Cervical vertebra state recovery evaluation system based on deep learning

Abstract

The invention discloses a cervical vertebra state recovery evaluation system based on deep learning, and relates to the technical field of cervical vertebra recovery evaluation. The cervical vertebra state recovery evaluation system based on deep learning comprises a recovery data acquisition unit for acquiring recovery state data of a cervical vertebra region; the recovery feature extraction unit is used for performing feature extraction processing on the recovery state data based on a pre-trained deep learning model to obtain a recovery state feature set; according to the invention, by collecting multi-dimensional data such as electromyogram signals, joint stress, movement range, local temperature and blood flow rate in real time, tiny changes in the recovery process of a patient can be accurately captured; feature extraction is carried out on the time series data through a CNN-LSTM hybrid model, so that the recovery state after treatment is dynamically evaluated, and the limitation of static evaluation in the traditional technology is overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of cervical spine recovery assessment, and particularly to a cervical spine state recovery assessment system based on deep learning. Background Art

[0002] Cervical spine diseases are relatively common health problems among modern people. Especially with the changes in lifestyle and the long-term use of electronic products, the incidence of cervical spondylosis has been increasing year by year. Cervical spondylosis is usually caused by degenerative changes, injuries or improper postures of the cervical spine, which may lead to a series of symptoms such as neck pain, limited mobility, and nerve compression, seriously affecting the quality of life and work efficiency of patients. In order to effectively relieve symptoms and accelerate recovery, physical therapy, drug therapy, surgical therapy and other means are often used clinically. However, no matter which treatment method is used, the monitoring and assessment of the state during the recovery process are very crucial.

[0003] The prior art, such as the invention patent application with publication number CN113162028B, discloses a method and system for assessing cervical spine state recovery based on deep learning. The method includes: generating a cervical spine recovery assessment request through a user terminal; sending the cervical spine recovery assessment request to a temperature-controlled cervical traction device to collect the real-time abnormal curvature of the patient's cervical spine and the patient's cervical spine treatment data; setting constraint conditions according to the real-time abnormal curvature and the patient's body data, generating an optimal treatment curve based on a time scale, determining the target treatment function of the patient, and determining the ideal cervical traction trajectory; evaluating the ideal cervical traction trajectory and the patient's cervical spine treatment data through a trained deep neural network supervised model on the time scale, and outputting an assessment result of cervical spine recovery. Through the above technical solution, the recovery situation of the patient is judged, and at the same time, it is judged whether the patient's treatment plan has achieved the optimal treatment effect.

[0004] Based on the above scheme, it is found that the limitations of the prior art at least include the following problems. First, the cervical spine recovery assessment system in the prior art usually relies on traditional static assessment methods, that is, simple comparison of the states before and after treatment through the measurement data of a single sensor. It lacks dynamic monitoring of the recovery process after treatment. In the assessment after treatment, many subtle changes are difficult to capture through simple comparison methods. For example, subtle fluctuations in muscle activity intensity or joint force may have a significant impact on the recovery process, but the prior art is difficult to continuously track these changes after treatment and adjust the recovery strategy in a timely manner. Since these changes are difficult to be captured and analyzed in a timely manner, the recovery effect of the patient is difficult to be dynamically optimized after treatment, and the treatment plan may lag behind as a result, prolonging the recovery time or resulting in incomplete recovery. Summary of the Invention

[0005] In view of the deficiencies of the prior art, the present invention provides a cervical spine state recovery evaluation system based on deep learning, which solves the problems in the prior art that there is a lack of dynamic monitoring of patient individual differences, which easily leads to possible lag in treatment plans and prolongs the recovery time.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A cervical spine state recovery evaluation system based on deep learning, including: a recovery data acquisition unit, which is used to acquire the recovery state data of the cervical spine area after the end of the current recovery cycle, including electromyogram signal time series data, joint force time series data, range of motion time series data, local temperature time series data, local blood flow time series data, and perform preprocessing on each of them; a recovery feature extraction unit, which is used to input the preprocessed recovery state data of the cervical spine area into a pre-trained deep learning model for feature extraction processing to obtain a recovery state feature set of the cervical spine area; a recovery feature evaluation unit, which is used to comprehensively analyze the recovery state feature set of the cervical spine area to obtain a recovery state feature index set of the cervical spine area, and perform comprehensive analysis to obtain a comprehensive evaluation index of the recovery state of the cervical spine area. The recovery state feature index set includes a muscle vitality recovery index, a joint load recovery index, a movement recovery index, a thermoregulation recovery index, and a blood circulation recovery index. The specific formulas are as follows: Among them, HzP is the comprehensive evaluation index of the recovery state of the cervical spine area, JhL is the muscle vitality recovery index of the cervical spine area, GjF is the joint load recovery index of the cervical spine area, HhF is the movement recovery index of the cervical spine area, RtJ is the thermoregulation recovery index of the cervical spine area, λ 1 is the recovery interaction adjustment coefficient stored in the database, XxH is the blood circulation recovery index of the cervical spine area, λ 2 is the movement recovery adjustment coefficient stored in the database.

[0007] Further, the electromyogram signal time series data includes muscle activity intensity values and muscle activity frequency values at several time points, the joint force time series data includes joint force values at several time points, the range of motion time series data includes activity angle values of the cervical spine in several preset directions at several time points, the local temperature time series data includes regional temperature values at several time points, and the local blood flow time series data includes regional blood flow velocity values at several time points.

[0008] Further, the deep learning model is specifically a CNN-LSTM hybrid model, including an input layer, a convolutional layer, an activation layer, a pooling layer, a recurrent layer, a fully connected layer, and an output layer.

[0009] Further, the specific steps to obtain the recovery state feature set of the cervical vertebra region are as follows: In the input layer of the CNN-LSTM hybrid model, receive the preprocessed recovery state data of the cervical vertebra region; in the convolutional layer of the CNN-LSTM hybrid model, perform feature extraction processing on the received preprocessed recovery state data of the cervical vertebra region; in the activation layer of the CNN-LSTM hybrid model, perform non-linear activation processing on the features output by the convolutional layer; in the pooling layer of the CNN-LSTM hybrid model, perform downsampling processing on the features output by the convolutional layer; in the recurrent layer of the CNN-LSTM hybrid model, perform recurrent processing on the preprocessed recovery state data of the cervical vertebra region; in the fully connected layer of the CNN-LSTM hybrid model, integrate the features extracted by the convolutional layer and the recurrent layer, and generate the recovery state feature set of the cervical vertebra region; in the output layer of the CNN-LSTM hybrid model, output the recovery state feature set generated by the fully connected layer, and the recovery state feature set includes the electromyogram signal time series feature set, the joint force time series feature set, the range of motion time series feature set, the local temperature time series feature set, and the local blood flow time series feature set.

