A remote intelligent monitoring system and method for bone injury rehabilitation

Abstract

This application provides a remote intelligent monitoring system and method for bone injury rehabilitation. The method acquires bone injury information of a target user and determines multiple skeletal rehabilitation observation points based on this information. It collects continuous image data through a user terminal and extracts skeletal posture information from the continuous image data. Based on the skeletal posture information, it extracts the posture contrast of each skeletal rehabilitation observation point. It identifies the rehabilitation trend of the posture contrast at each skeletal rehabilitation observation point and performs observability analysis based on the rehabilitation trends of each skeletal rehabilitation observation point to determine the rehabilitation monitoring weight of that skeletal rehabilitation observation point. Based on the rehabilitation monitoring weight, it identifies the rehabilitation progress of the target user and sends the rehabilitation progress and personalized rehabilitation strategies to the user terminal. This system can identify rehabilitation progress based on the posture contrast of skeletal observation points, improving the safety of bone injury rehabilitation care.

Description

Technical Field

[0001] This application relates to the field of bone injury rehabilitation nursing technology, and more specifically, to a remote intelligent monitoring system and method for bone injury rehabilitation. Background Technology

[0002] The purpose of rehabilitation training is to relieve pain, improve joint mobility, strengthen muscles, promote functional recovery, accelerate wound healing, effectively prevent secondary injuries caused by post-fracture sequelae, and help patients return to a normal life as soon as possible. With the rapid advancement of science and technology, the application of orthopedic rehabilitation training has become increasingly widespread and has become one of the important strategies for functional recovery after fractures and injuries. Especially for the recovery of limb function, orthopedic rehabilitation training has shown significant effectiveness and has effectively reduced the incidence of post-injury complications.

[0003] In existing technologies, bone injury rehabilitation monitoring is mainly conducted through regular hospital examinations. This method relies on patients coming to the hospital for imaging examinations and clinical assessments, which not only consumes a lot of time and medical resources, but also, for bone injury patients with limited mobility, frequent trips to the hospital may increase the risk of secondary injury and affect the continuity of rehabilitation. Furthermore, traditional rehabilitation assessment methods are based on manual test results at discrete time points, making it difficult to achieve continuous tracking and dynamic analysis of the rehabilitation process. This makes it difficult to capture subtle changes in the rehabilitation process in a timely manner, reducing the timeliness and accuracy of rehabilitation intervention. Therefore, how to monitor the rehabilitation process in a timely manner during bone injury rehabilitation has become an urgent problem to be solved. Summary of the Invention

[0004] This application provides a remote intelligent monitoring system and method for bone injury rehabilitation, which can identify rehabilitation progress based on the posture contrast of bone observation points, thereby improving the safety of bone injury rehabilitation nursing.

[0005] In a first aspect, this application provides a remote intelligent monitoring method for bone injury rehabilitation. This method can be executed by a network device, or by a chip configured in the network device, and this application does not limit the execution of such method.

[0006] Specifically, the method includes: Obtain bone injury information of the target user, and determine multiple skeletal rehabilitation observation points based on the bone injury information; Continuous image data is collected through a user terminal, and skeleton pose information is extracted from the continuous image data. For any skeletal rehabilitation observation point, extract the posture contrast of the skeletal rehabilitation observation point based on the skeletal posture information. Identify the rehabilitation trend of the posture contrast at the skeletal rehabilitation observation point, and perform observability analysis based on the rehabilitation trend of each skeletal rehabilitation observation point to determine the rehabilitation monitoring weight of the skeletal rehabilitation observation point. The rehabilitation progress of the target user is identified based on the rehabilitation monitoring weight, and the rehabilitation progress and personalized rehabilitation strategy are sent to the user terminal.

[0007] In conjunction with the first aspect, in certain implementations of the first aspect, extracting skeleton pose information from the continuous image data specifically includes: The continuous image data is acquired, and the skeletal key points are extracted from the continuous image data using a pre-trained human pose recognition model to obtain the skeletal key point coordinate information corresponding to multiple continuous images respectively. For any consecutive image, construct the skeleton topology based on the coordinate information of the skeletal key points corresponding to the consecutive image, and perform temporal correlation processing on the skeleton topology of adjacent frames according to the time frames corresponding to the consecutive images to obtain the skeleton pose information.

[0008] In conjunction with the first aspect, in certain implementations of the first aspect, extracting the posture contrast of the skeletal rehabilitation observation point based on the skeletal posture information specifically includes: Obtain the coordinate sequence of the skeletal key points corresponding to the skeletal rehabilitation observation points in the skeletal posture information; Action type identification is performed based on the coordinate sequence of the skeletal key points, and the same type of coordinate sequence is extracted from the coordinate sequence of symmetrical skeletal points based on the action type; The coordinate sequences of the skeletal key points and the symmetrical skeletal points are temporally aligned, and the spatial difference value of the corresponding joints is calculated as the pose contrast for the aligned skeletal key point coordinate data.

[0009] In conjunction with the first aspect, in some implementations of the first aspect, identifying the rehabilitation trend of the posture contrast of the skeletal rehabilitation observation point specifically includes: obtaining the skeletal posture information of the historical records of the skeletal rehabilitation observation point; extracting the posture contrast information corresponding to the skeletal rehabilitation observation point in each rehabilitation cycle window based on a preset rehabilitation cycle window; and constructing a rehabilitation trend based on the timestamps corresponding to each rehabilitation cycle window.

