A rehabilitation nursing intelligent monitoring method and system based on image acquisition

Abstract

The application discloses a rehabilitation nursing intelligent monitoring method and system based on image acquisition, and belongs to the technical field of rehabilitation nursing intelligent monitoring. The method comprises the following steps: collecting image data of a patient's limbs and synchronous pain response values to generate a monitoring model; processing current image data based on the monitoring model to evaluate and output a pain risk value; analyzing fluctuation trends of a pain response curve formed based on the pain risk value to generate and output a nursing scheme; collecting subsequent image data in the process of executing the nursing scheme, monitoring the recovery state of the limbs, and outputting recovery feedback information to adjust the nursing scheme; and through the establishment of a quantitative correlation between visual features and pain response, the application realizes continuous evaluation of pain risk, generates a nursing scheme based on dynamic fluctuation trend analysis, forms a closed-loop management process through recovery feedback information, and improves the efficiency and safety of rehabilitation nursing.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent monitoring of rehabilitation nursing, and particularly relates to an intelligent monitoring method and system for rehabilitation nursing based on image acquisition. BACKGROUND

[0002] With the continuous rise of population aging and the incidence of chronic diseases, the demand for professional rehabilitation nursing services is increasing. The traditional rehabilitation nursing mode highly depends on the on-site observation and manual assessment of medical staff. This mode not only has high labor intensity and low efficiency, but also is easily affected by subjective factors, and is difficult to realize standardization and scaling. Therefore, it is very important to introduce intelligent and automatic monitoring technology to objectively and real-time monitor and evaluate the rehabilitation process of patients.

[0003] The traditional scheme relies on physiological electrical signals such as electroencephalogram or electromyogram to judge the pain response or muscle activation state of patients. However, the expression of pain perception and physiological electrical signals has strong individual difference, and different patients have completely different responses to the same stimulus, which makes it difficult to make reliable cross-individual comparison and evaluation. In addition, physiological electrical signals mainly reflect the electrical activity of neuromuscular, and cannot directly capture the kinematic information which is crucial in rehabilitation training, such as joint activity angle, smoothness of motion trajectory and existence of compensatory action, etc. The lack of information dimension makes the system not comprehensive enough in evaluating the overall situation of the recovery of patients' limb function, which is not conducive to the design of nursing scheme.

[0004] Therefore, there is an urgent need for a new rehabilitation nursing monitoring method and system in the industry. SUMMARY

[0005] The purpose of the present application is to provide an intelligent monitoring method for rehabilitation nursing based on image acquisition, which can accurately and carefully evaluate the pain and recovery of the patient's limbs, and also can feedback the corresponding nursing scheme according to the evaluation results.

[0006] To achieve the above-mentioned purpose of the application, the technical scheme adopted by the present application is as follows: an intelligent monitoring method for rehabilitation nursing based on image acquisition, comprising the following steps:

[0007] When monitoring the rehabilitation nursing of patients, the current image data of the patients is collected in real time, and the pain risk value representing the current state is evaluated and output based on the pre-generated monitoring model;

[0008] Based on the plurality of pain risk values recorded in the preset time period, the fluctuation trend of the pain response curve is analyzed;

[0009] According to the fluctuation trend, the nursing scheme is generated and output;

[0010] And, in the process of executing the care program, in response to subsequent image data collected, the limb recovery state is monitored, and recovery feedback information for adjusting the care program is output.

[0011] For example, the step of pre-generating the monitoring model comprises:

[0012] Image data of a patient's limb is collected, and a pain response value synchronized with the image data collection time of the patient's limb is determined;

[0013] First to-be-measured parameters and first to-be-measured features are extracted based on the image data of the patient's limb and the pain response value, respectively;

[0014] And, the first to-be-measured parameters and the first to-be-measured features are paired to form training data pairs, so as to generate the monitoring model by using a plurality of sets of the training data pairs.

[0015] For example, the step of extracting the first to-be-measured parameters comprises:

[0016] Key areas of the patient's limb are identified in the image data of the patient's limb, and the first to-be-measured parameters are extracted based on the key areas.

[0017] For example, the step of evaluating and outputting the pain risk value representing the current state comprises:

[0018] Second to-be-measured parameters extracted based on the current image data are input into the monitoring model, and the pain risk value is determined and output by calculating a similarity value between the second to-be-measured parameters and the first to-be-measured parameters contained in the monitoring model.