[0010] Further, the electromyogram signal time series feature set includes the peak intensity of muscle activity, the average intensity of muscle activity, the average frequency value of muscle activity, and the fluctuation amplitude value of muscle activity. The specific steps to obtain the muscle vitality recovery index of the cervical vertebra region are as follows: Read the peak intensity of muscle activity, the average intensity of muscle activity, the average frequency value of muscle activity, and the fluctuation amplitude value of muscle activity in the cervical vertebra region, and perform comprehensive analysis to obtain the muscle vitality recovery index of the cervical vertebra region. The specific formula is as follows: Among them, HhL is the muscle vitality recovery index of the cervical vertebra region, FzQ is the peak intensity of muscle activity in the cervical vertebra region, α 1 is the peak intensity adjustment coefficient stored in the database, PjQ is the average intensity of muscle activity in the cervical vertebra region, α 2 is the average intensity adjustment coefficient stored in the database, BdF is the fluctuation amplitude value of muscle activity in the cervical vertebra region, α 3 is the fluctuation amplitude adjustment coefficient stored in the database, PjP is the average frequency value of muscle activity in the cervical vertebra region, α 4 is the average frequency adjustment coefficient stored in the database.

[0011] Further, the joint force time series feature set includes the maximum joint force value, the joint force fluctuation value, and the joint load frequency value. The specific steps to obtain the joint load recovery index of the cervical vertebra region are as follows: Read the maximum joint force value, the joint force fluctuation value, and the joint load frequency value in the cervical vertebra region, and perform comprehensive analysis to obtain the joint load recovery index of the cervical vertebra region. The specific formula is as follows: Among them, GjF is the joint load recovery index of the cervical vertebra region, GzS is the maximum joint force value of the cervical vertebra region, μ 1 is the maximum force adjustment coefficient stored in the database, GfP is the joint force fluctuation value of the cervical vertebra region, μ 2 is the force fluctuation adjustment coefficient stored in the database, GsB is the joint load frequency value of the cervical vertebra region, μ 3 is the load frequency adjustment coefficient stored in the database.

[0012] Furthermore, the range of motion time series feature set includes the maximum motion angle value, minimum motion angle value, and motion amplitude value of the cervical vertebra in a plurality of preset directions. The specific steps to obtain the motion recovery index of the cervical vertebra region are as follows: Read the maximum motion angle value, minimum motion angle value, and motion amplitude value of the cervical vertebra in a plurality of preset directions, and perform comprehensive analysis to obtain the motion recovery index of the cervical vertebra region.

[0013] Furthermore, the local temperature time series feature set includes the regional temperature fluctuation amplitude value, regional temperature standard deviation value, and regional temperature change rate value. The specific steps to obtain the heat regulation recovery index of the cervical vertebra region are as follows: Read the regional temperature fluctuation amplitude value, regional temperature standard deviation value, and regional temperature change rate value of the cervical vertebra region, and perform comprehensive analysis to obtain the heat regulation recovery index of the cervical vertebra region.

[0014] Furthermore, the local blood flow time series feature set includes the regional maximum blood flow rate value, regional blood flow fluctuation value, and regional blood flow change rate value. The specific steps to obtain the blood circulation recovery index of the cervical vertebra region are as follows: Read the regional maximum blood flow rate value, regional blood flow fluctuation value, and regional blood flow change rate value of the cervical vertebra region, and perform comprehensive analysis to obtain the blood circulation recovery index of the cervical vertebra region.

[0015] Furthermore, after obtaining the comprehensive evaluation index of the recovery state of the cervical vertebra region, it is judged and analyzed with a preset recovery evaluation interval. The specific steps are as follows: Judge whether the comprehensive evaluation index of the recovery state of the cervical vertebra region is within the preset recovery evaluation interval; if it is within the preset recovery evaluation interval, it is regarded as normal recovery; if it is outside the preset recovery evaluation interval, it is regarded as abnormal recovery, and an abnormal alarm is sent to the relevant staff.

[0016] The present invention has the following beneficial effects:

[0017] (1) The cervical spine state recovery evaluation system based on deep learning can accurately capture the minute changes during the patient's recovery process by collecting multi-dimensional data such as electromyogram signals, joint forces, range of motion, local temperature, and blood flow rate in real time. It extracts features from these time-series data through a CNN-LSTM hybrid model, thereby realizing the dynamic evaluation of the recovery state after treatment. This dynamic monitoring not only overcomes the limitations of static evaluation in traditional technologies but also can promptly detect subtle deviations during the recovery process, provide real-time feedback, and ensure that the treatment plan can be adjusted according to the changes in the recovery process. The system can effectively optimize the treatment effect, shorten the recovery time, and reduce the possible delayed reactions or incomplete recoveries during the treatment process.

[0018] (2) The cervical spine state recovery evaluation system based on deep learning can conduct personalized recovery evaluations according to the specific physiological characteristics and recovery processes of each patient. Through the analysis of multi-dimensional physiological data, the system can identify potential special problems that may occur during the patient's recovery, such as poor local blood flow or excessive accumulation of muscle fatigue. By extracting personalized features through a deep learning model, the system can automatically adjust the recovery strategy according to the specific situation of the patient. Different from traditional fixed treatment plans, this system can accurately match the recovery needs of patients, provide customized treatment evaluations and adjustments for each patient, ensure the smoothness and effectiveness of the recovery process, and avoid treatment deviations where individual differences are not fully considered.

