Intelligent monitoring system and method for rehabilitation of fracture patients

Abstract

The present application provides a fracture patient rehabilitation intelligent monitoring system and method, which relates to the field of intelligent monitoring. The system first collects the patient's blood flow and affected area vibration data through a blood flow sensor and a vibration sensor, and uploads these data to a cloud server for storage, and then retrieves these data from the cloud server for analysis to monitor the patient's fracture rehabilitation, and generates an early warning signal based on the obtained monitoring results, and finally displays the monitoring results and early warning signals. In this way, the patient's fracture rehabilitation status can be monitored more accurately.

Description

Technical Field

[0001] The present application relates to the field of intelligent monitoring, and more specifically, to an intelligent monitoring system and method for the rehabilitation of fracture patients. Background Art

[0002] Fracture is a common trauma, and patients usually need to go through a long recovery process after treatment. Effective rehabilitation monitoring is crucial, as it can help the medical team identify potential complications such as infection and nonunion in a timely manner, so that precise intervention measures can be taken before the problem worsens to prevent secondary injuries. Continuous monitoring can not only improve rehabilitation outcomes, but also enhance patients' confidence in recovery, ensuring that they receive the best quality care and treatment.

[0003] The invention with publication number CN110097978A discloses a method for subsequent rehabilitation monitoring of patients with fractures and joint replacements using fixation plates. Biocompatible wearable devices are set on the fixation plates and on the patients to monitor vital signs and rehabilitation status. The data is wirelessly transmitted to a cloud service platform. The cloud server intelligently analyzes the data, evaluates health and plate conditions, automatically alarms and notifies relevant personnel when abnormalities occur, and provides the location of the nearest hospital.

[0004] In this invention, a specific medical information algorithm is used to process and analyze the collected data to evaluate the patient's health status. The specific medical information algorithm here mainly relies on traditional statistical analysis, rule-based logical judgment and some classic machine learning models. However, although the patent mentions personalized adjustment, its adaptive learning ability is limited and relies more on preset parameters and thresholds. As the patient's rehabilitation process progresses, their physiological parameters may change, and fixed thresholds may not accurately reflect the latest health status, which means that the system is less adaptable to new data or special situations, and it cannot automatically adjust its analysis strategy as individual differences in patients or changes in rehabilitation progress change. In addition, for complex rehabilitation processes or special cases, rule-based methods may find it difficult to provide sufficient flexibility and support, because they usually require pre-definition of all possible situations and their corresponding treatment methods, which is almost impossible to fully cover in practice.

[0005] Therefore, an optimized intelligent monitoring solution for the rehabilitation of fracture patients is desired. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide an intelligent monitoring system and method for the rehabilitation of fracture patients.

[0007] According to one aspect of the present application, a fracture patient rehabilitation intelligent monitoring system is provided, comprising:

[0008] A patient data acquisition module, used to respectively acquire a blood flow time series data set and an affected part vibration signal of a fracture patient object through a blood flow sensor and a vibration sensor; a patient data transmission and storage module, used to transmit the data acquired by the patient data acquisition module to a cloud server for storage; a patient data analysis module, used to retrieve the stored blood flow time series data set and the affected part vibration signal from the cloud server and analyze them to obtain a fracture rehabilitation monitoring result; an early warning module, used to generate an early warning signal based on the fracture rehabilitation monitoring result; a result display module, used to display the fracture rehabilitation monitoring result and the early warning signal; wherein the patient data analysis module includes:

[0009] A patient blood flow encoding unit, used for performing time series encoding based on the patient blood flow on the blood flow time series data set to obtain a patient object blood flow time series feature;

[0010] An affected part vibration feature encoding unit, used for calculating the kurtosis value distribution and extracting the vibration abnormal pattern feature of the affected part vibration signal to obtain a semantic representation of the affected part vibration abnormal pattern;

[0011] The affected part feature joint encoding unit is used to perform cross-modal prompt significant joint encoding on the patient object's blood flow timing characteristics and the affected part vibration abnormality pattern semantic representation to obtain the patient object's blood flow timing-vibration abnormality semantic cross-modal prompt significant joint encoding features, and based on the patient object's blood flow timing-vibration abnormality semantic cross-modal prompt significant joint encoding features, obtain the fracture rehabilitation monitoring results.

[0012] Furthermore, the patient blood flow encoding unit is used to: pass the blood flow time series data set through a patient blood flow time series feature encoder based on a one-dimensional causal convolutional neural network model to obtain a patient object blood flow time series feature vector as the patient object blood flow time series feature.

[0013] Furthermore, the affected part vibration characteristic encoding unit comprises:

[0014] The affected part vibration signal kurtosis value distribution calculation subunit is used to calculate the kurtosis value distribution diagram of the affected part vibration signal to obtain the affected part vibration frequency kurtosis value distribution diagram;

[0015] The affected part vibration abnormal pattern feature extraction subunit is used to pass the affected part vibration frequency kurtosis value distribution diagram through the vibration abnormal pattern feature extractor based on the FCN model to obtain the affected part vibration abnormal pattern semantic representation diagram as the affected part vibration abnormal pattern semantic representation.

[0016] Furthermore, the affected part feature joint encoding unit comprises:

[0017] A blood flow autocorrelation calculation subunit, used to calculate the blood flow autocorrelation characteristics of the blood flow time series feature vector of the patient object;

[0018] A blood flow-vibration local cross-modal interaction subunit is used to perform local cross-modal prompt interaction on the affected part vibration abnormality pattern semantic representation map based on the blood flow autocorrelation feature to obtain a patient blood flow time series-vibration abnormality semantic cross-modal prompt significant joint coding feature map as the patient blood flow time series-vibration abnormality semantic cross-modal prompt significant joint coding feature;

[0019] The fracture rehabilitation monitoring result generating subunit is used to obtain the fracture rehabilitation monitoring result based on the patient's blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature map, and the fracture rehabilitation monitoring result is used to indicate whether there is an abnormal fracture rehabilitation situation.

[0020] Furthermore, the blood flow autocorrelation calculation subunit is used to calculate the product between the patient object blood flow time series feature vector and the transposed vector of the patient object blood flow time series feature vector to obtain the patient object blood flow autocorrelation coding matrix as the blood flow autocorrelation feature.