[0010] In conjunction with the first aspect, in certain implementations of the first aspect, observability analysis is performed based on the rehabilitation trends of each skeletal rehabilitation observation point to determine the rehabilitation monitoring weights of the skeletal rehabilitation observation points. Specifically, this includes: Multiple skeletal rehabilitation observation points are divided into regions to identify the set of reference observation points in the bone injury area and the set of related observation points in the injury-related area; For any reference observation point, the intensity of the injury is determined based on the recovery trend of that reference observation point. For any associated observation point, determine the adjacent bone injury area of ​​the associated observation point; determine the associated rehabilitation trend and average injury intensity of multiple reference observation points within the adjacent bone injury area; A correlation analysis is performed between the rehabilitation trend of the associated observation point and the associated rehabilitation trend of the adjacent bone injury area to determine the trend correlation; the rehabilitation monitoring weight of the associated observation point is determined based on the trend correlation and the average injury intensity.

[0011] In conjunction with the first aspect, in some implementations of the first aspect, identifying the rehabilitation progress of the target user based on the rehabilitation monitoring weight specifically includes: Obtain the rehabilitation trend corresponding to each skeletal rehabilitation observation point, and predict the rehabilitation trend value for the next time period based on the rehabilitation trend prediction model. The prediction results are weighted and fused according to the rehabilitation monitoring weights corresponding to each skeletal rehabilitation observation point to determine the predicted value of the rehabilitation trend of the target user. Based on the predicted rehabilitation trend values, a rehabilitation trend mapping is performed to determine the rehabilitation progress level, and a personalized rehabilitation strategy is generated based on the rehabilitation progress level.

[0012] In conjunction with the first aspect, in some implementations of the first aspect, before extracting skeleton pose information from the continuous image data, the method further includes: performing image preprocessing on the continuous image data.

[0013] Secondly, this application provides a remote intelligent monitoring system for bone injury rehabilitation, which includes a rehabilitation monitoring unit, the rehabilitation monitoring unit comprising: The user information acquisition module is used to acquire the bone injury information of the target user and determine multiple skeletal rehabilitation observation points based on the bone injury information. The image acquisition module is used to acquire continuous image data through a user terminal and extract skeleton pose information from the continuous image data; The posture recognition module extracts the posture contrast of any skeletal rehabilitation observation point based on the skeletal posture information. The posture recognition module is also used to identify the rehabilitation trend of the posture contrast of the skeletal rehabilitation observation point, and to perform observability analysis based on the rehabilitation trend of each skeletal rehabilitation observation point to determine the rehabilitation monitoring weight of the skeletal rehabilitation observation point. The rehabilitation strategy generation module is used to identify the rehabilitation progress of the target user based on the rehabilitation monitoring weight, and send the rehabilitation progress and personalized rehabilitation strategy to the user terminal.

[0014] Thirdly, this application provides a computer terminal device, which includes a memory and a processor. The memory stores code, and the processor is configured to acquire the code and execute the aforementioned remote intelligent monitoring method for bone injury rehabilitation.

[0015] Fourthly, this application provides a computer-readable storage medium storing at least one computer program, which is loaded and executed by a processor to perform the operations described above in a remote intelligent monitoring method for bone injury rehabilitation.

[0016] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: This application provides a remote intelligent monitoring system and method for bone injury rehabilitation. First, bone injury information of the target user is acquired, and multiple skeletal rehabilitation observation points are determined based on this information. Continuous image data is collected through a user terminal, and skeletal posture information is extracted from the continuous image data. Posture contrast of the skeletal rehabilitation observation point is extracted based on the skeletal posture information. The rehabilitation trend of the posture contrast of the skeletal rehabilitation observation point is identified, and observability analysis is performed based on the rehabilitation trends of each skeletal rehabilitation observation point to determine its rehabilitation monitoring weight. The rehabilitation progress of the target user is identified based on the rehabilitation monitoring weight, and the rehabilitation progress and personalized rehabilitation strategies are sent to the user terminal.

[0017] Therefore, this application introduces the concept of skeletal rehabilitation observation points, which structurally express the motion changes of different skeletal structural nodes and use posture contrast as a quantitative indicator to characterize the functional recovery of different skeletal nodes during the rehabilitation process. Time series analysis of the posture contrast of each skeletal rehabilitation observation point is performed to obtain continuous rehabilitation trend information, thereby realizing dynamic tracking of the rehabilitation process. A rehabilitation monitoring weight mechanism is introduced, and the degree of influence of different observation points in the rehabilitation process is distinguished by the observability analysis of the rehabilitation trends of different skeletal rehabilitation observation points. Based on the weighted multi-observation point rehabilitation trend fusion results, the overall rehabilitation status of the target user can be uniformly quantitatively expressed, which can reflect the overall functional recovery of the bones, reduce the intermittency and lag of traditional manual review methods, identify rehabilitation risks and intervene in a timely manner, and improve the monitoring timeliness of the bone injury rehabilitation nursing process.

[0018] In summary, this application enables rehabilitation progress recognition based on the posture contrast of skeletal observation points, thereby improving the safety of bone injury rehabilitation nursing. Attached Figure Description

[0019] Figure 1 This is an exemplary flowchart of a remote intelligent monitoring method for bone injury rehabilitation, according to some embodiments of this application; Figure 2 This is a schematic diagram of the structure of a rehabilitation monitoring unit according to some embodiments of this application; Figure 3 This is a schematic diagram of the structure of a computer terminal device that implements a remote intelligent monitoring method for bone injury rehabilitation, according to some embodiments of this application. Detailed Implementation

[0020] This application obtains the bone injury information of the target user and determines multiple skeletal rehabilitation observation points based on the bone injury information; it collects continuous image data through the user terminal and extracts skeletal posture information from the continuous image data; it extracts the posture contrast of the skeletal rehabilitation observation point based on the skeletal posture information; it identifies the rehabilitation trend of the posture contrast of the skeletal rehabilitation observation point, and performs observability analysis based on the rehabilitation trend of each skeletal rehabilitation observation point to determine the rehabilitation monitoring weight of the skeletal rehabilitation observation point; it identifies the rehabilitation progress of the target user based on the rehabilitation monitoring weight and sends the rehabilitation progress and personalized rehabilitation strategy to the user terminal. It can identify the rehabilitation progress based on the posture contrast of the skeletal observation point, thus improving the safety of bone injury rehabilitation care.