[0019] For example, the step of determining and outputting the pain risk value comprises:

[0020] The similarity value is compared with a preset decision threshold value;

[0021] And, when the similarity value is greater than or equal to the decision threshold value, a treatment period is calculated based on the similarity value, and a decision risk value is determined as the pain risk value according to the treatment period.

[0022] For example, the step of analyzing the fluctuation trend forming the pain response curve comprises:

[0023] A fluctuation amplitude and a fluctuation frequency are calculated based on the pain response curve;

[0024] And, when the fluctuation amplitude and the fluctuation frequency respectively satisfy the conditions of preset amplitude threshold value and frequency threshold value, the fluctuation trend is determined as a high fluctuation stage, so as to generate the care program based on the high fluctuation stage.

[0025] For example, the step of monitoring the limb recovery state comprises:

[0026] extracting a third to-be-measured parameter based on the subsequent image data;

[0027] and determining the limb recovery state by calculating an average similarity value between the third to-be-measured parameter and the first to-be-measured parameter contained in the monitoring model to generate the recovery feedback information.

[0028] For example, the step of adjusting the care plan comprises:

[0029] in response to the recovery feedback information, when the recovery feedback information represents an abnormal state, calculating an average pain risk value;

[0030] and when the average pain risk value is higher than a preset risk threshold, outputting a suppression care plan as the adjusted care plan;

[0031] or when the average pain risk value is lower than or equal to the risk threshold, outputting a maintenance care plan as the adjusted care plan.

[0032] The application also provides a rehabilitation care intelligent monitoring system based on image acquisition, comprising:

[0033] an image data acquisition module for acquiring current image data representing a patient state and subsequent image data in a care plan execution process;

[0034] a pain risk assessment module configured in response to the current image data acquired by the image data acquisition module, for evaluating and outputting a pain risk value based on a preset monitoring model, and analyzing a plurality of pain risk values generated within a preset time period to determine a fluctuation trend of a pain response curve;

[0035] a care plan generation module configured in response to the fluctuation trend determined by the pain risk assessment module, for generating and outputting a care plan;

[0036] and a recovery state monitoring module configured in response to the subsequent image data acquired by the image data acquisition module, for monitoring a limb recovery state and outputting recovery feedback information for adjusting the care plan by the care plan generation module.

[0037] For example, the care plan generation module is further configured to output a suppression care plan as the care plan when the recovery feedback information represents an abnormal state and the average pain risk value calculated based thereon is higher than a preset risk threshold. Advantages

[0038] 1) The present application collects image data of the patient's limb and determines the pain response value synchronized with the collection time, and then extracts the first to be measured parameter and the first to be measured feature to generate a monitoring model. In the monitoring stage, the model can output a quantitative pain risk value by comparing the second to be measured parameter with the first to be measured parameter contained in the model based on the newly collected current image data, thereby the present application establishes a direct quantitative correlation between objective limb visual representation and subjective pain perception, realizes objective evaluation of the patient's pain risk, and improves the reliability of monitoring.

[0039] 2) The present application records a plurality of pain risk values in a predetermined time period to form a pain response curve, and analyzes the fluctuation trend of the curve. By calculating the fluctuation amplitude and frequency and comparing with the corresponding threshold, the high fluctuation stage or low fluctuation stage can be identified to represent different recovery states, thereby the present application can grasp the recovery process of the patient from the macroscopic perspective of fluctuation trend, so that the generated nursing plan is more targeted and predictable, and active intervention can be made according to the dynamic evolution of the patient's state.

[0040] 3) The present application collects subsequent image data to evaluate the limb recovery state after executing the nursing plan, and outputs recovery feedback information, which will be directly used to guide the selection of subsequent nursing plan, for example, adaptive adjustment between inhibition nursing plan and maintenance nursing plan according to the limb recovery state and the calculated average pain risk value, realizing dynamic optimization and intelligent decision of rehabilitation nursing, ensuring that the nursing measures can continuously adapt to the actual recovery situation of the patient. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 is a method flowchart of the present application. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application will be further described in detail below combined with specific embodiments, it should be understood that the specific embodiments described here are only used to explain the present application, and are not used to limit the protection scope of the present application.