[0019] (3) The cervical spine state recovery evaluation system based on deep learning conducts multi-dimensional comprehensive analysis by combining multiple physiological data such as electromyogram signals, joint forces, range of motion, local temperature, and blood flow, providing a comprehensive recovery evaluation. Traditional evaluation methods usually rely on single data or static comparisons and are difficult to fully reflect the patient's recovery state. However, this system can comprehensively evaluate the patient's rehabilitation situation from multiple perspectives, including key factors such as muscle vitality, joint load, and blood circulation. This comprehensiveness not only can provide a more accurate analysis of the recovery state but also can promptly identify potential problems during the recovery process and make targeted adjustments to ensure the coordination and effectiveness of the recovery in all aspects, thereby improving the overall treatment efficiency and the patient's recovery quality.

[0020] Of course, it is not necessary for any product implementing the present invention to simultaneously achieve all the above-mentioned advantages. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a block diagram of a cervical spine state recovery evaluation system based on deep learning according to the present invention.

[0022] Figure 2 It is a specific step flow chart for obtaining the recovery state feature set of the cervical spine region in a cervical spine state recovery evaluation system based on deep learning according to the present invention. Detailed implementation mode

[0023] Please refer to Figure 1 , an embodiment of the present invention provides a technical solution: a cervical spine state recovery evaluation system based on deep learning, including: a recovery data acquisition unit, configured to acquire recovery state data of the cervical spine region after the end of the current recovery cycle, including electromyogram signal time series data, joint force time series data, range of motion time series data, local temperature time series data, local blood flow time series data, and perform preprocessing on each of them; a recovery feature extraction unit, configured to input the preprocessed recovery state data of the cervical spine region into a pre-trained deep learning model for feature extraction processing to obtain a recovery state feature set of the cervical spine region; a recovery feature evaluation unit, configured to comprehensively analyze the recovery state feature set of the cervical spine region to obtain a recovery state feature index set of the cervical spine region, and perform comprehensive analysis to obtain a comprehensive evaluation index of the recovery state of the cervical spine region. The recovery state feature index set includes a muscle vitality recovery index, a joint load recovery index, a movement recovery index, a thermoregulation recovery index, and a blood circulation recovery index.

[0024] The electromyogram signal time series data includes muscle activity intensity values and muscle activity frequency values at several time points. The joint force time series data includes joint force values at several time points. The range of motion time series data includes activity angle values of the cervical spine in several preset directions at several time points. The local temperature time series data includes regional temperature values at several time points. The local blood flow time series data includes regional blood flow velocity values at several time points.

[0025] Among them, the muscle activity intensity value refers to the voltage fluctuation amplitude of the electromyogram (EMG) signal within a certain period of time, which reflects the activation intensity of the muscle and can be measured and obtained through an electromyogram (EMG) sensor.

[0026] The muscle activity frequency value refers to the number of fluctuations of the muscle activity signal per unit time, which represents the rhythm and speed of muscle activity, that is, the frequency of muscle contraction and relaxation within a certain period of time, and can be measured and obtained through an electromyogram (EMG) sensor.

[0027] The joint force value refers to the magnitude of the force borne by the joint, indicating the intensity of joint force and reflecting the load-bearing capacity of the joint under specific activities, and can be measured and obtained through a force sensor or a pressure sensor (for example, a pressure pad, a pressure sensor gasket, etc.).

[0028] The activity angle value reflects the range of motion of the cervical spine in different directions. For example, the angle changes of the cervical spine in forward flexion, backward extension, left and right rotation, etc. represent the angle range that the cervical spine can move during different movements, and can be measured and obtained through an accelerometer, a gyroscope or an optical motion capture system.

[0029] Regional temperature value, which reflects the temperature changes in a local area (such as the neck, shoulder, etc.). Such temperature changes are usually closely related to physiological processes such as local blood flow, inflammation, and metabolic activities, and can be measured and obtained through an infrared sensor or a thermal imaging device.

[0030] Regional blood flow velocity value, which reflects the blood flow condition in a local area (such as the neck, shoulder, etc.). The change in blood flow rate is closely related to physiological processes such as blood circulation and the delivery of oxygen and nutrients, and can be measured and obtained through a blood flow sensor or an ultrasonic blood flow meter.

[0031] Among them, the specific formula for calculating the comprehensive evaluation index of the recovery status of the cervical spine region is as follows: Among them, HzP is the comprehensive evaluation index of the recovery status of the cervical spine region, JhL is the muscle vitality recovery index of the cervical spine region, GjF is the joint load recovery index of the cervical spine region, HhF is the activity recovery index of the cervical spine region, RtJ is the thermal regulation recovery index of the cervical spine region, λ 1 is the recovery interaction adjustment coefficient stored in the database, XxH is the blood circulation recovery index of the cervical spine region, λ 2 is the activity recovery adjustment coefficient stored in the database.

[0032] It should be explained that the recovery interaction adjustment coefficient λ stored in the database 1 and the activity recovery adjustment coefficient λ 2 The specific acquisition steps are as follows: For the recovery interaction adjustment coefficient λ 1 : First, the system will input data of different recovery cycles, including muscle vitality recovery, joint load recovery, and activity recovery data. These data are collected in real time through sensors or other measurement tools. The obtained parameter data such as muscle vitality, joint load, and activity recovery are stored in the database, and based on the analysis of historical data and existing data, machine learning or statistical models are used to evaluate the interaction between these parameters. According to the input data and the historical data model, the relationship strength between recovery cycles is calculated, and λ 1 is adjusted to optimize the weight of recovery interaction. If a certain recovery index in a recovery cycle has a greater impact, then the corresponding λ 1 will increase, otherwise it will decrease to ensure the balance of the recovery process; For the activity recovery adjustment coefficient λ 2 : The system collects real-time data from the thermal regulation recovery index, blood circulation recovery index, and activity recovery. These data are obtained through devices such as temperature sensors and blood flow sensors. The physiological data related to activity recovery are stored in the database. Through the analysis of historical recovery data, it is evaluated how thermal regulation and blood circulation affect the process of activity recovery. According to the analysis of the input data and the model, λ 2The value will be adjusted during the active recovery process of each recovery cycle. For example, if blood flow increases and thermoregulation becomes more effective, λ 2 The value will increase to indicate a higher active recovery effect.

[0033] The specific implementation example of calculating the comprehensive evaluation index of the recovery status of the cervical spine area is as follows. The following data is available:

[0034] The muscle vitality recovery index of the cervical spine area is approximately: 2.657.