[0021] To better understand the above technical solutions, a detailed description of the solutions will be provided below in conjunction with the accompanying drawings and specific implementation methods. (Reference) Figure 1 The figure is an exemplary flowchart of a remote intelligent monitoring method for bone injury rehabilitation according to some embodiments of this application. The remote intelligent monitoring method 100 for bone injury rehabilitation mainly includes the following steps: In step S101, bone injury information of the target user is obtained, and multiple skeletal rehabilitation observation points are determined based on the bone injury information.

[0022] In some embodiments, the bone injury information includes at least: anatomical location information identifying the bone injury site of the target user, wherein the anatomical location information includes the fracture location, the postoperative fixation area, and the spatial location information corresponding to the damaged joint.

[0023] It should be noted that the bone injury information is used to characterize the location and extent of bone injuries of the target user. The bone injury information includes, but is not limited to, at least one of medical imaging data, electronic medical record data, doctor annotation information, and user terminal input information. The corresponding analysis method is adopted according to the specific form of the bone injury information. For example, when the bone injury information is medical imaging data, the bone structure in the image is segmented and located by an image recognition model to extract the spatial location information of the fracture area. When the bone injury information is text-based medical record data, doctor annotation information, or other text information types, the anatomical descriptions in the text are semantically parsed by a natural language processing model to determine the corresponding standard human anatomical location and map it to the corresponding bone node in the bone model. This application will not elaborate further on this.

[0024] Preferably, in some embodiments, determining multiple skeletal rehabilitation observation points based on the bone injury information specifically includes: The affected area of ​​the bone injury is determined based on the bone injury information; Based on the pre-constructed human skeletal model, a subset of skeletal structures associated with the bone injury area is determined, and the associated injured area is extracted based on the subset of skeletal structures. Extract multiple key skeletal points from the injured bone region and the associated region, identify symmetrical skeletal points of the key skeletal points, and use the key skeletal points and their corresponding symmetrical skeletal points as observation points for bone rehabilitation.

[0025] In specific implementation, the bone injury area is the fixed anatomical space range after the target user's bone is damaged. For example, when the bone injury information is medical imaging data, the bone structure is identified by an image segmentation model, and the fracture area, bone crack area and the area adjacent to the postoperative implant are extracted and used as the bone injury area.

[0026] In the process of determining the subset of skeletal structures associated with the injured area based on the pre-constructed human skeletal model, the human skeletal model is a graph structure model constructed based on the connection relationship between joint nodes and bones, where nodes represent key points of bones and edges represent bone connection relationships. Based on the spatial adjacency relationship of the injured area in the skeletal model, the subset of skeletal structures is expanded according to a preset neighborhood expansion rule. The neighborhood expansion rule includes taking the injured area as the central node and expanding the associated skeletal nodes step by step according to the second-order neighborhood to obtain the associated injured area.

[0027] In some embodiments, multiple skeletal key points are extracted from the bone injury area and the associated injury area. The skeletal key points include articular nodes and their endpoint nodes connected to the bones. In the process of identifying the symmetrical skeletal points of the skeletal key points, the symmetrical skeletal points are determined by establishing a mapping relationship based on the left and right symmetrical anatomical structure of the human body, and matching the corresponding left and right nodes in the skeletal topology one by one. It is foreseeable that when the left and right skeletal structures are occluded or there are special circumstances, such as the absence of detection of some organs, the corresponding symmetrical skeletal points are determined by an interpolation method based on the geometric constraints of the skeletal structure.

[0028] In step S102, continuous image data is collected through the user terminal, and skeleton pose information is extracted from the continuous image data.

[0029] In some embodiments, continuous image data is obtained by continuously acquiring images through a high-definition camera mounted on the user terminal. The continuous image data includes multiple images of the user's posture and corresponding time tags. In some embodiments, the continuous image data can be acquired by the user performing a preset rehabilitation movement in front of the user terminal. This application will not elaborate further on this.

[0030] Optionally, in some embodiments, before extracting skeleton pose information from the continuous image data, the method further includes: performing image preprocessing on the continuous image data.

[0031] In specific implementation, image preprocessing of the continuous image data includes: image denoising, size normalization, and inter-frame sampling filtering. The denoising process uses median filtering to suppress high-frequency noise in the image, reducing noise interference during the acquisition process, and normalizes the pixel values ​​of the continuous images after scaling them to a preset standard size. Inter-frame sampling filtering is performed on the continuous image sequence, and effective keyframes are selected from the continuous frames by setting a time interval threshold. The filtered image data is then subjected to histogram equalization.

[0032] In some other embodiments, the preprocessing of continuous image data can also include human region detection and background suppression. By cropping the human region based on the target detection model, only the image portion containing the human skeleton region is retained, thereby reducing background interference in the skeleton pose extraction process.

[0033] Preferably, in some embodiments, extracting skeleton pose information from the continuous image data specifically includes: The continuous image data is acquired, and the skeletal key points are extracted from the continuous image data using a pre-trained human pose recognition model to obtain the skeletal key point coordinate information corresponding to multiple continuous images respectively. For any consecutive image, construct the skeleton topology based on the coordinate information of the skeletal key points corresponding to the consecutive image, and perform temporal correlation processing on the skeleton topology of adjacent frames according to the time frames corresponding to the consecutive images to obtain the skeleton pose information.