[0021] Furthermore, the blood flow-vibration local cross-modal interaction subunit comprises:

[0022] A secondary subunit for decoupling the feature of the abnormal vibration pattern of the affected part is used to perform feature decoupling along the channel dimension on the semantic representation graph of the abnormal vibration pattern of the affected part to obtain a set of semantic local feature matrices of the abnormal vibration pattern of the affected part;

[0023] A patient object blood flow feature linear transformation secondary subunit, used for linearly transforming the patient object blood flow autocorrelation coding matrix to obtain a patient object blood flow semantic query coding matrix and a patient object blood flow semantic value coding matrix;

[0024] A secondary subunit for generating blood flow-vibration cross-modal semantic features is used to perform Transformer-based cross-modal prompt information encoding on each affected part vibration abnormality pattern semantic local feature matrix in the set of the patient object blood flow semantic query encoding matrix, the patient object blood flow semantic value encoding matrix, and the affected part vibration abnormality pattern semantic local feature matrix to obtain a set of blood flow-vibration semantic cross-modal prompt information semantic encoding matrices;

[0025] The blood flow-vibration feature weighted attention interaction secondary subunit is used to perform gated weighted attention interaction optimization on the set of semantic coding matrices of the blood flow-vibration semantic cross-modal prompt information to obtain the significant joint coding feature map of the patient's blood flow timing-vibration abnormal semantic cross-modal prompt.

[0026] Furthermore, the blood flow-vibration feature weight attention interaction secondary subunit is used to:

[0027] Performing decoding gating processing on the set of blood flow-vibration semantic cross-modal prompt information semantic encoding matrices to obtain a set of blood flow-vibration cross-modal semantic interaction attention weights;

[0028] Multiplying the blood flow autocorrelation coding matrix of the patient object by each affected part vibration abnormality pattern semantic local feature matrix in the set of affected part vibration abnormality pattern semantic local feature matrices by position point to obtain a set of blood flow-vibration cross-modal interaction local feature matrices;

[0029] The set of the blood flow-vibration cross-modal semantic interaction attention weights and the set of the blood flow-vibration cross-modal interaction local feature matrices are cross-modally jointly encoded to obtain the patient subject's blood flow timing-vibration abnormal semantic cross-modal prompt significant joint encoding feature map.

[0030] Furthermore, the fracture rehabilitation monitoring result generating subunit is used to: pass the patient subject's blood flow timing-vibration abnormal semantic cross-modal prompt significant joint encoding feature map through a classifier-based rehabilitation status monitor to obtain the fracture rehabilitation monitoring result, and the fracture rehabilitation monitoring result is used to indicate whether there is an abnormal fracture rehabilitation situation.

[0031] According to another aspect of the present application, a method for intelligently monitoring the rehabilitation of fracture patients is provided, which is used to execute the intelligent monitoring system for the rehabilitation of fracture patients as described above, and comprises:

[0032] The blood flow time series data set and the vibration signal of the affected part of the fracture patient are collected by the blood flow sensor and the vibration sensor respectively; the data collected by the patient data collection module is transmitted to the cloud server for storage; the stored blood flow time series data set and the vibration signal of the affected part are retrieved from the cloud server and analyzed to obtain the fracture rehabilitation monitoring result; based on the fracture rehabilitation monitoring result, an early warning signal is generated; the fracture rehabilitation monitoring result and the early warning signal are displayed;

[0033] The blood flow time series data set and the affected part vibration signal stored in the cloud server are retrieved and analyzed to obtain the fracture rehabilitation monitoring result, including:

[0034] Performing time series encoding based on the patient's blood flow on the blood flow time series data set to obtain a blood flow time series feature of the patient object;

[0035] Calculating the distribution of kurtosis values ​​and extracting the features of abnormal vibration patterns on the affected part vibration signal to obtain semantic representation of the abnormal vibration pattern of the affected part;

[0036] The patient's blood flow timing characteristics and the affected area vibration abnormality pattern semantic representation are cross-modal significantly jointly encoded to obtain the patient's blood flow timing-vibration abnormality semantic cross-modal significantly jointly encoded features, and based on the patient's blood flow timing-vibration abnormality semantic cross-modal significantly jointly encoded features, the fracture rehabilitation monitoring results are obtained.

[0037] Furthermore, the blood flow timing data set is time-series encoded based on the patient's blood flow to obtain the patient's blood flow timing characteristics, including: passing the blood flow timing data set through a patient's blood flow timing characteristic encoder based on a one-dimensional causal convolutional neural network model to obtain a patient's blood flow timing characteristic vector as the patient's blood flow timing characteristic.

[0038] Compared with the prior art, the fracture patient rehabilitation intelligent monitoring system and method provided by the present application first collects the patient's blood flow and affected area vibration data through a blood flow sensor and a vibration sensor, and uploads these data to a cloud server for storage, and then retrieves these data from the cloud server for analysis to monitor the patient's fracture rehabilitation, and generates an early warning signal based on the obtained monitoring results, and finally displays the monitoring results and early warning signals. In this way, it is possible to more accurately monitor the patient's fracture rehabilitation status. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0040] Figure 1 This is a system block diagram of an intelligent monitoring system for fracture patient rehabilitation according to an embodiment of the present application.

[0041] Figure 2 This is a block diagram of a patient data analysis module in the intelligent monitoring system for fracture patient rehabilitation according to an embodiment of the present application.

[0042] Figure 3 It is a block diagram of the affected part feature joint encoding unit in the intelligent monitoring system for the rehabilitation of fracture patients according to an embodiment of the present application.

[0043] Figure 4 This is a flow chart of the intelligent monitoring method for the rehabilitation of fracture patients according to an embodiment of the present application. DETAILED DESCRIPTION

[0044] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0045] Fracture rehabilitation monitoring is crucial for preventing complications and improving treatment outcomes. The invention with publication number CN110097978A introduces a method for follow-up rehabilitation monitoring of fracture and joint replacement patients using fixed plates, which uses biocompatible wearable devices to collect the patient's vital signs data and transmit the data to a cloud service platform via wireless transmission. Specific medical information algorithms are used on the cloud server to analyze the data, assess the patient's health and plate condition, and automatically alarm when abnormalities occur. Although this method provides personalized adjustments, its adaptive learning capabilities are limited, it relies on preset parameters and thresholds, and it is difficult to adapt to changes in physiological parameters during the patient's rehabilitation process, and it lacks flexibility and support for complex cases.

[0046] Based on this, the present application provides an intelligent monitoring system for the rehabilitation of fracture patients. Figure 1 FIG. 1 is a system block diagram of an intelligent monitoring system for fracture patient rehabilitation according to an embodiment of the present application. Figure 1 As shown, in the intelligent monitoring system 100 for the rehabilitation of fracture patients, it includes: a patient data acquisition module 110, which is used to collect the blood flow time series data set and the vibration signal of the affected part of the fracture patient through a blood flow sensor and a vibration sensor respectively; a patient data transmission and storage module 120, which is used to transmit the data collected by the patient data acquisition module to a cloud server for storage; a patient data analysis module 130, which is used to retrieve the stored blood flow time series data set and the affected part vibration signal from the cloud server and analyze them to obtain the fracture rehabilitation monitoring result; an early warning module 140, which is used to generate an early warning signal based on the fracture rehabilitation monitoring result; and a result display module 150, which is used to display the fracture rehabilitation monitoring result and the early warning signal.