[0035] The joint load recovery index of the cervical spine area is approximately: 1.236.

[0036] The active recovery index of the cervical spine area is approximately: 1.435.

[0037] The thermoregulation recovery index of the cervical spine area is approximately: 2.751.

[0038] The recovery interaction regulation coefficient stored in the database is approximately: 1.121.

[0039] The blood circulation recovery index of the cervical spine area is approximately: 1.559.

[0040] The active recovery regulation coefficient stored in the database is approximately: 0.863.

[0041] Substitute the above data into the specific formula for calculating the comprehensive evaluation index of the recovery status of the cervical spine area respectively, and we get:

[0042] The comprehensive evaluation index of the recovery status of the cervical spine area = (((2.657×1.236) / (1 + 1.435 + 2.751))^1.121)×(1.559 / (1+(0.863×(1.435 / (2.657 + 1)))))≈((3.284 / 5.176)^1.121)×(1.559 / (1+(0.863×0.392)))≈0.6004×1.165≈0.699.

[0043] Specifically, the deep learning model is specifically a CNN-LSTM hybrid model (i.e., a hybrid model of convolutional neural network and long short-term memory network), including an input layer, a convolutional layer, an activation layer, a pooling layer, a recurrent layer, a fully connected layer, and an output layer.

[0044] The pre-training steps of the deep learning model are as follows:

[0045] Data collection: Collect a large amount of cervical spine recovery data, including electromyogram signals, joint forces, range of motion, temperature, blood flow, etc. The data should include data points at multiple time steps, and each data point should contain relevant numerical features.

[0046] Data preprocessing: Standardize and normalize the data to ensure that the data can be fed into the deep learning model for training.

[0047] Select a CNN-LSTM hybrid model, which is suitable for processing time series data and can capture both local features and temporal dependencies simultaneously.

[0048] Training data: Use the preprocessed training dataset (including all time series data and related labels) to train the model.

[0049] Loss function: Select an appropriate loss function according to the objective of the task. For example, if it is a regression problem, use the mean squared error (MSE) as the loss function; if it is a classification problem, use the cross-entropy loss function.

[0050] Optimizer: Commonly used optimizers such as Adam are used to optimize the model parameters.

[0051] Evaluate the performance of the model on the validation set and optimize the model by adjusting hyperparameters (such as learning rate, network depth, etc.).

[0052] Use accuracy, F1 score, AUC, etc. to evaluate the performance of the model.

[0053] In this implementation, by using a CNN-LSTM hybrid model, the system can effectively process time series data, capture local features and long-term temporal dependencies, provide in-depth analysis of the cervical spine recovery process. The pre-training steps of the model include data collection, data preprocessing, training, and optimization to ensure that the model can accurately process complex physiological data and perform personalized recovery assessments. The standardization and normalization of the data make the numerical features from different sources have a unified scale, improving the stability and accuracy of model training. By adopting appropriate loss functions (such as mean squared error or cross-entropy) and optimizers (such as Adam), the system can continuously optimize the model parameters, accurately fit the data, and ensure the best recovery assessment effect. The evaluation on the validation set and hyperparameter adjustment further enhance the generalization ability and performance of the model. The model is evaluated from multiple angles through indicators such as accuracy, F1 score, and AUC to ensure that it can efficiently and accurately process cervical spine recovery data, providing reliable support for practical applications.

[0054] Specifically, as Figure 2As shown below, the specific steps to obtain the recovery state feature set of the cervical spine region are as follows: In the input layer of the CNN-LSTM hybrid model, it receives the preprocessed recovery state data of the cervical spine region, that is, for each type of time-series data (such as electromyogram signals, joint forces, etc.), it is organized into a data sequence according to time steps. The data will undergo standardization or normalization processing to ensure that the values are within an appropriate range, avoiding problems such as gradient explosion or gradient disappearance during the training process. All the data is combined into a unified input format for subsequent model layers to process; In the convolutional layer of the CNN-LSTM hybrid model, it performs feature extraction processing on the received preprocessed recovery state data of the cervical spine region. That is, the convolutional kernel performs convolution with the input data through a sliding window operation to generate a feature map. Each convolutional kernel searches for specific patterns (such as peaks and frequency changes of electromyogram signals) in the data at different time steps. During the convolution operation, the weights of the convolutional kernel will be updated during the training process to optimize the feature extraction ability. The output of the convolutional layer is a new feature map, indicating the activation situation of each local area in the input data; In the activation layer of the CNN-LSTM hybrid model, it performs non-linear activation processing on the features output by the convolutional layer. The ReLU (Rectified Linear Unit) activation function is usually used in this layer. ReLU sets negative values to zero and keeps positive values unchanged. It can effectively avoid the problem of gradient disappearance and accelerate the training process of the model. The role of the activation layer is to introduce non-linearity, enabling the model to learn complex patterns and non-linear relationships in the data and enhancing the expressive ability of the model. Through non-linear transformation, the activation function enables the network to recognize non-linear patterns such as muscle activity or complex changes in joint load; In the pooling layer of the CNN-LSTM hybrid model, it performs downsampling processing on the features output by the convolutional layer, including max pooling: within a small area (such as a 2x2 or 3x3 window), the maximum value in the area is selected as the pooling result, which helps to extract the most significant features and reduce unnecessary details; average pooling: the average value in the area is selected for downsampling, which is usually used to smooth the data; Through the pooling operation, the pooling layer reduces the size of the feature map while retaining the most important information in the input data, enhancing the robustness of the model to minor changes; In the recurrent layer of the CNN-LSTM hybrid model, it performs recurrent processing on the preprocessed recovery state data of the cervical spine region. That is, the LSTM unit stores the information of the previous time step through a memory unit and decides which information needs to be retained and which needs to be forgotten based on the input and the previous state. The LSTM contains three main gating mechanisms: the input gate, the forget gate, and the output gate, which help the LSTM network control the flow of information. The LSTM layer will capture the long-term dependencies of time-series data, such as the long-term change trend of muscle activity and the evolution of joint forces over time;In the fully connected layer of the CNN-LSTM hybrid model, the features extracted by the convolutional layer and the recurrent layer are integrated to generate a set of recovery state features for the cervical region, that is, all local and temporal features extracted by the convolutional layer and the LSTM layer are integrated; in the output layer of the CNN-LSTM hybrid model, the set of recovery state features for the cervical region generated by the fully connected layer is output. The set of recovery state features includes the set of electromyogram signal temporal features, the set of joint force temporal features, the set of range of motion temporal features, the set of local temperature temporal features, and the set of local blood flow temporal features.