[0034] In specific implementation, the continuous image data is acquired, and human target detection is performed on each frame of the image to determine the region of interest (ROI) containing the human body, thereby reducing the interference of background information on the detection of skeleton key points. The cropped ROI image is then input into a pre-trained human pose recognition model. The human pose recognition model can be one of the key point detection models commonly used in the prior art for human skeleton point feature extraction, which is based on convolutional neural networks, HRNet structures, or Transformer structures. It is used to extract features and regress key points of the human skeletal structure. This application does not limit the specific type of human pose recognition model. In the process of detecting key points in each frame of an image using the human pose recognition model to obtain the skeletal key point coordinate information of the corresponding frame, the skeletal key point coordinate information includes, but is not limited to, the two-dimensional or three-dimensional coordinates of the shoulder joint, elbow joint, wrist joint, hip joint, knee joint, and ankle joint. In some embodiments, the skeletal key point coordinate information is subjected to confidence screening to remove key point data with a confidence level lower than a preset threshold, and missing key points are repaired using neighborhood interpolation as a completion method. For any consecutive image frame, a skeleton topology is constructed based on the skeletal key point coordinate information corresponding to that frame. The skeleton topology is defined by preset skeletal connection relationships, such as establishing joint connection edges based on human anatomy, including skeletal connection relationships such as shoulder-elbow-wrist and hip-knee-ankle, forming a graph-structured human skeleton model.

[0035] In some implementations, the skeleton topology is represented as a weighted graph structure, where nodes represent skeletal key points, edges represent joint connections, and weights are assigned to edges based on changes in joint distance and angle to characterize the degree of motion change of different joints. Further, temporal correlation processing is performed on the skeleton topology of adjacent frames based on the time frames corresponding to consecutive images. Specifically, key points are indexed and matched based on inter-frame timestamps, and Euclidean distance is used to correlate the trajectories of key points in adjacent frames to form a continuous sequence of skeletal motion trajectories. In a specific implementation, a moving average can also be performed based on a preset time window to achieve temporal smoothing of the skeletal motion trajectory sequence, obtaining skeleton pose information containing time dimension information. The skeleton pose information includes the coordinates of each skeletal key point in the skeletal motion trajectory sequence and their corresponding time labels.

[0036] In step S103, for any skeletal rehabilitation observation point, the posture contrast of the skeletal rehabilitation observation point is extracted based on the skeletal posture information.

[0037] It should be noted that the skeletal rehabilitation observation points are observation point pairs consisting of key skeletal points in the bone injury area and corresponding symmetrical skeletal points on the human anatomy. By extracting the posture comparison information of the key skeletal points and symmetrical skeletal points in the same type of movement and performing posture difference analysis, the rehabilitation status of the bone injury area can be observed and quantified. That is, this application identifies the rehabilitation progress of the bone injury area by comparing the movement posture between the bone injury area and the symmetrical healthy area.

[0038] Preferably, in some embodiments, extracting the posture contrast of the skeletal rehabilitation observation point based on the skeletal posture information specifically includes: Obtain the coordinate sequence of the skeletal key points corresponding to the skeletal rehabilitation observation points in the skeletal posture information; Action type identification is performed based on the coordinate sequence of the skeletal key points, and the same type of coordinate sequence is extracted from the coordinate sequence of symmetrical skeletal points based on the action type; The coordinate sequences of the skeletal key points and the symmetrical skeletal points are temporally aligned, and the spatial difference value of the corresponding joints is calculated as the pose contrast for the aligned skeletal key point coordinate data.

[0039] In specific implementation, the coordinate sequence of skeletal key points corresponding to the skeletal rehabilitation observation point in the skeletal posture information is obtained. The coordinate sequence of skeletal key points is three-dimensional spatial coordinate data extracted from continuous image frames and sorted according to time labels to form a continuous motion trajectory. Action type recognition is performed based on the coordinate sequence of skeletal key points. The action type recognition includes classifying the coordinate change sequence within a continuous time window according to the velocity change characteristics to determine the action type of the current skeletal movement. In some other embodiments, the joint angular velocity change rate can also be extracted based on the nearby skeletal key points for rule classification. It is foreseeable that action type recognition can also be performed through a pre-trained action recognition model. This application does not limit this.

[0040] In this embodiment, after determining the action type, the symmetrical bone point coordinate sequence corresponding to that action type is extracted from the skeleton posture information. In some implementations, the skeletal keypoint coordinate sequence and the symmetrical bone point coordinate sequence are time-aligned. The time-alignment process includes direct matching based on time tags, and linear interpolation to unify the time scale when the time steps are inconsistent. Further, after completing the time-alignment, the point-by-point difference calculation is performed on the skeletal keypoint coordinates at the corresponding time points. The difference calculation can be performed through joint angle deviation. Specifically, multiple bone point coordinates adjacent to the target skeletal keypoint are extracted from the skeleton posture information. The skeletal point includes three adjacent skeletal nodes constituting the target joint, including a proximal skeletal node, a joint node (target skeletal key point), and a distal skeletal node. Based on the spatial coordinates of the three adjacent skeletal nodes, a joint vector is constructed, and the angle of the joint is calculated. The angle of the joint is calculated by the angle between two vectors formed by the joint node of the skeletal key point pointing to two adjacent skeletal nodes. The difference values ​​of the rate of change of the angle of the skeletal key point and the symmetrical skeletal point at multiple time points are weighted and summarized to obtain the posture contrast of the skeletal rehabilitation observation point under the current action type. The weighting method can be based on the time position and decrease or increase the weight, which will not be elaborated in this application.

[0041] In step S104, the rehabilitation trend of the posture contrast of the skeletal rehabilitation observation point is identified, and observability analysis is performed based on the rehabilitation trend of each skeletal rehabilitation observation point to determine the rehabilitation monitoring weight of the skeletal rehabilitation observation point.