[0043] Embodiment one

[0044] Please refer to Figure 1 The present embodiment provides a rehabilitation nursing intelligent monitoring method based on image collection, which is used to objectively evaluate the pain risk value and limb recovery state of the patient in the rehabilitation process by non-contact method, and generates an adaptive nursing plan accordingly, the specific process includes the following steps:

[0045] Step S100: when monitoring the rehabilitation nursing of a patient, a pain risk value characterizing the current state is evaluated and output in response to real-time collected current image data of the patient and based on a pre-generated monitoring model;

[0046] Step S200: based on a plurality of said pain risk values recorded within a preset time period, a fluctuation trend of a pain response curve is analyzed;

[0047] Step S300: a nursing scheme is generated and output according to the fluctuation trend;

[0048] Step S400: in the process of executing the nursing scheme, the limb recovery state is monitored and recovery feedback information for adjusting the nursing scheme is output in response to collected subsequent image data.

[0049] Specifically, a monitoring model capable of associating the first to-be-measured parameter with the first to-be-measured feature is established, and image data of a patient's limb in rehabilitation training is continuously collected by an image collection device, such as a high-definition camera, which records visual information of a single frame or continuous multiple frames of the patient's limb in the rehabilitation training process, including the shape, position and dynamic changes of the patient's limb, and at the same time, the pain response of the patient is recorded synchronously through a wearable device or the guidance of medical staff, so as to determine the pain response value synchronized with the image data collection time of the patient's limb, thereby obtaining quantitative data representing the pain degree of the subjective feeling or physiological response of the patient.

[0050] The monitoring model is a calculation model for mapping the limb state parameters extracted from the image data into a quantitative pain risk value, and the specific formula is as follows:

[0051] ; in the formula, represents the pain risk value, which means the quantitative evaluation result of the model finally output, representing the possibility of pain occurrence under the current limb state; represents the current limb state vector (i.e. the first to-be-measured parameter or the second to-be-measured parameter), which means the feature vector describing the current comprehensive state of the patient's limb extracted in real time; represents the training sample vector, which means the th limb state parameter vector sample in the training data set, i.e. a specific first to-be-measured parameter; represents the kernel function, which means a function for calculating the similarity of two input vectors in a high-dimensional feature space, such as a Gaussian kernel function ; represents the Lagrange multiplier, which means the weight coefficient learned by the model through the training data, used to determine the contribution of each training sample to the final prediction result; represents the bias term, which means the bias parameter of the model, used to adjust the decision boundary; Support vector number, which means the number of training samples that determine the position of the decision boundary after model training.

[0052] Further, in order to enable the monitoring model to accurately learn the relationship between pain and limb performance, it is necessary to extract the first to-be-measured parameter and the first to-be-measured feature from the original data. The first to-be-measured parameter is calculated by extracting and fusing the training image data collected in the model training stage, which can comprehensively reflect the quantitative value of the dynamic and static features of the limb. The first to-be-measured feature is generated based on the pain response value, which can be used as a target label or feature vector for the monitoring model to learn.

[0053] On the one hand, based on the collected patient limb image data, the first to-be-measured parameter is extracted. Preferably, it is necessary to identify the key regions of the patient's limbs in the patient's limb image data, such as joint centers or muscle outlines, which are important for evaluating the rehabilitation state. The identification process can specifically include: in a pre-set scanning region, for example, according to the pre-defined limb position of an ergonomic model, setting the starting point in the region, such as the geometric center of the region, as the initial key point, based on the initial key point, and using prior knowledge of the limb skeletal structure to extract the pre-set position relationship for describing the relative direction and distance ratio between different key regions, such as the relative direction and distance ratio between the knee joint and the ankle joint; according to the position relationship, a series of possible joint positions are calculated from the initial key point to form a plurality of candidate regions which are a plurality of image sub-regions that may contain real key regions; through image feature matching or morphological analysis, the region with the highest confidence is selected from the plurality of candidate regions as the final key region, and a key region list is formed.

[0054] After identifying the key regions, the limb tremor amplitude of the key regions is determined by inter-frame difference method or optical flow method analysis from the patient's limb image data. This non-autonomous tremor is an important objective indicator of pain stress response, and the limb tremor amplitude can be a quantitative indicator representing the non-autonomous micro-vibration of the limb by analyzing the position changes of the key regions between consecutive image frames; other key parameters such as joint activity angle and movement speed are extracted in combination with the limb tremor amplitude; the key parameters are weighted and fused by a pre-set formula to obtain a calculation value that can comprehensively reflect the state of the limb, and the calculation value is taken as the first to-be-measured parameter. The purpose of the pre-set formula is to normalize parameters of different physical units and dimensions, and to give higher weights to parameters that can better reflect the pain state, so as to generate a single calculation value that can comprehensively reflect the state of the limb.