[0055] In this implementation, through the detailed CNN-LSTM hybrid model process, complex temporal data in cervical spine recovery assessment is effectively processed. First, by standardizing and normalizing the data, the training stability of the model is ensured, and common gradient problems are avoided. The convolutional layer is used to extract local features from the input data and can identify subtle changes in muscle activity, joint load, etc., further improving the accuracy of recovery state assessment. The activation layer introduces non-linear functions, enabling the model to handle complex patterns, especially the non-linear features of physiological signals such as muscle activity and joint force. The pooling layer reduces unnecessary details through downsampling, enhancing the model's adaptability to subtle changes. The recurrent layer captures the long-term dependencies of temporal data through LSTM units, ensuring that the model can handle the dynamic changes during the recovery process. Finally, feature integration is completed in the fully connected layer, generating a comprehensive set of recovery state features for the cervical region, providing strong data support for comprehensive assessment. These steps ensure that this system can accurately, stably, and effectively perform personalized recovery assessment and provide real-time optimization guidance for treatment plans.

[0056] Specifically, the set of electromyogram signal temporal features includes the peak intensity of muscle activity, the average intensity of muscle activity, the average frequency value of muscle activity, and the fluctuation amplitude value of muscle activity. The specific steps to obtain the muscle vitality recovery index for the cervical region are as follows: Read the peak intensity of muscle activity, the average intensity of muscle activity, the average frequency value of muscle activity, and the fluctuation amplitude value of muscle activity in the cervical region and perform comprehensive analysis to obtain the muscle vitality recovery index for the cervical region.

[0057] Among them, the peak intensity of muscle activity represents the maximum amplitude of the electromyogram signal at several time points, which reflects the highest activation intensity of the muscle.

[0058] The average intensity of muscle activity represents the average amplitude of the electromyogram signal within a certain time at several time points.

[0059] The average frequency value of muscle activity represents the average value of the frequency of muscle activity signals per unit time.

[0060] The muscle activity fluctuation amplitude value represents the fluctuation range of the electromyogram signal within several time points, that is, the change amplitude of the signal.

[0061] Among them, the specific formula for calculating the muscle vitality recovery index of the cervical region is as follows: Among them, JhL is the muscle vitality recovery index of the cervical region, FzQ is the peak muscle activity intensity of the cervical region, α 1 is the peak activity intensity adjustment coefficient stored in the database, PjQ is the average muscle activity intensity of the cervical region, α 2 is the average activity intensity adjustment coefficient stored in the database, BdF is the muscle activity fluctuation amplitude value of the cervical region, α 3 is the activity fluctuation amplitude adjustment coefficient stored in the database, PjP is the average muscle activity frequency value of the cervical region, α 4 is the average activity frequency adjustment coefficient stored in the database.

[0062] It should be explained that the specific acquisition steps of the peak activity intensity adjustment coefficient α 1 , the average activity intensity adjustment coefficient α 2 , the activity fluctuation amplitude adjustment coefficient α 3 , and the average activity frequency adjustment coefficient α 4 are as follows: Real-time collect various activity data through sensors, including the peak intensity, average intensity, fluctuation amplitude, and frequency of muscles. These data are stored in the database. Then, through regression analysis or deep learning modeling of historical recovery data, extract the relationship between these activity parameters and the recovery state, and determine the influence degree of each factor on the recovery. Based on these analysis results, dynamically adjust the corresponding adjustment coefficients (such as α 1 , α 2 , α 3 , α 4 ) to ensure that each coefficient can accurately reflect the regulatory effect of its corresponding activity factor on the recovery process, thereby optimizing the overall recovery state.

[0063] In this implementation, by elaborating on the calculation process of the electromyogram signal timing feature set and its relationship with the recovery status, it fully demonstrates how the system utilizes multiple physiological data to precisely evaluate the cervical spine recovery process. Specifically, through the analysis of four important features, namely the peak intensity, average intensity, frequency, and fluctuation amplitude of muscle activity, the system can comprehensively reflect the state of muscle vitality. Each feature plays an important role in the recovery process, revealing muscle fatigue, activation intensity, and recovery conditions at different stages. For example, the peak intensity reflects the highest activation level of the muscle, indicating whether the muscle has recovered to a sufficient vitality level; the average intensity can evaluate the stability of muscle activity; the frequency and fluctuation amplitude help to judge the persistence and stability of the muscle, avoiding problems such as excessive fatigue. By combining the adjustment coefficients stored in the database (such as the peak activity intensity adjustment coefficient, activity average intensity adjustment coefficient, etc.), the system can not only analyze the characteristics of muscle activity but also make dynamic adjustments according to the specific situation of the patient. This approach makes the evaluation process more personalized and enables precise adjustment of treatment strategies at different recovery stages, ensuring the effectiveness and personalization of treatment.

[0064] Specifically, the joint force timing feature set includes the maximum joint force value, joint force fluctuation value, and joint load frequency value. The specific steps to obtain the joint load recovery index for the cervical spine region are as follows: Read the maximum joint force value, joint force fluctuation value, and joint load frequency value of the cervical spine region and conduct comprehensive analysis to obtain the joint load recovery index for the cervical spine region.

[0065] Among them, the maximum joint force value represents the maximum load borne by the joint during movement at several time points.

[0066] The joint force fluctuation value represents the fluctuation range of the joint load change during movement at several time points, that is, the amplitude of the load change.

[0067] The joint load frequency value represents the frequency of joint load change per unit time.

[0068] Among them, the specific formula for calculating the joint load recovery index for the cervical spine region is as follows: Among them, GjF is the joint load recovery index for the cervical spine region, GzS is the maximum joint force value for the cervical spine region, μ 1 is the maximum force adjustment coefficient stored in the database, GfP is the joint force fluctuation value for the cervical spine region, μ 2 is the force fluctuation adjustment coefficient stored in the database, GsB is the joint load frequency value for the cervical spine region, μ 3 is the load frequency adjustment coefficient stored in the database.