[0047] In the embodiment of the present application, the patient data acquisition module 110 is used to collect the blood flow time series data set and the vibration signal of the fractured patient object through the blood flow sensor and the vibration sensor respectively. It should be understood that the blood flow time series data set contains the data sequence information of the blood flow volume of the fracture site of the fracture patient over a period of time. These data reflect the dynamic process of local blood circulation, including the change of blood flow velocity, that is, the fluctuation of the rate at which blood passes through a specific area per unit time; the value of the blood flow and its distribution at different times, which reflects the degree of blood perfusion in the site; and may also contain rhythmic information of blood flow, such as whether there is a normal physiological fluctuation cycle. These information can be combined to reflect the blood supply status of the fracture site. In the process of fracture healing, sufficient and stable blood supply is one of the key factors. Good blood circulation helps to transport nutrients and oxygen to the fracture site, promote the regeneration and repair of bone cells, and take away metabolic waste. If the blood flow time series data is abnormal, such as continuous low blood flow, abnormal fluctuations or rhythm disorders, it indicates that the fracture healing is facing problems, such as blood vessel damage affecting blood supply, inflammation of local tissues or other factors that hinder blood circulation, which is closely related to abnormal fracture recovery. The vibration signal of the affected part contains information related to the vibration generated by the fracture site and its surrounding tissues during movement or stress, specifically covering the vibration characteristics generated by changes in muscle tension. The change in muscle strength during contraction and relaxation will cause corresponding changes in vibration frequency and amplitude, which is closely related to the recovery of muscle function and the support and protection of muscles to the fracture site; the tension signal data of the medial and lateral collateral ligaments is also reflected in the vibration signal. The tension and stability of the ligaments affect the stability of the fracture site, and the vibration mode generated by the change in its tension can reflect the health status and recovery degree of the ligaments; in addition, it also includes the overall vibration caused by physical activity or external factors acting on the affected part, such as the vibration response generated by the impact force on the fracture site when the patient walks or moves his limbs. This information is helpful to judge the recovery status of the fracture site. Normal muscle and ligament function and stable fracture site status are important signs of recovery. If the vibration signal is abnormal, such as abnormal high-frequency vibration, vibration amplitude beyond the normal range, or disordered vibration pattern, it may indicate abnormal recovery conditions such as muscle strength imbalance, unhealed ligament injury, or unstable fixation of the fracture site. In general, by comprehensively analyzing these two types of data, it is possible to more comprehensively and accurately judge whether the fracture recovery process is normal, and timely discover potential abnormal recovery conditions so that the medical team can take targeted intervention measures to improve the recovery effect.

[0048] The blood flow sensor and vibration sensor are used to collect the blood flow time series data set and the vibration signal of the affected part of the fracture patient. This step involves the selection of sensor type, sensor placement, and the determination of data collection frequency and duration. The following is a detailed description of its implementation process:

[0049] For the collection of blood flow time series data, we must first select the appropriate sensor type, such as non-invasive sensors such as laser Doppler blood flow meters, which measure blood flow velocity and calculate blood flow by emitting lasers to illuminate the skin surface and detecting changes in the frequency of reflected light. They have high accuracy and applicability for skin surface blood flow detection. In some cases where more precise measurements are required, implantable blood flow sensors can also be considered, but these need to be placed during surgery and biocompatibility and long-term stability must be considered. The placement of the sensor is also critical. For fracture patients, it is usually placed on the skin surface near the fracture site. Upper limb fractures can be placed on the forearm or upper arm, and lower limb fractures can be placed on the calf or thigh near the fracture site. If it is an implantable sensor, it may be placed directly in the tissue near the fracture end during fracture surgery, but the surgical specifications must be strictly followed and subsequent monitoring must be performed to ensure normal operation and no complications such as infection. The frequency and duration of data collection vary depending on the patient's recovery stage. A higher frequency may be required in the early stages of a fracture, such as collecting data every 1-5 minutes. As rehabilitation progresses, the frequency can be appropriately reduced, such as collecting data every 10-30 minutes. Collection should continue throughout the entire rehabilitation process, from the early stages after fracture treatment to basic recovery, which may last for weeks or even months to obtain a complete blood flow time series data set to reflect the dynamic changes in blood supply.

[0050] There are also many key points in vibration sensor collection. The sensor type can be an accelerometer, which can measure the acceleration changes of an object in three axes to obtain vibration information. It can be installed on the skin surface near the fracture site or fixed on an external support device. For example, for wrist fractures, it can be pasted on the skin on the back of the wrist, and for leg fractures, it can be pasted on the knees, ankles and other parts. If an external support device is used, it can also be installed on it; you can also choose a strain gauge sensor, which reflects the vibration by measuring the strain when the object is subjected to force. It can be pasted on the skin near the muscles or ligaments around the fracture site. For example, for upper limb fractures, it can be pasted near the biceps brachii, and for lower limb fractures, it can be pasted near the quadriceps femoris and the inner and outer collateral ligaments. The frequency of data collection is also related to the rehabilitation stage. In the early rehabilitation stage, when patients begin to engage in slight activities, a higher frequency may be required, such as collecting data every 2-5 minutes. As rehabilitation progresses, the amount of activity increases and the fracture site is relatively stable, the frequency can be appropriately reduced, such as collecting data every 10-20 minutes. Moreover, collection should continue throughout the entire rehabilitation process, from early activities after fracture treatment to the time when the patient can move normally and the doctor determines that the rehabilitation is basically completed, so as to obtain a complete vibration signal of the affected area to reflect the vibration changes of the fracture site and its surrounding tissues during the rehabilitation process.

[0051] In summary, collecting blood flow time series data sets and vibration signals of the affected area from fracture patients through blood flow sensors and vibration sensors is a complex process that requires careful design and implementation. It can ensure that complete data is obtained, thereby providing a reliable data basis for subsequent intelligent monitoring of fracture patient rehabilitation.

[0052] In an embodiment of the present application, the patient data storage and storage module 120 is used to transmit the data collected by the patient data acquisition module to the cloud server for storage. It should be understood that in the rehabilitation process of fracture patients, data collected by various types of sensors are involved. In order to centrally manage these scattered data, the collected data can be transmitted to the cloud server for storage. The cloud server has a large-capacity storage capacity and can store the patient's complete data records from the early to late stages of rehabilitation. It can also analyze long-term series of patient data through powerful computing power, and can more accurately detect subtle changes and potential problems in the patient's rehabilitation process, thereby providing strong support for personalized treatment and precise intervention. And by storing patient data on a cloud server, the patient's latest data can be easily and quickly obtained without being limited to specific equipment or locations, which greatly improves the convenience and flexibility of data acquisition and helps to make medical decisions in a timely manner.