[0042] It should be noted that the posture contrast described in this application characterizes the degree of spatial deviation of key skeletal points relative to a reference symmetrical structure during movement. The degree of deviation can reflect the degree of limitation and recovery status of skeletal function. Since the skeletal rehabilitation process is accompanied by the gradual recovery of joint range of motion and motion symmetry, the sequence formed by the change of posture contrast over time can characterize the dynamic evolution trend of the rehabilitation process. Compared with the traditional static evaluation method based on single-frame images, posture contrast, by introducing continuous change analysis in the time dimension, enables the skeletal function recovery status to be quantified into a calculable trend signal, thereby achieving dynamic identification of the rehabilitation process and improving the accuracy of rehabilitation progress identification.

[0043] Preferably, in some embodiments, identifying the rehabilitation trend of the posture contrast at the skeletal rehabilitation observation point specifically includes: Obtain the skeletal posture information of the historical records of the skeletal rehabilitation observation point. Based on the preset rehabilitation cycle window, extract the posture contrast information corresponding to the skeletal rehabilitation observation point in each rehabilitation cycle window, and construct the rehabilitation trend according to the timestamps corresponding to each rehabilitation cycle window.

[0044] In specific implementation, the historical data includes skeletal posture information and corresponding posture contrast at multiple time sampling points. The continuous time series is divided into multiple rehabilitation cycle windows according to a preset time length. The rehabilitation cycle window can be a time interval divided by day, week, or preset rehabilitation stage, thus forming multiple discrete time periods. For each rehabilitation cycle window, the posture contrast data corresponding to all time sampling points within the time window is extracted, and the posture contrast within the window is statistically aggregated to obtain the average contrast value corresponding to the rehabilitation cycle window. The average contrast values ​​corresponding to each rehabilitation cycle window are arranged in chronological order. In some embodiments, sequence fitting is performed based on the timestamps corresponding to each rehabilitation cycle window, and the fitting function is used as the rehabilitation trend corresponding to the skeletal rehabilitation observation point. Alternatively, resampling can be performed after time series fitting to obtain a discrete rehabilitation trend sequence with a higher sampling rate as the rehabilitation trend corresponding to the skeletal rehabilitation observation point. It is foreseeable that the rehabilitation trend in this application can be a continuous function model with time as the independent variable, or a discrete rehabilitation trend value time series, and its specific form is not limited.

[0045] It should be noted that the bone injury area is significantly affected by the bone injury during posture analysis, exhibiting jitter and instability during movement. Therefore, the rehabilitation observation points in the bone injury area are only used for modeling reference. The injured associated area with motion coupling is used as the basis for trend modeling, thereby improving the robustness of rehabilitation trend prediction. Preferably, in some embodiments, observability analysis is performed based on the rehabilitation trend of each skeletal rehabilitation observation point to determine the rehabilitation monitoring weight of the skeletal rehabilitation observation point, specifically including: Multiple skeletal rehabilitation observation points are divided into regions to identify the set of reference observation points in the bone injury area and the set of related observation points in the injury-related area; For any reference observation point, the intensity of the injury is determined based on the recovery trend of that reference observation point. For any associated observation point, determine the adjacent bone injury area of ​​the associated observation point; determine the associated rehabilitation trend and average injury intensity of multiple reference observation points within the adjacent bone injury area; A correlation analysis is performed between the rehabilitation trend of the associated observation point and the associated rehabilitation trend of the adjacent bone injury area to determine the trend correlation; the rehabilitation monitoring weight of the associated observation point is determined based on the trend correlation and the average injury intensity.

[0046] In specific implementation, multiple skeletal rehabilitation observation points are divided into regions according to their spatial positions in the topological structure of the human skeleton. The skeletal rehabilitation observation points are divided into a set of reference observation points and a set of associated observation points. The reference observation points are joint nodes with a core area of ​​bone injury, and the associated observation points are extended joint nodes that have structural connections and motion coupling relationships with the bone injury area. The region division is determined based on the adjacency relationship in the human skeleton diagram structure. The associated observation points are obtained by expanding according to the second-order neighborhood, with the bone injury center node as the starting node. In some embodiments, the reference observation points in the second-order neighborhood can be used as the adjacent bone injury areas of the associated observation points.

[0047] For any reference observation point, its recovery trend is determined based on the historical posture contrast sequence of that observation point, and the injury intensity is further calculated by a moving average of the slope change within a time window. The injury intensity characterizes the degree of functional impairment of the reference observation point. For any associated observation point, its adjacent bone injury area is determined based on the topology of the human skeleton. The adjacent bone injury area is a set of reference observation points that are directly adjacent to the associated observation point in the skeletal diagram structure. The associated recovery trend of multiple reference observation points in the adjacent bone injury area is calculated, where the associated recovery trend is the average of the recovery trends of all reference observation points adjacent to the associated observation point in the bone injury area. Specifically: associated recovery trend = arithmetic mean of the recovery trends of all reference observation points adjacent to the associated observation point in the bone injury area. In some implementations, the average injury intensity of multiple reference observation points within the adjacent bone injury area is calculated simultaneously, where the average injury intensity is the average of the injury intensity of each reference observation point. Further, the rehabilitation trend of the associated observation point is correlated with the associated rehabilitation trend of the adjacent bone injury area. Specifically, the Pearson correlation coefficient can be used as the trend correlation index. The rehabilitation monitoring weight of the associated observation point is obtained by fusing the trend correlation and the average injury intensity, using a product-normalized method. The rehabilitation monitoring weight of the associated observation point characterizes the degree of influence of the associated observation point in the overall skeletal rehabilitation process.

[0048] Optionally, in some embodiments, since the bone injury area is greatly affected by the bone injury, it is generally not directly involved in the prediction and modeling of the rehabilitation trend. The rehabilitation monitoring weight of the reference observation point in the bone injury area can be reset to zero, or the rehabilitation trend of the reference observation point can be analyzed for trend stability. For example, the rate of change of rehabilitation trend within multiple sliding time windows can be extracted, and the sum of the ratio between the standard value of variance and the variance of the rate of change of rehabilitation trend and the preset basic weight value can be used as the rehabilitation monitoring weight of the reference observation point.