[0069] It should be noted that the maximum force adjustment coefficient μ stored in the database1 、Force fluctuation adjustment coefficient μ 2 、Load frequency adjustment coefficient μ 3 The specific acquisition steps are as follows: By collecting relevant data during the activity process in real time, such as the maximum force, force fluctuation, and load frequency, etc., these data are collected by sensors (such as pressure sensors) and stored in the database. Next, using the historical data in the database, statistical analysis or machine learning models are used to analyze the relationship between these parameters and the recovery effect. Through regression analysis and optimization algorithms, the adjustment effect of each coefficient on the recovery is extracted. Through these analyses, the system dynamically adjusts these adjustment coefficients (μ 1 、μ 2 、μ 3 ), to ensure that the influence of each factor is maximized and the recovery process is optimized.

[0070] In this implementation plan, through the detailed analysis of the maximum force value, force fluctuation value, and load frequency value of the joint, the recovery status of the joints in the patient's cervical region can be comprehensively evaluated. Each parameter reflects different dimensions of the joint during the recovery process and can effectively detect whether the joint is overloaded or unstable. More importantly, the system dynamically adjusts the joint load for each recovery cycle by combining the adjustment coefficients in the database, thereby optimizing the recovery process. Through these precise personalized adjustments, the system can not only prevent possible problems during the joint recovery process but also effectively improve the recovery effect and speed. This method of combining real-time data collection and historical data analysis makes the recovery evaluation more scientific and accurate, thus providing a more precise and personalized treatment plan for patients.

[0071] Specifically, the range of motion time series feature set includes the maximum motion angle value, minimum motion angle value, and motion amplitude value of the cervical vertebra in several preset directions. The specific steps to obtain the motion recovery index of the cervical region are as follows: Read the maximum motion angle value, minimum motion angle value, and motion amplitude value of the cervical vertebra in several preset directions and conduct comprehensive analysis to obtain the motion recovery index of the cervical region.

[0072] Among them, the maximum motion angle value of the cervical vertebra in several preset directions represents the maximum motion angle of the cervical vertebra in the preset directions (such as forward flexion, backward extension, left and right rotation, etc.) at several time points.

[0073] The minimum motion angle value of the cervical vertebra in several preset directions represents the minimum motion angle of the cervical vertebra in the preset directions (such as forward flexion, backward extension, left and right rotation, etc.) at several time points.

[0074] The range of motion values of the cervical vertebra in a number of preset directions represents the difference between the maximum and minimum motion angles of the cervical vertebra in the preset directions (such as forward flexion, backward extension, left and right rotation, etc.) within a number of time points.

[0075] Among them, the specific formula for calculating the activity recovery index of the cervical vertebra region is as follows: Among them, HhF is the activity recovery index of the cervical vertebra region, ZdH i is the maximum motion angle value of the cervical vertebra in the i-th preset direction, HdF i is the range of motion value of the cervical vertebra in the i-th preset direction, ZxH i is the minimum motion angle value of the cervical vertebra in the i-th preset direction, i = 1, 2, 3,..., i 0 i 0 is the number of preset directions.

[0076] In this implementation scheme, by analyzing in detail the maximum motion angle value, minimum motion angle value and range of motion value of the cervical vertebra in multiple directions, the activity recovery state of the cervical vertebra can be accurately evaluated. Each parameter can reveal different aspects of the cervical vertebra during the recovery process, including the recovery of the range of motion, the improvement of motion flexibility, and the alleviation of joint stiffness problems. By synthesizing these features, the system provides a comprehensive assessment of the recovery state of the cervical vertebra, helping the therapist to monitor and optimize the treatment plan in real time. Especially by dynamically adjusting the range of motion, the system can ensure the smooth progress of the recovery process, avoid uneven recovery, and ultimately improve the overall treatment effect and promote the patient's faster and more comprehensive recovery.

[0077] Specifically, the local temperature time series feature set includes the regional temperature fluctuation amplitude value, regional temperature standard deviation value, and regional temperature change rate value. The specific steps to obtain the thermoregulatory recovery index of the cervical vertebra region are as follows: Read the regional temperature fluctuation amplitude value, regional temperature standard deviation value, and regional temperature change rate value of the cervical vertebra region, and conduct comprehensive analysis to obtain the thermoregulatory recovery index of the cervical vertebra region.

[0078] Among them, the regional temperature fluctuation amplitude value represents the fluctuation range of the local regional temperature within a number of time points, that is, the difference between the maximum temperature and the minimum temperature.

[0079] The regional temperature standard deviation value represents the standard deviation of the temperature data within a number of time points, that is, the amplitude of temperature fluctuation.

[0080] The regional temperature change rate value represents the rate of temperature change per unit time.

[0081] Among them, the specific formula for calculating the thermoregulatory recovery index of the cervical vertebra region is as follows: Wherein, RtJ is the thermal regulation recovery index of the cervical region, WbD is the regional temperature fluctuation amplitude value of the cervical region, WbH is the regional temperature change rate value of the cervical region, WbZ is the regional temperature standard deviation value of the cervical region, and δ 1 is the temperature stability regulation coefficient stored in the database, and δ 2 is the temperature regulation sensitivity coefficient stored in the database.

[0082] It should be explained that the specific steps for obtaining the temperature stability regulation coefficient δ 1 and the temperature regulation sensitivity coefficient δ 2 in the database are as follows: The temperature data of the local area, including data such as temperature fluctuation amplitude and temperature change rate, are monitored in real time through a temperature sensor. These data are stored in the database in real time for analysis. Then, through regression analysis or deep learning modeling of historical temperature recovery data, the roles of temperature change and stability in the recovery are identified. By calculating the relationship between temperature fluctuation and change and the recovery process, the model will generate adjustment values for the temperature stability and sensitivity coefficients. Finally, based on these analysis results, the δ 1 and δ 2 coefficients are dynamically adjusted to adapt to different recovery situations and ensure the effective recovery of the system under different temperature conditions.

