starting point-target decision anchor adaptive splicing factors.

9. The intelligent analysis method for rehabilitation treatment of lumbar disc herniation according to claim 1, characterized in that: Generate a personalized rehabilitation plan based on the rehabilitation starting point-goal semantic dynamic response encoding vector, including: The rehabilitation starting point-target semantic dynamic response encoding vector is input into the rehabilitation plan generator based on the big model to obtain a personalized rehabilitation plan.

10. An intelligent analysis system for rehabilitation treatment of lumbar disc herniation, characterized in that: include: A target patient object medical condition document acquisition module is used to acquire the target patient object's medical condition document, which includes a medical condition status part and a motor function recovery demand target part; A semantic information extraction module is used to extract the semantic information of the medical condition document to obtain a set of semantic coding vectors of motor function recovery requirements and semantic coding vectors of the current medical condition; A rehabilitation starting point-target semantic analysis module is used to use the motor function recovery demand semantic coding vector as the rehabilitation target vector and the set of the disease status partial semantic coding vector as the set of rehabilitation starting point vectors, and input them into the rehabilitation starting point-target semantic analysis network to obtain the rehabilitation starting point-target semantic dynamic response coding vector; The personalized rehabilitation program generation module is used to generate a personalized rehabilitation program based on the rehabilitation starting point-target semantic dynamic response encoding vector.

Images