[0049] In step S105, the rehabilitation progress of the target user is identified based on the rehabilitation monitoring weight, and the rehabilitation progress and personalized rehabilitation strategy are sent to the user terminal.

[0050] It should be noted that the physiological recovery information reflected by different skeletal rehabilitation observation points varies during the skeletal rehabilitation process. Among them, the observation points in the bone injury area directly correspond to the injury site, and their posture changes are more sensitive, reflecting the intensity of local functional recovery. However, they are also easily affected by factors such as pain suppression, motor compensation, and acquisition noise, resulting in large fluctuations in the rehabilitation trend and relatively weak stability. In order to improve the stability and reliability of rehabilitation trend modeling, this application uses the observation points in the bone injury area as a reference observation point set, and determines the injury intensity based on its rehabilitation trend to characterize the degree of local injury, thereby reflecting the core influencing factors in the rehabilitation process. At the same time, although the observation points in the injury-related area do not directly correspond to the center of the bone injury, they have an adjacency relationship or motor coupling relationship with the bone injury area in the topology of the human skeleton. Their posture changes can reflect the coordination and compensation of the overall movement pattern, and have better stability and continuity than the observation points in the injury area.

[0051] Based on the above characteristics, for any associated observation point, this application determines the adjacent bone injury area in the skeletal topology and extracts the associated rehabilitation trend and average injury intensity of multiple reference observation points in the adjacent area as a comprehensive representation of the overall rehabilitation status of the area. This avoids the uncertainty caused by fluctuations of a single observation point. Furthermore, by conducting correlation analysis between the rehabilitation trend of the associated observation point and the associated rehabilitation trend of the adjacent bone injury area, the trend correlation is determined to measure the consistency between local movement patterns and the overall rehabilitation status, and at the same time reflects the synergistic relationship of functional recovery. This avoids the interference of abnormalities of a single associated observation point on the prediction of the overall rehabilitation trend, thereby realizing differentiated modeling of different types of skeletal rehabilitation observation points and improving the robustness of rehabilitation trend prediction.

[0052] Optionally, in some embodiments, identifying the rehabilitation progress of the target user based on the rehabilitation monitoring weight specifically includes: Obtain the rehabilitation trend corresponding to each skeletal rehabilitation observation point, and predict the rehabilitation trend value for the next time period based on the rehabilitation trend prediction model. The prediction results are weighted and fused according to the rehabilitation monitoring weights corresponding to each skeletal rehabilitation observation point to determine the predicted value of the rehabilitation trend of the target user. Based on the predicted rehabilitation trend values, a rehabilitation trend mapping is performed to determine the rehabilitation progress level, and a personalized rehabilitation strategy is generated based on the rehabilitation progress level.

[0053] In practical implementation, the moving average autoregressive model, commonly used in time series forecasting in existing technologies, can be adopted as the recovery trend prediction model to predict the recovery trend value for the next time period. The following is a specific embodiment of this application using the moving average autoregressive model to predict the recovery trend value: The preset rehabilitation prediction sample period is a preset number of days, and the rehabilitation trend prediction interval is a preset time granularity, such as 1 day. Rehabilitation trend values ​​of the target skeletal rehabilitation observation point within a historical time range are continuously collected to obtain a rehabilitation trend time series dataset. This dataset consists of posture contrast statistics within multiple rehabilitation period windows, where each time point corresponds to a rehabilitation trend value, representing the functional recovery status of the skeletal rehabilitation observation point within that time period. Further, a change curve for the rehabilitation trend time series is constructed, and a stationarity analysis is performed on this curve to eliminate fluctuations caused by local posture jitter or observation noise. In some embodiments, the stationarity analysis includes exponential smoothing of the sequence. An autocorrelation analysis plot and a partial autocorrelation analysis plot are constructed based on the rehabilitation trend time series. The autoregressive order p of the rehabilitation trend sequence is determined based on the autocorrelation analysis plot. When the autocorrelation coefficient rapidly decays to near zero after a certain lag order, this order is determined as the autoregressive order. The moving average order q is determined based on the partial autocorrelation analysis plot. When the partial autocorrelation coefficient rapidly decays to near zero after a certain lag order, this order is determined as the moving average order. Further, based on the determined autoregressive order p and moving average order q, an autoregressive moving average prediction model ARMA(p,q) for skeletal rehabilitation trends is constructed, and the historical rehabilitation trend time series is input into the model. In some embodiments, the parameters of the ARMA model are estimated using the least squares method, and the significance of the parameters is tested to screen effective model parameters, wherein the confidence level of the significance test can be set to a preset threshold. Further, the predicted value of the rehabilitation trend for the next time period is obtained based on the model prediction results, and the predicted value is used as the estimated result of the rehabilitation trend of the target skeletal rehabilitation observation point in the next time period.

[0054] In specific implementation, the weighted rehabilitation trend prediction values ​​of multiple skeletal rehabilitation observation points are fused and calculated. Specifically, the overall rehabilitation trend prediction value of the target user is obtained by weighted summation, i.e., the overall rehabilitation trend prediction result of the target user = the sum of the rehabilitation trend prediction values ​​of each rehabilitation monitoring weight and the corresponding observation point. Based on the overall rehabilitation trend prediction value, rehabilitation trend mapping processing is performed to map the continuous numerical prediction result into discrete rehabilitation progress levels. The rehabilitation progress levels include, but are not limited to, the initial recovery stage, the intermediate recovery stage, and the stable recovery stage. Specifically, the mapping can be achieved by dividing the data into preset threshold intervals. For example, when the rehabilitation trend prediction value is lower than the first threshold, it corresponds to the initial recovery stage; when it is between the first threshold and the second threshold, it corresponds to the intermediate recovery stage; and when it is higher than the second threshold, it corresponds to the stable recovery stage. The thresholds can be preset based on the historical test data of bone injury patients, which will not be elaborated on in this application.