[0055] On the other hand, based on the pain response value fed back by the patient, the first to-be-tested feature is extracted, which can specifically include: through the interactive device or artificial inquiry, the pain onset reported by the patient is accurately recorded, based on the time stamp of the pain onset, the pain onset time and the pain recovery period from the peak to the baseline are calculated, and the length of the pain recovery period reflects the adjustment ability of the nervous system of the patient; according to the calculated pain onset time and pain recovery period, the pain response value is comprehensively determined, for example, quantified as a numerical value in a specific range; based on the pain response value, the first to-be-tested feature for model learning is generated, which can be the numerical value itself or a vector after encoding.

[0056] The first to-be-tested parameter extracted above is time-stamped and paired with the first to-be-tested feature to form a training data pair, which is a data pair composed of one first to-be-tested parameter and one first to-be-tested feature aligned in time, used as a basic unit for training the monitoring model. By collecting multiple sets of training data pairs at different rehabilitation stages, using machine learning algorithms such as support vector machines, neural networks or gradient boosting trees, a monitoring model capable of predicting pain risk values based on the first to-be-tested parameters is trained and generated.

[0057] Further, in the rehabilitation care process, enter the real-time monitoring mode, collect the current image data of the patient, and extract the second to-be-tested parameter based on the current image data. The second to-be-tested parameter is a quantitative numerical value reflecting the current state of the limb extracted from the current image data using the same method as extracting the first to-be-tested parameter in the real-time monitoring stage.

[0058] Further, the second to-be-tested parameter is input into the monitoring model, which outputs the pain risk value by comparing the input second to-be-tested parameter with the first to-be-tested parameter contained in the monitoring model in high-dimensional space. Specifically, it can include: calculating the similarity value between the second to-be-tested parameter and the first to-be-tested parameter contained in the monitoring model, for example, by calculating the Euclidean distance or cosine similarity; comparing the similarity value with a preset judgment threshold, which is set to ensure the reliability of the prediction, if the similarity value is less than the judgment threshold, it means that the current state of the limb has low correlation with all known pain states, at this time, output the preset default risk value as the pain risk value, the default risk value usually represents a low risk or unknown risk pain risk value; if the similarity value is greater than or equal to the judgment threshold, then calculate the treatment period based on the similarity value, the treatment period can be used to guide the duration of subsequent treatment or rehabilitation actions, and determine and output the judgment risk value as the pain risk value according to the treatment period.

[0059] In order to macroscopically grasp the change rule of the pain of the patient, a plurality of pain risk values generated in a preset time period are continuously recorded, and the values are connected in time sequence to form a pain response curve for intuitively displaying the change trend of the pain risk.

[0060] Further, the fluctuation trend of the pain response curve is analyzed, and a nursing scheme is output based on the analysis result. The fluctuation trend analysis can analyze the change mode of the pain response curve in a specific time period, mainly focusing on the degree and speed of the change. Specifically, the fluctuation amplitude and the fluctuation frequency of the pain response curve in the preset time period are calculated based on the pain response curve by a signal processing algorithm. The fluctuation amplitude reflects the change amplitude of the pain degree, and the fluctuation frequency reflects the rate of the change of the pain. The calculated fluctuation amplitude is compared with a preset amplitude threshold, and the fluctuation frequency is compared with a preset frequency threshold. When the fluctuation amplitude meets the condition of the amplitude threshold and the fluctuation frequency meets the condition of the frequency threshold, the current fluctuation trend is determined as a high fluctuation stage, which usually means that the patient is in an unstable pain period. In other cases, the fluctuation trend is determined as a low fluctuation stage, which indicates that the pain state is relatively stable. The step of outputting the nursing scheme is performed based on the judgment of the high fluctuation stage or the low fluctuation stage, so as to realize the dynamic adjustment of the nursing strategy.