[0053] Transmitting the data collected by the patient data collection module to the cloud server for storage is a complex process involving multiple links. The following is a detailed description of its implementation process:

[0054] First, the blood flow sensor and vibration sensor need to be connected to the data acquisition device. There are various ways to connect the sensor to the acquisition device. For wired sensors, analog signals are often transmitted to the acquisition device through USB interfaces, RS-232 interfaces, etc., and the acquisition device then converts them into digital signals. Wireless sensors are connected to the acquisition device with the help of wireless communication protocols such as Bluetooth, ZigBee or Wi-Fi. This method reduces the connection constraints on the patient's body and improves comfort. At the same time, the data acquisition device must be reasonably configured, and its parameter settings are crucial. For example, the sampling frequency needs to be set according to the sensor characteristics and actual needs. For example, the blood flow sensor may collect data every 1-5 minutes in the early stage, and the vibration sensor may collect data every 2-5 minutes in the early stage. In addition, the appropriate signal gain should be set according to the output signal strength of the sensor to ensure that the collected data is within the appropriate range to avoid signal distortion or inaccurate collection.

[0055] Next, in the data transmission network link, the network selection should be based on the actual situation. If there is a LAN environment inside the medical institution, the acquisition equipment can be connected to the LAN by wire (such as Ethernet cable) or wireless (such as Wi-Fi), and then the data is transmitted to the cloud server through the network server of the medical institution. This method can provide stable and high-speed transmission. If the medical institution does not have a suitable LAN environment, or the patient is recovering at home, the acquisition equipment can directly transmit data to the cloud server through a mobile network (such as 4G or 5G network). The wide coverage of the mobile network can meet the monitoring needs of different locations. In addition, in order to ensure data security, the data must be encrypted during the transmission process. Encryption algorithms such as AES are often used to encrypt blood flow and vibration data, so that even if the data is intercepted, it cannot be easily interpreted. At the same time, a secure transmission channel should be established, such as using the SSL / TLS protocol to encrypt and authenticate the data at the network layer and the transport layer to ensure the integrity and confidentiality of the data and prevent the data from being tampered with or leaked.

[0056] Finally, regarding the configuration and storage of cloud servers, medical institutions or relevant service providers should choose appropriate cloud servers. It can be cloud computing resources of public cloud service providers or private cloud servers. The cloud server needs to have sufficient storage capacity to store complete data records of patients from the early to late stages of rehabilitation. For medium-sized medical institutions, it may be necessary to configure storage capacity of several TB to tens of TB. At the same time, it should also have strong computing power, by configuring high-performance CPUs, large amounts of memory and appropriate storage architectures (such as distributed storage systems) to support the analysis of long-term patient data series and accurately discover subtle changes and potential problems in the rehabilitation process. When the data is transferred to the cloud server, effective storage management should be carried out. Database management systems (such as MySQL, Oracle, etc.) can be used to store data, and blood flow and vibration data can be stored according to key information such as patient ID and collection time. A separate database table is created for each patient, which contains relevant data fields. In addition, in order to improve the efficiency of data query and analysis, an index mechanism should be established, such as indexing the collection time field, so as to quickly locate relevant data records and reduce query time. At the same time, data should be backed up regularly, using a combination of full backup and incremental backup, such as weekly full backup and daily incremental backup, to ensure data security and recoverability.

[0057] In the embodiment of the present application, the patient data analysis module 130 is used to retrieve the stored blood flow time series data set and the affected part vibration signal from the cloud server and analyze them to obtain the fracture rehabilitation monitoring result.

[0058] Correspondingly, in the patient data analysis module 130, the technical concept of the present application is to retrieve the blood flow time series data set and the affected part vibration signal of the fracture patient object from the cloud server, and use the data analysis and signal processing technology based on machine learning to perform time series feature analysis on the blood flow of the fracture patient object, and at the same time perform kurtosis distribution calculation and vibration anomaly analysis on the affected part vibration signal, so as to intelligently judge the abnormal situation of fracture rehabilitation based on the cross-modal prompt coding representation between the patient object blood flow time series characteristics and the semantic representation of the abnormal vibration pattern of the affected part. In this way, the unique health pattern of each patient can be identified based on the unique characteristics of each patient and the dynamic changes in the rehabilitation process, and the evaluation criteria can be optimized accordingly. In addition, this method can also provide a more comprehensive understanding of the patient's state changes, and can also provide more accurate evaluation and support for patients with complex rehabilitation processes or special situations that are difficult to predict with established rules.

[0059] Specifically, in an embodiment of the present application, the blood flow time series data set and the affected part vibration signal of the fracture patient object are first retrieved from the cloud server. It should be understood that since the data stored in the cloud server is complete and has been sorted to a certain extent, the quality of the retrieved data is guaranteed, and a reliable data basis can be provided for subsequent analysis. Based on reliable data for analysis, the fracture rehabilitation monitoring results obtained are more credible, reducing the risk of misjudgment due to missing or incorrect data. And the cloud server supports fast data retrieval, which enables timely detection and intervention measures to be taken in the patient's rehabilitation process once abnormal signs appear, avoiding delays in the disease.

[0060] Specifically, Figure 2 FIG. 1 is a block diagram of a patient data analysis module in a fracture patient rehabilitation intelligent monitoring system according to an embodiment of the present application. Figure 2 As shown, the patient data analysis module 130 includes: a patient blood flow encoding unit 131, which is used to perform time series encoding on the blood flow time series data set based on the patient blood flow to obtain the patient object blood flow time series characteristics; an affected part vibration feature encoding unit 132, which is used to perform kurtosis value distribution calculation and vibration abnormality pattern feature extraction on the affected part vibration signal to obtain the affected part vibration abnormality pattern semantic representation; an affected part feature joint encoding unit 133, which is used to perform cross-modal prompt significant joint encoding on the patient object blood flow time series characteristics and the affected part vibration abnormality pattern semantic representation to obtain the patient object blood flow time series-vibration abnormality semantic cross-modal prompt significant joint encoding characteristics, and obtain the fracture rehabilitation monitoring result based on the patient object blood flow time series-vibration abnormality semantic cross-modal prompt significant joint encoding characteristics.

[0061] In the embodiment of the present application, the patient blood flow encoding unit 131 is used to perform time series encoding based on the patient blood flow on the blood flow time series data set to obtain the patient object blood flow time series characteristics. Specifically, in the embodiment of the present application, the patient blood flow encoding unit 131 is used to: pass the blood flow time series data set through a patient blood flow time series feature encoder based on a one-dimensional causal convolutional neural network model to obtain a patient object blood flow time series feature vector as the patient object blood flow time series feature. It should be understood that the blood flow of the fracture patient object is constantly changing with the change of time points. Therefore, in order to more accurately capture and mine the time series feature information of blood flow in different local time ranges, such as the continuous increase or fluctuation of blood flow in a local time period, in the technical solution of the present application, the blood flow time series data set of the fracture patient object is passed through a patient blood flow time series feature encoder based on a one-dimensional causal convolutional neural network model to slide on the time axis using a convolution kernel of appropriate size through a special filling and delay mechanism to capture the local time characteristics of blood flow, such as periodic fluctuations and trend changes, and obtain a patient object blood flow time series feature vector. It should be understood that the one-dimensional causal convolutional neural network is a deep learning model specifically designed to process time series data. In particular, it combines the concepts of one-dimensional convolutional neural network and causal convolution to ensure that only the current and previous time point information is used in the prediction or analysis process without leaking future data, which ensures the authenticity and reliability of the model prediction, that is, it ensures that the model's prediction is based on actual events rather than hypothetical future data, thereby enhancing the model's ability to capture the time dependency of blood flow to accurately understand the patient's blood flow timing information.