[0055] In some implementations, personalized rehabilitation strategies are generated based on the rehabilitation progress level. These personalized rehabilitation strategies include suggestions for adjusting exercise intensity, suggestions for adjusting rehabilitation training frequency, and risk warning information. The frequency and priority of strategy push are dynamically adjusted according to the rehabilitation progress level.

[0056] In another aspect, in some embodiments, this application provides a remote intelligent monitoring system for bone injury rehabilitation, the system including a rehabilitation monitoring unit, referenced... Figure 2 The figure is a schematic diagram of the exemplary hardware and / or software structure of a rehabilitation monitoring unit according to some embodiments of this application. The rehabilitation monitoring unit 200 includes: a user information acquisition module 201, an image acquisition module 202, a posture recognition module 203, and a rehabilitation strategy generation module 204, which are described below: User information acquisition module 201 is used to acquire bone injury information of target users and determine multiple skeletal rehabilitation observation points based on the bone injury information; Image acquisition module 202 is used to acquire continuous image data through a user terminal and extract skeleton pose information from the continuous image data; The posture recognition module 203 extracts the posture contrast of any skeletal rehabilitation observation point based on the skeletal posture information. The posture recognition module 203 is also used to identify the rehabilitation trend of the posture contrast of the skeletal rehabilitation observation point, and to perform observability analysis based on the rehabilitation trend of each skeletal rehabilitation observation point to determine the rehabilitation monitoring weight of the skeletal rehabilitation observation point. The rehabilitation strategy generation module 204 is used to identify the rehabilitation progress of the target user based on the rehabilitation monitoring weight, and send the rehabilitation progress and personalized rehabilitation strategy to the user terminal.

[0057] The foregoing has provided a detailed example of a remote intelligent monitoring system and method for bone injury rehabilitation provided in the embodiments of this application. It is understood that the corresponding device includes hardware structures and / or software modules for performing each function in order to achieve the above functions.

[0058] Those skilled in the art should readily recognize that, based on the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a certain function in the application is executed in a manner that drives hardware or computer software depends on the specific application and design constraints of the technical solution. Therefore, those skilled in the art can use different methods to implement the described function for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0059] In addition, this application also provides a computer terminal device, which includes a memory and a processor. The memory stores code, and the processor is configured to acquire the code and execute the above-described remote intelligent monitoring method for bone injury rehabilitation.

[0060] In some embodiments, reference Figure 3 This figure is a schematic diagram of the structure of a computer terminal device implementing a remote intelligent monitoring method for bone injury rehabilitation, according to some embodiments of this application. The remote intelligent monitoring method for bone injury rehabilitation described in the above embodiments can... Figure 3 The computer terminal device 300 shown is used to implement this, and the computer terminal device 300 includes at least one communication bus 301, communication interface 302, processor 303 and memory 304.

[0061] The processor 303 may be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more devices used to control the execution of a remote intelligent monitoring method for bone injury rehabilitation as described in this application.

[0062] The communication bus 301 may include a path for transmitting information between the aforementioned components.

[0063] Memory 304 may be read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions, random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 304 may exist independently and be connected to processor 303 via communication bus 301. Memory 304 may also be integrated with processor 303.

[0064] The memory 304 stores program code for executing the scheme of this application, and its execution is controlled by the processor 303. The processor 303 executes the program code stored in the memory 304. The program code may include one or more software modules. In the above embodiments, the determination of the attitude contrast parameter can be achieved by the processor 303 and one or more software modules in the program code in the memory 304.

[0065] Communication interface 302 uses any transceiver-like device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area network (WLAN), etc.

[0066] Optionally, the computer terminal device 300 may also include a power supply 305 for providing power to various devices or circuits in the real-time computer terminal device.

[0067] In a specific implementation, as one example, a computer terminal device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0068] The aforementioned computer terminal device can be a general-purpose computer terminal device or a dedicated computer terminal device. In specific implementations, the computer terminal device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer terminal device.

[0069] In addition, other aspects of this application also provide a computer-readable storage medium storing at least one computer program, which is loaded and executed by a processor to perform the operations described above in a remote intelligent monitoring method for bone injury rehabilitation.

[0070] In summary, the remote intelligent monitoring system and method for bone injury rehabilitation disclosed in this application first acquires the bone injury information of the target user and determines multiple skeletal rehabilitation observation points based on the bone injury information; continuous image data is collected through the user terminal, and skeletal posture information is extracted from the continuous image data; the posture contrast of the skeletal rehabilitation observation point is extracted based on the skeletal posture information; the rehabilitation trend of the posture contrast of the skeletal rehabilitation observation point is identified, and observability analysis is performed based on the rehabilitation trend of each skeletal rehabilitation observation point to determine the rehabilitation monitoring weight of the skeletal rehabilitation observation point; the rehabilitation progress of the target user is identified based on the rehabilitation monitoring weight, and the rehabilitation progress and personalized rehabilitation strategies are sent to the user terminal. This system can identify the rehabilitation progress based on the posture contrast of the skeletal observation points, thus improving the safety of bone injury rehabilitation care.

[0071] The above descriptions are merely embodiments of this application, and common knowledge such as specific technical solutions or characteristics in the solutions are not described in detail here. It should be noted that those skilled in the art can make several modifications and improvements without departing from the technical solutions of this application, and these should also be considered within the scope of protection of this application, without affecting the effectiveness of the implementation of this application or the practicality of the patent.