[0061] In the process of executing the nursing scheme output by the system, in order to form a closed-loop feedback system, it is necessary to continuously track the nursing effect, to continue to collect subsequent image data, and to extract a third to-be-measured parameter from the subsequent image data based on the subsequent image data by using the same method as the above steps. The third to-be-measured parameter is a quantitative value extracted from the subsequent image data in the effect evaluation stage after the execution of the nursing scheme by using the same method as the extraction of the first to-be-measured parameter.

[0062] In order to judge the recovery state of the limb, an average similarity value between the third to-be-measured parameter and the first to-be-measured parameters contained in the monitoring model is calculated. The average similarity value can macroscopically reflect the overall matching degree between the current limb state and all known states in the model. According to the average similarity value, the recovery state of the limb is judged, and recovery feedback information is output, which is information representing whether the limb is currently in a normal recovery state or an abnormal recovery state. For example, the average similarity value can be compared with a preset reference threshold. If the average similarity value is higher than the reference threshold, the recovery feedback information represents the normal recovery state. If the average similarity value is lower than or equal to the reference threshold, the recovery feedback information represents the abnormal recovery state. It should be noted that the average similarity value is an average value obtained by calculating the similarity between the third to-be-measured parameter and all first to-be-measured parameters contained in the monitoring model, which is used to macroscopically evaluate the overall matching degree between the current limb state and all states learned by the model.

[0063] In response to the recovery feedback information, decision and scheme optimization are performed. If the recovery feedback information represents a non-recovery normal state, the average pain risk value in the monitoring period is calculated. If the recovery feedback information represents a recovery normal state, the current care scheme is continued to be executed, and the subsequent average pain risk value is recorded. The average pain risk is a value obtained by averaging a plurality of pain risk values output by the system, and is used to reflect the overall pain risk level in the period. The calculated or recorded average pain risk value is compared with a preset risk threshold. The risk threshold is an acceptable upper limit of pain in the rehabilitation plan. If the average pain risk value is higher than the risk threshold, it indicates that the current care scheme may be too aggressive, and the inhibited care scheme is output as the care scheme, for example, suggesting to reduce the training intensity. If the average pain risk value is lower than or equal to the risk threshold, it indicates that the current scheme is safe and effective, and the care scheme is output as the care scheme.

[0064] The embodiment can evaluate the pain risk value and the limb recovery state of the patient, and can generate a care scheme according to the feedback of the evaluation result, form a closed-loop management process in actual application, and effectively improve the execution quality and safety of rehabilitation care through the execution of the intervention feedback and the multiple data calibration.

[0065] Embodiment two

[0066] The embodiment provides a rehabilitation care intelligent monitoring system based on image acquisition. The system is used to execute the rehabilitation care intelligent monitoring method based on image acquisition described above. The system can objectively evaluate the pain risk value by analyzing the image data of the patient in real time, and automatically generate and adjust the care scheme based on the fluctuation trend of the pain response curve, so as to realize the intelligent closed-loop management process of the rehabilitation process. The system can be divided into the following modules which work cooperatively:

[0067] An image data acquisition module is used to execute the acquisition task of image data. At the beginning of monitoring, the module is responsible for acquiring current image data representing the state of the patient in real time, for example, capturing static or dynamic images of a specific limb or face of the patient during rehabilitation activities. During the execution of the care scheme, the module continuously acquires subsequent image data for subsequent recovery state monitoring. The image data can be a two-dimensional color image or a three-dimensional image data containing depth information.

[0068] A pain risk evaluation module is responsible for converting the original image data into a quantitative pain risk value and analyzing the dynamic change. The specific work flow is as follows:

[0069] receiving current image data provided by the image data acquisition module, extracting a second to-be-measured parameter from the image data, such as a visual indicator related to pain including facial micro-expression, limb muscle tension, joint activity angle, etc. identified by a computer vision algorithm.

[0070] inputting the extracted second to-be-measured parameter into a pre-generated monitoring model, which internally stores a large amount of reference data, i.e. the mapping relationship between the first to-be-measured parameter (derived from a large amount of patient limb image data) and the first to-be-measured feature (derived from pain response values collected synchronously with the image) established in the model training phase, and the monitoring model evaluates the degree of association between the current state and the historical pain state by calculating the similarity value between the input second to-be-measured parameter and the first to-be-measured parameter stored in the model.