[0083] In this implementation plan, through the refined analysis of the regional temperature fluctuation amplitude value, regional temperature standard deviation value, and regional temperature change rate value, the thermal regulation recovery situation of the cervical region can be comprehensively evaluated. The real-time monitoring of these temperature characteristics enables the system to identify problems such as excessive temperature fluctuation and change rate in the local area, preventing instability during the treatment process. Combining the temperature stability regulation coefficient and temperature regulation sensitivity coefficient stored in the database, the system can dynamically adjust the recovery strategy according to the specific recovery status of the patient to ensure that the temperature change is within the optimal range, optimize the local blood flow and metabolic processes. Through this personalized adjustment and real-time feedback method, the system can ensure the temperature stability during the recovery process of the cervical region, thereby improving the overall recovery efficiency. Finally, this technology ensures that the patient obtains the most suitable thermal regulation support during the treatment process, promoting the maximization of the cervical recovery effect.

[0084] Specifically, the local blood flow time series feature set includes the regional maximum blood flow rate value, regional blood flow fluctuation value, and regional blood flow change rate value. The specific steps for obtaining the blood circulation recovery index of the cervical region are as follows: Read the regional maximum blood flow rate value, regional blood flow fluctuation value, and regional blood flow change rate value of the cervical region and conduct comprehensive analysis to obtain the blood circulation recovery index of the cervical region.

[0085] Among them, the regional maximum blood flow rate value represents the maximum blood flow rate of the local area at several time points.

[0086] The regional blood flow fluctuation value represents the fluctuation of the local regional blood flow rate at several time points and indicates the stability of blood flow.

[0087] The regional blood flow change rate value represents the change rate of the blood flow rate per unit time.

[0088] Among them, the specific formula for calculating the blood circulation recovery index of the cervical vertebra region is as follows: Among them, XxH is the blood circulation recovery index of the cervical vertebra region, DxS is the maximum regional blood flow rate value of the cervical vertebra region, XbD is the regional blood flow fluctuation value of the cervical vertebra region, ξ 1 is the fluctuation adjustment coefficient stored in the database, XbH is the regional blood flow change rate value of the cervical vertebra region, ξ 2 is the change adjustment coefficient stored in the database, ξ 3 is the combined change and fluctuation adjustment coefficient stored in the database.

[0089] It should be explained that the specific acquisition steps of the fluctuation adjustment coefficient ξ 1 , the change adjustment coefficient ξ 2 , and the combined change and fluctuation adjustment coefficient ξ 3 are as follows: By real-time monitoring the fluctuations and changes in blood flow, collecting data (such as blood flow rate, flow change rate, etc.), and storing these data in the database. Then, through historical data analysis, a model is established to evaluate the impact of fluctuations and changes on blood circulation recovery. Through regression analysis or other statistical methods, the adjustment rules for each coefficient are obtained. According to the data analysis results, the system will dynamically adjust ξ 1 , ξ 2 , and ξ 3 to ensure that each adjustment coefficient can accurately reflect the fluctuations, changes, and their combined effects of blood flow at different recovery stages.

[0090] In this implementation plan, by analyzing the maximum regional blood flow rate value, regional blood flow fluctuation value, and regional blood flow change rate value, a comprehensive evaluation of the blood circulation recovery of the cervical vertebra region is carried out. The maximum blood flow rate value can measure the overall level of blood flow, the fluctuation value reflects the blood flow stability, and the change rate value reveals the dynamic adjustment situation during the recovery process. By combining the fluctuation adjustment coefficient, change adjustment coefficient, and combined change and fluctuation adjustment coefficient in the database, the system can dynamically adjust the blood circulation for each recovery cycle. This method ensures personalized adjustment of the blood flow recovery process, avoids the influence of excessive fluctuations or unstable blood flow on the recovery effect. Through this precise analysis and adjustment, the system can respond promptly to changes in blood circulation at different recovery stages, thereby promoting the optimization of the overall recovery process and improving the treatment effect.

[0091] Specifically, after obtaining the comprehensive evaluation index of the recovery status of the cervical region, a judgment and analysis are carried out with a preset recovery evaluation interval. The specific steps are as follows: Determine whether the comprehensive evaluation index of the recovery status of the cervical region is within the preset recovery evaluation interval; if it is within the preset recovery evaluation interval, it is regarded as normal recovery; if it is outside the preset recovery evaluation interval, it is regarded as abnormal recovery, and an abnormal alarm is sent to the relevant staff.

[0092] In this implementation plan, by calculating the comprehensive evaluation index of the recovery status of the cervical region and comparing it with the preset recovery evaluation interval, the recovery status of the patient can be accurately judged. If the recovery index is within the normal range, it indicates that the recovery process is smooth; if it exceeds the normal range, the system will immediately issue an abnormal alarm to prompt the relevant staff to take measures for intervention. This function can ensure that the patient is monitored and intervened in a timely manner at each stage of the recovery process, avoiding the situation of treatment lag or neglect. Through this real-time evaluation and alarm mechanism, the system effectively improves the pertinence and timeliness of the treatment plan, helps to optimize the overall recovery effect, avoids the occurrence of an adverse recovery process, and ensures that the patient can achieve the best treatment effect.

[0093] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications to these embodiments once they learn the basic creative concept. Therefore, the appended claims are intended to be construed as including the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0094] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. A cervical vertebra recovery assessment system based on deep learning, characterized in that: include: A recovery data acquisition unit is used to acquire recovery status data of the cervical spine area after the current recovery cycle ends, including electromyography signal time series data, joint force time series data, range of motion time series data, local temperature time series data, and local blood flow time series data, and pre-process them respectively; A recovery feature extraction unit is used to input the preprocessed recovery state data of the cervical vertebra area into a pre-trained deep learning model for feature extraction processing to obtain a recovery state feature set of the cervical vertebra area; The recovery feature evaluation unit is used to comprehensively analyze the recovery state feature set of the cervical vertebrae region to obtain the recovery state feature index set of the cervical vertebrae region, and to comprehensively analyze to obtain the recovery state comprehensive evaluation index of the cervical vertebrae region. The recovery state feature index set includes a muscle vitality recovery index, a joint load recovery index, an activity recovery index, a thermal regulation recovery index, and a blood circulation recovery index. The specific formula is as follows: Among them, HzP, JhL, GjF, HhF, RtJ, and XxH are the comprehensive evaluation index of recovery status of the cervical spine area, muscle vitality recovery index, joint load recovery index, activity recovery index, thermal regulation recovery index, and blood circulation recovery index, respectively; λ1 and λ2 are the recovery interaction adjustment coefficient and activity recovery adjustment coefficient stored in the database, respectively.