[0072] The scope of protection claimed in this application shall be determined by the content of its claims. The specific embodiments described in the specification can be used to interpret the content of the claims. Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of the invention. Therefore, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.

Claims

1. A remote intelligent monitoring method for bone injury rehabilitation, characterized in that, include: Obtain bone injury information of the target user, and determine multiple skeletal rehabilitation observation points based on the bone injury information; Continuous image data is collected through a user terminal, and skeleton pose information is extracted from the continuous image data. For any skeletal rehabilitation observation point, extract the posture contrast of the skeletal rehabilitation observation point based on the skeletal posture information. Identify the rehabilitation trend of the posture contrast at the skeletal rehabilitation observation point, and perform observability analysis based on the rehabilitation trend of each skeletal rehabilitation observation point to determine the rehabilitation monitoring weight of the skeletal rehabilitation observation point. The rehabilitation progress of the target user is identified based on the rehabilitation monitoring weight, and the rehabilitation progress and personalized rehabilitation strategy are sent to the user terminal.

2. The method as described in claim 1, characterized in that, Extracting skeleton pose information from the continuous image data specifically includes: The continuous image data is acquired, and the skeletal key points are extracted from the continuous image data using a pre-trained human pose recognition model to obtain the skeletal key point coordinate information corresponding to multiple continuous images respectively. For any consecutive image, construct the skeleton topology based on the coordinate information of the skeletal key points corresponding to the consecutive image, and perform temporal correlation processing on the skeleton topology of adjacent frames according to the time frames corresponding to the consecutive images to obtain the skeleton pose information.

3. The method as described in claim 1, characterized in that, Extracting the posture contrast of the skeletal rehabilitation observation point based on the skeletal posture information specifically includes: Obtain the coordinate sequence of the skeletal key points corresponding to the skeletal rehabilitation observation points in the skeletal posture information; Action type identification is performed based on the coordinate sequence of the skeletal key points, and the same type of coordinate sequence is extracted from the coordinate sequence of symmetrical skeletal points based on the action type; The coordinate sequences of the skeletal key points and the symmetrical skeletal points are temporally aligned, and the spatial difference value of the corresponding joints is calculated as the pose contrast for the aligned skeletal key point coordinate data.

4. The method as described in claim 1, characterized in that, The rehabilitation trend of identifying the posture contrast of the skeletal rehabilitation observation point specifically includes: obtaining the skeletal posture information of the historical records of the skeletal rehabilitation observation point; extracting the posture contrast information corresponding to the skeletal rehabilitation observation point in each rehabilitation cycle window based on a preset rehabilitation cycle window; and constructing a rehabilitation trend based on the timestamps corresponding to each rehabilitation cycle window.

5. The method as described in claim 1, characterized in that, Based on the rehabilitation trends at each skeletal rehabilitation observation point, an observability analysis was conducted to determine the rehabilitation monitoring weights for each observation point, specifically including: Multiple skeletal rehabilitation observation points are divided into regions to identify the set of reference observation points in the bone injury area and the set of related observation points in the injury-related area; For any reference observation point, the intensity of the injury is determined based on the recovery trend of that reference observation point. For any associated observation point, determine the adjacent bone injury area of ​​the associated observation point; determine the associated rehabilitation trend and average injury intensity of multiple reference observation points within the adjacent bone injury area; A correlation analysis is performed between the rehabilitation trend of the associated observation point and the associated rehabilitation trend of the adjacent bone injury area to determine the trend correlation; the rehabilitation monitoring weight of the associated observation point is determined based on the trend correlation and the average injury intensity.

6. The method as described in claim 1, characterized in that, The specific steps for identifying the rehabilitation progress of the target user based on the rehabilitation monitoring weights include: Obtain the rehabilitation trend corresponding to each skeletal rehabilitation observation point, and predict the rehabilitation trend value for the next time period based on the rehabilitation trend prediction model. The prediction results are weighted and fused according to the rehabilitation monitoring weights corresponding to each skeletal rehabilitation observation point to determine the predicted value of the rehabilitation trend of the target user. Based on the predicted rehabilitation trend values, a rehabilitation trend mapping is performed to determine the rehabilitation progress level, and a personalized rehabilitation strategy is generated based on the rehabilitation progress level.

7. The method as described in claim 1, characterized in that, Before extracting skeleton pose information from the continuous image data, the method further includes: performing image preprocessing on the continuous image data.

8. A remote intelligent monitoring system for bone injury rehabilitation, comprising a rehabilitation monitoring unit, wherein the rehabilitation monitoring unit is used to execute the remote intelligent monitoring method for bone injury rehabilitation as described in any one of claims 1 to 7, characterized in that, The rehabilitation monitoring unit includes: The user information acquisition module is used to acquire the bone injury information of the target user and determine multiple skeletal rehabilitation observation points based on the bone injury information. The image acquisition module is used to acquire continuous image data through a user terminal and extract skeleton pose information from the continuous image data; The posture recognition module extracts the posture contrast of any skeletal rehabilitation observation point based on the skeletal posture information. The posture recognition module is also used to identify the rehabilitation trend of the posture contrast of the skeletal rehabilitation observation point, and to perform observability analysis based on the rehabilitation trend of each skeletal rehabilitation observation point to determine the rehabilitation monitoring weight of the skeletal rehabilitation observation point. The rehabilitation strategy generation module is used to identify the rehabilitation progress of the target user based on the rehabilitation monitoring weight, and send the rehabilitation progress and personalized rehabilitation strategy to the user terminal.

9. A computer terminal device, characterized in that, The computer terminal device includes a memory and a processor, the memory storing code, and the processor being configured to retrieve the code, such that when the code is executed in the processor, the method as described in any one of claims 1-7 is implemented.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed on a computer, cause the method as described in any one of claims 1-7 to be implemented.

Images