[0071] In order to improve the accuracy of the evaluation, the calculated similarity value is compared with a preset decision threshold, and when the similarity value is greater than or equal to the decision threshold, it indicates that the current state is highly similar to a certain known pain state, and further based on the similarity value, a treatment period is calculated, and a decision risk value is determined according to the treatment period, which is finally output as a pain risk value representing the current state.

[0072] Within a preset time period, a plurality of output pain risk values are recorded continuously, and based on these data points, a pain response curve is drawn, by analyzing the curve, the fluctuation amplitude and frequency are calculated, when the calculated fluctuation amplitude and frequency meet the preset amplitude threshold and frequency threshold conditions, for example, the fluctuation is intense and frequent, the current fluctuation trend is determined as a high fluctuation stage, and the judgment result is transmitted to the care plan generation module.

[0073] The care plan generation module is responsible for generating and adjusting the care plan according to the analysis results, and its functions include:

[0074] initial scheme generation, in response to the fluctuation trend determined by the pain risk assessment module, specifically, when receiving the signal of "high fluctuation stage", a targeted care plan is generated and output according to the preset care knowledge base or rule engine, for example, the plan may suggest suspending or reducing the intensity of the current rehabilitation activity, reminding the nursing staff to intervene manually or suggesting to take specific relaxation measures.

[0075] The scheme is dynamically adjusted. During the execution of the nursing scheme, recovery feedback information is received from the recovery state monitoring module. When the recovery feedback information indicates that the patient's limb is not in a normal state, an average pain risk value is calculated based on a series of recent pain risk values. The average pain risk value is compared with a preset risk threshold. If the average pain risk value is higher than the risk threshold, it indicates that the current nursing scheme is not effective or the patient's pain is increasing, and an adjusted inhibition nursing scheme is output. For example, the activity intensity is further reduced or the activity is completely stopped. Otherwise, if the average pain risk value is lower than or equal to the risk threshold, it indicates that the current scheme is controllable, and a maintenance nursing scheme is output, that is, the current scheme is continued to be executed and monitored.

[0076] The recovery state monitoring module is responsible for continuously tracking and evaluating the recovery state of the patient's limb during the execution of the nursing scheme. The workflow is as follows:

[0077] The third to-be-measured parameter is extracted based on the subsequent image data collected by the image data collection module during the execution of the nursing scheme. The parameter is consistent with the second to-be-measured parameter in type and dimension.

[0078] The overall closeness between the current limb state and the various reference states recorded in the model is quantitatively evaluated by calculating the average similarity value between the third to-be-measured parameter and the first to-be-measured parameter contained in the monitoring model. This average similarity value can comprehensively reflect whether the limb is tending towards a normal pain-free state or still showing pain-related visual features.

[0079] Based on the calculated average similarity value, the current limb recovery state is determined, and corresponding recovery feedback information (e.g., normal state or not normal state) is generated and sent to the nursing scheme generation module as the basis for adjusting the nursing scheme.

[0080] In addition, in order to enable the above-mentioned system to operate effectively, the monitoring model needs to be generated in advance. The generation process is usually completed offline before the system is deployed, which specifically includes: collecting a large amount of patient limb image data, and simultaneously acquiring and recording a pain response value through a standard pain assessment scale (such as NRS digital scoring method) or a physiological sensor while collecting each frame of image data; identifying key areas (such as joints, around the wound) in the collected patient limb image data, and extracting the first to-be-measured parameter based on these key areas, and extracting the first to-be-measured feature based on the corresponding pain response value; pairing the first to-be-measured parameter with the first to-be-measured feature to form a training data pair, using multiple groups of such training data pairs, and training through a machine learning algorithm (such as support vector machine, neural network) to finally generate a monitoring model that can accurately reflect the relationship between image parameters and pain degree.

[0081] The application can objectively and continuously monitor the pain risk value of the patient through the cooperative work of the image data acquisition module, the pain risk assessment module, the nursing scheme generation module and the recovery state monitoring module, avoids the delay and deviation of subjective report, and intelligently generates and adjusts the nursing scheme according to the fluctuation trend of the pain response curve, thereby improving the scientificity and individualization level of rehabilitation nursing, and being suitable for postoperative rehabilitation, chronic pain management and nursing scene of patients with limited communication ability.

[0082] The above only describes the preferred embodiments of the application and is not intended to limit the application. Various modifications and changes can be made to the application by those skilled in the art, and any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the application shall be included in the protection scope of the application.