[0062] In the embodiment of the present application, the affected part vibration feature encoding unit 132 is used to perform kurtosis value distribution calculation and vibration abnormal pattern feature extraction on the affected part vibration signal to obtain a semantic representation of the affected part vibration abnormal pattern. Specifically, in the embodiment of the present application, the affected part vibration feature encoding unit 132 includes: an affected part vibration signal kurtosis value distribution calculation subunit, used to calculate the kurtosis value distribution map of the affected part vibration signal to obtain an affected part vibration frequency kurtosis value distribution map; an affected part vibration abnormal pattern feature extraction subunit, used to pass the affected part vibration frequency kurtosis value distribution map through a vibration abnormal pattern feature extractor based on the FCN model to obtain an affected part vibration abnormal pattern semantic representation map as the affected part vibration abnormal pattern semantic representation.

[0063] It should be understood that there may be times with higher instantaneous frequencies in the affected part vibration signal, and these frequencies are manifestations of some vibration abnormalities. It is understandable that kurtosis is a statistic used to measure the "peak state" of the probability distribution, that is, the sharpness of the distribution curve. For the affected part vibration signal, kurtosis can help identify whether there are abnormal high-amplitude events or shocks. These abnormal vibrations may be caused by improper exercise, external force or other adverse factors, which have a potential negative impact on fracture rehabilitation. Based on this, in the technical solution of the present application, the kurtosis value distribution map of the affected part vibration signal is calculated to obtain the kurtosis value distribution map of the affected part vibration frequency. That is to say, by analyzing the kurtosis value distribution map of the affected part vibration signal, it is possible to identify which frequency intervals have higher kurtosis values, which means that there may be abnormal activities in the frequency interval. For example, if the kurtosis value of certain specific frequency intervals is found to increase suddenly during the patient's rehabilitation process, this may be due to abnormal vibration modes caused by problems such as loosening or breaking of steel plates or bolts. Similarly, it may also be abnormal muscle tension and ligament tension caused by inappropriate exercise and functional training. At this time, the rehabilitation plan needs to be adjusted to avoid complications.

[0064] Calculating the kurtosis value distribution map of the affected part vibration signal to obtain the kurtosis value distribution map of the affected part vibration frequency involves multiple links such as kurtosis value calculation and distribution map drawing. The following is a detailed description of its implementation process:

[0065] First, the raw vibration signal needs to be preprocessed. These raw data may contain noise or irregular data points, so they must be filtered, denoised, and smoothed to ensure the quality of the basic data for subsequent analysis. This step is crucial to remove unnecessary interference factors, because high-quality data is a prerequisite for accurate evaluation.

[0066] Next, considering that vibration signals are usually continuous time series data, the entire time series can be divided into several shorter time segments (windows) for easier analysis and understanding. The data in each window can be regarded as a set of sample points, which helps to understand the changes in vibration characteristics in the local area. The process of segmentation and windowing not only simplifies the complexity of data analysis, but also makes it possible to capture subtle changes in different time periods in a more detailed manner.

[0067] Once the data is segmented, we can calculate the kurtosis value within each window. Kurtosis is a statistic that measures the "peakedness" of a probability distribution, that is, the sharpness of the distribution curve. For the vibration signal of the affected area, kurtosis can help identify whether there are abnormal high-amplitude events or shocks. These abnormal vibrations may be caused by improper movement, external forces or other adverse factors, which have a potential negative impact on fracture recovery. By applying the statistical formula, the kurtosis value of the sample point in each window can be calculated. :

[0068] ;

[0069] in, is the sample size, Indicates Sample points, is the sample mean, is the sample standard deviation. This formula reflects the probability of extreme values ​​appearing in the data set. A higher kurtosis value means there are more abnormal high-frequency events or shocks.

[0070] The last stage is to draw the distribution map. Determine the frequency axis range and scale according to the sampling frequency of the vibration signal. For example, if the sampling frequency is f, the frequency axis range is 0 to f / 2, and then divide it into several equally spaced intervals, such as an interval of 10Hz. For each time window, the kurtosis value calculated corresponds to the corresponding frequency interval. If the vibration frequency is concentrated in a certain interval, the kurtosis value is high. Then use professional drawing software such as Matlab, Python drawing library, etc., with the frequency interval as the horizontal axis and the kurtosis value as the vertical axis to draw a distribution map of the vibration frequency kurtosis value of the affected area. Through this map, the kurtosis of the vibration signal in different frequency intervals can be intuitively observed, which can help identify abnormal vibration frequency areas.

[0071] Afterwards, in order to further capture and refine the abnormal patterns in the patient's vibration signal to effectively improve the ability to understand and analyze the abnormal vibration patterns of the affected part, in the technical solution of the present application, the vibration frequency kurtosis value distribution map of the affected part is passed through a vibration abnormal pattern feature extractor based on the FCN model to capture vibration features at different scales using multi-level convolutional layers to obtain a semantic representation map of the abnormal vibration pattern of the affected part. In particular, FCN is a deep learning model that can process input images of any size and output feature maps with the same spatial dimension. It can simultaneously capture local details (such as abnormal vibration patterns in a specific area) and overall structures (such as the vibration distribution trend of the entire affected part) in the vibration frequency kurtosis value distribution map, which is crucial for identifying complex vibration anomalies. Moreover, compared with the traditional convolutional neural network (CNN), FCN converts the fully connected layer into a convolutional layer based on the CNN structure, and can process data without losing spatial resolution, that is, the features extracted from the vibration frequency kurtosis value distribution map can correspond more accurately to the position of the original image, making the semantic representation map finally obtained more accurate and intuitive.

[0072] The distribution diagram of the vibration frequency kurtosis value of the affected part is passed through the vibration abnormal pattern feature extractor based on the FCN model to obtain the semantic representation diagram of the vibration abnormal pattern of the affected part, which covers multiple key links such as data preparation and preprocessing, FCN model construction and training, and feature extraction. The following is a detailed description of its implementation process:

[0073] The first is the data preparation and preprocessing stage. To ensure that the data format meets the FCN model input requirements, it may be necessary to resize the original kurtosis value distribution map, such as scaling or cropping it to a fixed size required by the model, such as 224×224 pixels, and converting the data type to a floating point type suitable for calculation. Then the data is normalized, and it is common to map each pixel value of the data to a specific interval, such as the 0-1 interval.