2. The cervical vertebra recovery assessment system based on deep learning according to claim 1, characterized in that: The electromyography signal timing data includes muscle activity intensity values ​​and muscle activity frequency values ​​at several time points, the joint force timing data includes joint force values ​​at several time points, the range of motion timing data includes activity angle values ​​of the cervical vertebra in several preset directions at several time points, the local temperature timing data includes regional temperature values ​​at several time points, and the local blood flow timing data includes regional blood flow velocity values ​​at several time points.

3. The cervical vertebra recovery assessment system based on deep learning according to claim 1, characterized in that: The deep learning model is specifically a CNN-LSTM hybrid model, including an input layer, a convolutional layer, an activation layer, a pooling layer, a recurrent layer, a fully connected layer, and an output layer.

4. The cervical vertebra recovery assessment system based on deep learning according to claim 3, characterized in that: The specific steps to obtain the recovery state feature set of the cervical spine area are as follows: In the input layer of the CNN-LSTM hybrid model, the preprocessed recovery status data of the cervical spine area is received; In the convolution layer of the CNN-LSTM hybrid model, feature extraction is performed on the received preprocessed recovery state data of the cervical spine area; In the activation layer of the CNN-LSTM hybrid model, nonlinear activation processing is performed on the features output by the convolutional layer; In the pooling layer of the CNN-LSTM hybrid model, the features output by the convolutional layer are downsampled; In the recurrent layer of the CNN-LSTM hybrid model, the preprocessed recovery status data of the cervical spine area is recurrently processed; In the fully connected layer of the CNN-LSTM hybrid model, the features extracted by the convolutional layer and the recurrent layer are integrated and processed to generate a recovery state feature set of the cervical spine area; In the output layer of the CNN-LSTM hybrid model, the recovery state feature set of the cervical spine area generated by the fully connected layer is output processed, and the recovery state feature set includes the electromyography signal timing feature set, the joint force timing feature set, the range of motion timing feature set, the local temperature timing feature set, and the local blood flow timing feature set.

5. The cervical vertebra state recovery assessment system based on deep learning according to claim 4, characterized in that: The electromyography signal time series feature set includes muscle activity peak intensity, muscle activity average intensity, muscle activity average frequency value, and muscle activity fluctuation amplitude value. The specific steps for obtaining the muscle vitality recovery index of the cervical spine area are as follows: The peak intensity, average intensity, average frequency and fluctuation amplitude of muscle activity in the cervical spine area are read and analyzed comprehensively to obtain the muscle vitality recovery index of the cervical spine area. The specific formula is as follows: Among them, JhL, FzQ, PjQ, BdF, and PjP are the muscle vitality recovery index, muscle activity peak intensity, muscle activity average intensity, muscle activity fluctuation amplitude value, and muscle activity average frequency value of the cervical spine area respectively; α1, α2, α3, and α4 are the activity peak intensity adjustment coefficient, activity average intensity adjustment coefficient, activity fluctuation amplitude adjustment coefficient, and activity average frequency adjustment coefficient stored in the database respectively.

6. The cervical vertebra recovery assessment system based on deep learning according to claim 4, characterized in that: The joint force time series feature set includes the joint maximum force value, the joint force fluctuation value, and the joint load frequency value. The specific steps for obtaining the joint load recovery index of the cervical spine area are as follows: The maximum joint force value, joint force fluctuation value, and joint load frequency value of the cervical spine area are read and comprehensively analyzed to obtain the joint load recovery index of the cervical spine area. The specific formula is as follows: Among them, GjF, GzS, GfP, and GsB are the joint load recovery index, maximum joint force value, joint force fluctuation value, and joint load frequency value of the cervical spine area respectively, and μ1, μ2, and μ3 are the maximum force adjustment coefficient, force fluctuation adjustment coefficient, and load frequency adjustment coefficient stored in the database respectively.

7. The cervical vertebra recovery assessment system based on deep learning according to claim 4, characterized in that: The activity range time series feature set includes the maximum activity angle value, the minimum activity angle value, and the activity amplitude value of the cervical spine in several preset directions. The specific steps for obtaining the activity recovery index of the cervical spine area are as follows: The maximum activity angle value, minimum activity angle value, and activity amplitude value of the cervical spine in several preset directions are read, and a comprehensive analysis is performed to obtain the activity recovery index of the cervical spine area.

8. The cervical vertebra recovery assessment system based on deep learning according to claim 4, characterized in that: The local temperature time series feature set includes the regional temperature fluctuation amplitude value, the regional temperature standard deviation value, and the regional temperature change rate value. The specific steps for obtaining the thermal regulation recovery index of the cervical region are as follows: The regional temperature fluctuation amplitude value, regional temperature standard deviation value, and regional temperature change rate value of the cervical spine area are read, and a comprehensive analysis is performed to obtain the thermal regulation recovery index of the cervical spine area.

9. The cervical vertebra recovery assessment system based on deep learning according to claim 4, characterized in that: The local blood flow time series feature set includes the regional maximum blood flow rate value, the regional blood flow fluctuation value, and the regional blood flow change rate value. The specific steps for obtaining the blood circulation recovery index of the cervical spine region are as follows: The regional maximum blood flow rate value, regional blood flow fluctuation value, and regional blood flow change rate value of the cervical spine area are read, and a comprehensive analysis is performed to obtain the blood circulation recovery index of the cervical spine area.

10. The cervical vertebra recovery assessment system based on deep learning according to claim 1, characterized in that: After obtaining the comprehensive evaluation index of the recovery status of the cervical spine area, a judgment analysis is performed with the preset recovery evaluation interval. The specific steps are as follows: Determine whether the comprehensive evaluation index of the recovery status of the cervical spine area is within the preset recovery evaluation range; If it is within the preset recovery assessment range, it is considered to have returned to normal; If it is outside the preset recovery assessment interval, it is considered as recovery abnormality and an abnormality alert is sent to relevant staff.

Images