Claims

1. An image acquisition-based intelligent monitoring method for rehabilitation care, characterized in that, The method comprises the following steps: when monitoring the rehabilitation nursing of a patient, a pain risk value representing a current state is evaluated and output in response to real-time collected current image data of the patient and based on a pre-generated monitoring model; a fluctuation trend of a pain response curve is analyzed based on a plurality of the pain risk values recorded within a preset time period; a nursing scheme is generated and output according to the fluctuation trend; and, in the process of executing the nursing scheme, a limb recovery state is monitored in response to collected subsequent image data, and recovery feedback information for adjusting the nursing scheme is output; the step of pre-generating the monitoring model comprises: collecting image data of a patient's limb and determining a pain response value synchronized with the image data collection time of the patient's limb; a first to-be-measured parameter and a first to-be-measured feature are extracted based on the image data of the patient's limb and the pain response value, respectively; and, the first to-be-measured parameter and the first to-be-measured feature are paired to form a training data pair, so as to generate the monitoring model by using a plurality of the training data pairs; the step of evaluating and outputting the pain risk value representing the current state comprises: a second to-be-measured parameter extracted based on the current image data is input into the monitoring model, and the pain risk value is determined and output by calculating a similarity value between the second to-be-measured parameter and the first to-be-measured parameter contained in the monitoring model; the step of analyzing the fluctuation trend of the pain response curve comprises: a fluctuation amplitude and a fluctuation frequency are calculated based on the pain response curve; and, when the fluctuation amplitude and the fluctuation frequency respectively satisfy the conditions of preset amplitude threshold and frequency threshold, the fluctuation trend is determined as a high fluctuation stage, so as to generate the nursing scheme based on the high fluctuation stage. 2.The image collection-based rehabilitation nursing intelligent monitoring method according to claim 1, characterized in that, the step of extracting the first to-be-measured parameter comprises: a key area of the patient's limb is identified in the image data of the patient's limb, and the first to-be-measured parameter is extracted based on the key area. 3.The image collection-based rehabilitation nursing intelligent monitoring method according to claim 2, characterized in that, the step of determining and outputting the pain risk value comprises: the similarity value is compared with a preset judgment threshold; and, when the similarity value is greater than or equal to the judgment threshold, a treatment period is calculated based on the similarity value, and a judgment risk value is determined as the pain risk value according to the treatment period. 4.The image collection-based rehabilitation nursing intelligent monitoring method according to claim 1, characterized in that, the step of monitoring the limb recovery state comprises: a third to-be-measured parameter is extracted based on the subsequent image data; and, the limb recovery state is judged by calculating an average similarity value between the third to-be-measured parameter and the first to-be-measured parameter contained in the monitoring model, so as to generate the recovery feedback information.

5. The intelligent monitoring method for rehabilitation nursing based on image acquisition according to claim 1, characterized in that, the step of adjusting the nursing scheme comprises: in response to the recovery feedback information, when the recovery feedback information represents an un-recovered normal state, an average pain risk value is calculated; and, when the average pain risk value is higher than a preset risk threshold, an inhibition nursing scheme is output as the adjusted nursing scheme; or, when the average pain risk value is lower than or equal to the risk threshold, a maintenance nursing scheme is output as the adjusted nursing scheme.

6. An image acquisition-based intelligent monitoring system for rehabilitation care, characterized in that, The method comprises: an image data acquisition module configured to acquire current image data representing a state of a patient and subsequent image data during execution of a care plan; a pain risk assessment module configured to evaluate and output a pain risk value based on a preset monitoring model in response to the current image data acquired by the image data acquisition module, and analyze a plurality of the pain risk values generated within a preset time period to determine a fluctuation trend of a pain response curve; a care plan generation module configured to generate and output a care plan in response to the fluctuation trend determined by the pain risk assessment module; and a recovery state monitoring module configured to monitor a limb recovery state in response to the subsequent image data acquired by the image data acquisition module, and output recovery feedback information for adjusting the care plan by the care plan generation module.

7. The intelligent monitoring system for rehabilitation care based on image acquisition according to claim 6, characterized in that, The care plan generation module is further configured to output a suppressive care plan as the care plan when the recovery feedback information represents an abnormal state and a mean pain risk value calculated based thereon is higher than a preset risk threshold.

Images