[0074] Then comes the FCN model construction and training phase. It is necessary to select a suitable FCN model architecture according to application requirements and data characteristics, such as FCN-8s, and its different variants differ in network depth and convolutional layer configuration. To train the model, a large amount of labeled data must be prepared, including images of normal and abnormal vibration modes and corresponding annotation information. The annotation information should indicate whether there is an abnormality, type, and location. These data can be obtained by collecting a large amount of vibration data from fracture patients and annotating them by professionals. During the training process, common optimization algorithms such as stochastic gradient descent are used to adjust the model parameters, and regularization methods such as L2 regularization and Dropout are used to prevent overfitting. During the training process, parameters and hyperparameters need to be continuously adjusted to achieve the best results.

[0075] The last stage is feature extraction and semantic representation map generation. After receiving the preprocessed kurtosis value distribution map, the trained FCN model automatically extracts image features. Its multi-level convolutional layer structure can capture features of different scales, local details at the shallow level, and overall structure at the deep level. On this basis, a semantic representation map of the abnormal vibration pattern of the affected part is generated. This map is the result of feature extraction and semantic analysis of the input map, which contains local details and overall structural information and performs semantic abstraction. Each pixel value in the map represents a specific vibration abnormality pattern or feature. By analyzing the pixel value, the abnormal vibration of the affected part can be intuitively understood.

[0076] In an embodiment of the present application, the affected part feature joint encoding unit 133 is used to perform cross-modal prompt significant joint encoding on the patient object blood flow time series feature and the affected part vibration abnormal pattern semantic representation to obtain the patient object blood flow time series-vibration abnormal semantic cross-modal prompt significant joint encoding feature, and based on the patient object blood flow time series-vibration abnormal semantic cross-modal prompt significant joint encoding feature, obtain the fracture rehabilitation monitoring result. It should be understood that the patient object blood flow time series feature and the affected part vibration abnormal pattern semantic representation are data information of different modes, and there are complex inter-correlation characteristics between blood flow and vibration. Therefore, in order to capture the correlation between different modal data and strengthen the important complementary information between the two, so that the finally generated joint encoding feature map not only contains the dynamic changes in the time series, but also incorporates the abnormal distribution in space, in the technical solution of the present application, a cross-modal prompt significant joint encoding mechanism is introduced to process the patient object blood flow time series feature and the affected part vibration abnormal pattern semantic representation to obtain the patient object blood flow time series-vibration abnormal semantic cross-modal prompt significant joint encoding feature. In this way, the weight distribution can be adaptively adjusted to ensure that important features are not overlooked, while noise interference is suppressed, thereby improving sensitivity to subtle but critical changes in patient information.

[0077] Specifically, Figure 3 FIG. 1 is a block diagram of an affected part feature joint encoding unit in a fracture patient rehabilitation intelligent monitoring system according to an embodiment of the present application. Figure 3 As shown, the affected part feature joint encoding unit 133 includes: a blood flow autocorrelation calculation subunit 1331, which is used to calculate the blood flow autocorrelation feature of the blood flow time series feature vector of the patient object; a blood flow-vibration local cross-modal interaction subunit 1332, which is used to perform local cross-modal prompt interaction on the affected part vibration abnormality pattern semantic representation map based on the blood flow autocorrelation feature to obtain the patient object blood flow time series-vibration abnormality semantic cross-modal prompt significant joint encoding feature map as the patient object blood flow time series-vibration abnormality semantic cross-modal prompt significant joint encoding feature; a fracture rehabilitation monitoring result generation subunit 1333, which is used to obtain the fracture rehabilitation monitoring result based on the patient object blood flow time series-vibration abnormality semantic cross-modal prompt significant joint encoding feature map, and the fracture rehabilitation monitoring result is used to indicate whether there is a fracture rehabilitation abnormality.

[0078] In the embodiment of the present application, the blood flow autocorrelation calculation subunit 1331 is used to calculate the blood flow autocorrelation feature of the blood flow time series feature vector of the patient object. Specifically, in the embodiment of the present application, the blood flow autocorrelation calculation subunit 1331 is used to: calculate the product between the blood flow time series feature vector of the patient object and the transposed vector of the blood flow time series feature vector of the patient object to obtain the blood flow autocorrelation encoding matrix of the patient object as the blood flow autocorrelation feature. The above process can be expressed as:

[0079] ;

[0080] in, is the time series feature vector of the patient's blood flow, is matrix multiplication, for The transposed vector of is the patient object blood flow autocorrelation encoding matrix.

[0081] Here, by calculating the product between the blood flow time series feature vector of the patient object and the transposed vector of the blood flow time series feature vector of the patient object, the correlation between the blood flow features of the patient object at different time points can be revealed. That is, the blood flow autocorrelation coding matrix of the patient object obtained by this calculation method not only contains the information of the original blood flow time series features of the patient object, but also adds the correlation measurement information between the blood flow time series features, which can lay the foundation for the subsequent feature interaction of the vibration signal.

[0082] In an embodiment of the present application, the blood flow-vibration local cross-modal interaction subunit 1332 is used to perform local cross-modal prompt interaction on the affected part vibration abnormality pattern semantic representation map based on the blood flow autocorrelation feature to obtain the patient object blood flow timing-vibration abnormality semantic cross-modal prompt significant joint coding feature map as the patient object blood flow timing-vibration abnormality semantic cross-modal prompt significant joint coding feature. Specifically, in an embodiment of the present application, the blood flow-vibration local cross-modal interaction subunit 1332 includes: an affected part vibration abnormality pattern feature decoupling secondary subunit, used to perform feature decoupling along the channel dimension on the affected part vibration abnormality pattern semantic representation map to obtain a set of affected part vibration abnormality pattern semantic local feature matrices; a patient object blood flow feature linear transformation secondary subunit, used to perform linear transformation on the patient object blood flow autocorrelation coding matrix to obtain a patient object blood flow semantic query coding matrix and a patient object blood flow semantic value coding matrix; a blood flow-vibration cross-modal semantic feature generation secondary subunit, used to generate a blood flow map of the patient object blood flow. Each affected part vibration abnormal pattern semantic local feature matrix in the set of the semantic query encoding matrix, the patient object blood flow semantic value encoding matrix and the affected part vibration abnormal pattern semantic local feature matrix is ​​respectively subjected to Transformer-based cross-modal prompt information encoding to obtain a set of blood flow-vibration semantic cross-modal prompt information semantic encoding matrices; a blood flow-vibration feature weighted attention interaction secondary subunit is used to perform gated weighted attention interaction optimization on the set of blood flow-vibration semantic cross-modal prompt information semantic encoding matrices to obtain the patient object blood flow timing-vibration abnormality semantic cross-modal prompt significant joint encoding feature map.

[0083] Specifically, first, the semantic representation graph of the abnormal vibration pattern of the affected part is subjected to feature decoupling along the channel dimension to obtain a set of semantic local feature matrices of the abnormal vibration pattern of the affected part. The above process can be expressed as:

[0084] ;

[0085] in, is a semantic representation diagram of the abnormal vibration pattern of the affected part, For Perform feature decoupling operations. , and They are the first and the second in the set of semantic local feature matrices of abnormal vibration patterns of the affected area. and The semantic local feature matrix of abnormal vibration patterns of the affected area.

[0086] It should be understandable that the features of the original semantic representation map of the abnormal vibration pattern of the affected area are mixed with each other, and it is difficult for the model to focus on these local details, but the features on the local details often contain key information. Here, by decoupling the features of the semantic representation map of the abnormal vibration pattern of the affected area along the channel dimension, the model's attention to the local details of the original semantic representation map can be improved, and its own feature analysis can be performed more accurately. Moreover, the decoupling operation can make the characteristics of the vibration signal itself clearer and more organized. When it is subsequently interactively fused with the blood flow timing characteristics, it can enable the key information at the corresponding time points to be better combined with each other, which is conducive to improving the overall data interaction effect.

[0087] Next, the patient object blood flow autocorrelation coding matrix is ​​linearly transformed to obtain the patient object blood flow semantic query coding matrix and the patient object blood flow semantic value coding matrix. The above process can be expressed as:

[0088] ;

[0089] ;

[0090] in, is the patient blood flow autocorrelation encoding matrix, is the query embedding matrix, is the patient object blood flow semantic query encoding matrix, is the value embedding matrix, It is the semantic value encoding matrix of the patient object’s blood flow.

[0091] That is, by performing a linear transformation on the patient object blood flow autocorrelation coding matrix, the original coding matrix can be converted into a form suitable for calculation by the subsequent attention mechanism. Here, the converted patient object blood flow semantic query coding matrix is ​​used to find the correlation with the vibration signal characteristics, while the patient object blood flow semantic value coding matrix is ​​used to preserve the important information of the original blood flow time series characteristics. It is worth mentioning that the linear transformation is not just a simple mathematical operation, but by introducing trainable parameters to adjust the spatial distribution of the blood flow time series feature representation, thereby making the converted blood flow time series feature representation more suitable for cross-modal fusion.

[0092] Then, each affected part vibration abnormal pattern semantic local feature matrix in the set of the patient object blood flow semantic query encoding matrix, the patient object blood flow semantic value encoding matrix and the affected part vibration abnormal pattern semantic local feature matrix is ​​respectively subjected to Transformer-based cross-modal prompt information encoding to obtain a set of blood flow-vibration semantic cross-modal prompt information semantic encoding matrices. The above process can be expressed as:

[0093] ;

[0094] in, is the patient object blood flow semantic query encoding matrix, is the patient object blood flow semantic value encoding matrix, yes The transposed matrix of yes The scale is the matrix width multiplied by the matrix height, yes function, yes and The semantic encoding matrix of blood flow-vibration semantic cross-modal cue information, is matrix multiplication.

[0095] That is, by performing Transformer-based cross-modal prompt information encoding on each affected part vibration abnormality pattern semantic local feature matrix in the set of the patient object blood flow semantic query encoding matrix, the patient object blood flow semantic value encoding matrix, and the affected part vibration abnormality pattern semantic local feature matrix, the self-attention mechanism of the transformer structure can be used to effectively capture the deep semantic relationship between the two different modal information of the patient's blood flow and the affected part vibration signal. The set of blood flow-vibration semantic cross-modal prompt information semantic encoding matrices generated in this way not only retains the information of the original blood flow time series features, but also integrates the supplementary information from the vibration signal features, which is conducive to improving the richness and accuracy of feature representation.

[0096] More specifically, in the embodiment of the present application, the blood flow-vibration feature weight attention interaction secondary subunit is used to: decode and gate the set of blood flow-vibration semantic cross-modal prompt information semantic coding matrices to obtain a set of blood flow-vibration cross-modal semantic interaction attention weights; multiply the patient object blood flow autocorrelation coding matrix by each affected part vibration abnormality pattern semantic local feature matrix in the set of affected part vibration abnormality pattern semantic local feature matrices by position point to obtain a set of blood flow-vibration cross-modal interaction local feature matrices; cross-modally jointly encode the set of blood flow-vibration cross-modal semantic interaction attention weights and the set of blood flow-vibration cross-modal interaction local feature matrices to obtain the patient object blood flow timing-vibration abnormality semantic cross-modal prompt significant joint encoding feature map. The above process can be expressed by the formula:

[0097] ;

[0098] ;

[0099] in, yes function, is the decoding weight matrix, It is a decoder. yes and The semantic encoding matrix of blood flow-vibration semantic cross-modal cue information, It is the point multiplication by position. , and They are the first and the second in the set of semantic local feature matrices of abnormal vibration patterns of the affected area. and The semantic local feature matrix of abnormal vibration pattern of the affected part, , , They are The corresponding blood flow-vibration cross-modal semantic interaction attention weights, is the characteristic coupling along the channel dimension, It is a significant joint encoding feature map of semantic cross-modal cues of abnormal blood flow timing-oscillation of the patient.

[0100] It should be understood that by performing decoding gating processing on the set of semantic coding matrices of the blood flow-vibration semantic cross-modal prompt information, the importance of different modal features in the original coding matrix set can be adjusted, that is, by calculating the interactive attention weight to determine which information is more critical. In this way, it helps the model to allocate attention resources more flexibly when processing data, avoiding over-reliance on a certain modal information and ignoring other clues that may be equally important. In this way, the model can more accurately capture the complementarity and synergy between the two different modal data of blood flow and vibration signal. Afterwards, the blood flow autocorrelation coding matrix of the patient object is respectively multiplied by position with each affected part vibration abnormality mode semantic local feature matrix in the set of the affected part vibration abnormality mode semantic local feature matrix. The calculation method of position point multiplication can directly realize the direct interaction between the blood flow time series features and the vibration signal semantic features, and then effectively capture the complementary information of the blood flow data and the vibration signal data at the same spatial position, thereby enhancing the spatial consistency of the obtained interactive local feature representation. Finally, the set of the blood flow-vibration cross-modal semantic interaction attention weights and the set of the blood flow-vibration cross-modal interaction local feature matrices are cross-modally jointly encoded. Here, by combining the attention weights and the interactive local features, the weighted fusion of the blood flow timing features and the vibration signal semantic features can be achieved. The generated patient blood flow timing-vibration abnormal semantic cross-modal prompt significant joint encoding feature map has stronger expressive power and less redundant information. It not only maintains the fidelity to the original input data, but also can better serve the subsequent classification operations.

[0101] In an embodiment of the present application, the fracture rehabilitation monitoring result generation subunit 1333 is used to obtain the fracture rehabilitation monitoring result based on the patient object blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature map, and the fracture rehabilitation monitoring result is used to indicate whether there is an abnormal fracture rehabilitation situation. Specifically, in an embodiment of the present application, the fracture rehabilitation monitoring result generation subunit 1333 is used to: pass the patient object blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature map through a classifier-based rehabilitation status monitor to obtain the fracture rehabilitation monitoring result, and the fracture rehabilitation monitoring result is used to indicate whether there is an abnormal fracture rehabilitation situation. It should be understood that the patient object blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature map integrates the key information of two different modalities, blood flow timing data and affected area vibration signal. In order to judge the patient's fracture rehabilitation status based on the key information in the feature map, it is necessary to input the patient object blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature map into the classifier-based rehabilitation status monitor. The classifier is a model built based on a deep learning algorithm. By training a large amount of labeled (normal rehabilitation or abnormal rehabilitation) data, it can learn the difference patterns of data under normal rehabilitation status and abnormal rehabilitation status, and then can perform classification and discrimination when inputting a new joint coding feature map, realizing the transformation from complex data features to clear rehabilitation status judgment. The fracture rehabilitation monitoring results obtained are used to indicate whether there is abnormal fracture rehabilitation, which directly guides clinical decision-making and the formulation of intervention measures, so as to better meet the actual needs of patients, improve rehabilitation effects, and promote patients to recover faster and better.

[0102] Here, considering that the patient's blood flow timing feature vector and the affected part vibration abnormality pattern semantic representation map respectively represent the one-dimensional timing correlation characteristics of the fracture patient's blood flow and the image semantic characteristics of the kurtosis value distribution map of the affected part vibration signal, when performing cross-modal interactive optimization encoding based on cross-modal prompt gates, insufficient correspondence of the prompt gating of cross-modal features will also lead to the lack of aggregated feature instance determination of the patient's blood flow timing-vibration abnormality semantic cross-modal prompt significant joint encoding feature map, thereby affecting the accuracy of the fracture rehabilitation monitoring results obtained through the classifier-based rehabilitation status monitor.

[0103] Preferably, the blood flow time series-vibration abnormal semantic cross-modal prompt significant joint encoding feature map of the patient subject is passed through a classifier-based rehabilitation status monitor to obtain a fracture rehabilitation monitoring result, including:

[0104] Expanding the patient object blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature map into a patient object blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature vector;

[0105] The distance between each pair of eigenvalues ​​of the patient object blood flow timing series-vibration abnormal semantic cross-modal prompt significant joint coding feature vector is calculated, such as the L2 distance, and the square root of the distance is taken to obtain the patient object blood flow timing series-vibration abnormal semantic cross-modal prompt significant joint coding distance representation matrix. The process can be expressed as:

[0106] ;

[0107] in, Represents the significant joint encoding feature vector of the patient's blood flow time series-vibration abnormal semantic cross-modal prompts, The first The eigenvalues ​​at the positions, The first The eigenvalues ​​at the positions, express and The distance between express and The square root of the distance between

[0108] The patient blood flow time series-vibration abnormal semantic cross-modal prompt significant joint coding autocorrelation matrix of the patient blood flow time series-vibration abnormal semantic cross-modal prompt significant joint coding feature vector is obtained. The process can be expressed as:

[0109] ;

[0110] in, represents the significant joint encoding feature vector of the patient's blood flow timing-vibration abnormal semantic cross-modal prompts, and is a row vector, express The transposed vector of The autocorrelation matrix representing the significant joint encoding of the semantic cross-modal cues of blood flow timing-oscillation abnormality in the patient subject;

[0111] The patient blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature vector is matrix-multiplied with the patient blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding distance representation matrix to obtain the patient blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding primary mapping vector, and the process can be expressed as:

[0112] ;

[0113] in, Represents the significant joint encoding feature vector of the patient's blood flow time series-vibration abnormal semantic cross-modal prompts, represents matrix multiplication, Represents the significant joint encoding distance representation matrix of the patient's blood flow time series-vibration abnormal semantic cross-modal prompts, Represents the first-level mapping vector of the significant joint encoding of the semantic cross-modal cues of blood flow time series-vibration abnormality in the patient subject;

[0114] The first-level mapping vector of the significant joint coding of the abnormal semantic cross-modal prompt of blood flow timing series-vibration of the patient object is matrix-multiplied with the matrix product of the significant joint coding distance representation matrix of the abnormal semantic cross-modal prompt of blood flow timing series-vibration of the patient object and the autocorrelation matrix of the significant joint coding of the abnormal semantic cross-modal prompt of blood flow timing series-vibration of the patient object to obtain the multi-level mapping vector of the significant joint coding of the abnormal semantic cross-modal prompt of blood flow timing series-vibration of the patient object, and the process is expressed as:

[0115] ;

[0116] in, Indicates the first-level mapping vector of the significant joint encoding of the patient's blood flow timing-vibration abnormal semantic cross-modal prompts, represents matrix multiplication, Represents the significant joint encoding distance representation matrix of the patient's blood flow time series-vibration abnormal semantic cross-modal prompts, The autocorrelation matrix representing the significant joint encoding of the patient's blood flow time series-vibration abnormal semantic cross-modal prompts, Represents a multi-level mapping vector of blood flow time series-vibration abnormal semantic cross-modal prompts that are significantly jointly encoded;

[0117] The patient blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding multi-level mapping vector is interpolated with the patient blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding associated eigenvector (if the eigenvalue is insufficient, interpolation or zero filling) composed of the eigenvalue of the patient blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding autocorrelation matrix to obtain an optimized patient blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature vector;

[0118] The optimized patient blood flow timing-vibration abnormal semantic cross-modal prompt significant joint encoding feature vector is passed through the classifier-based rehabilitation status monitor to obtain the fracture rehabilitation monitoring result.

[0119] That is, through the linear target mapping representation of the similarity distance representation matrix of the patient's blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature vector after the patient's blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature map is expanded, the self-correlation complete similarity instantiation of the patient's blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature map is performed based on a secondary target mapping representation of the multi-level distribution hierarchy, and the association mismatch negative influence factor is compensated by the association fusion kernel bias to improve the eigenvalue instance judgment degree of the patient's blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature map under similarity constraints, that is, the significance degree of the eigenvalue as an instance for classification regression judgment, so as to improve the accuracy of the fracture rehabilitation monitoring result obtained by the classifier-based rehabilitation status monitor of the patient's blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature map.

[0120] In summary, the patient data analysis module 130 is clearly explained, which uses machine learning-based data analysis and signal processing technology to perform time series feature analysis on the blood flow of the fracture patient object, and at the same time performs kurtosis distribution calculation and vibration anomaly analysis on the vibration signal of the affected part, so as to intelligently judge the abnormal situation of fracture rehabilitation based on the cross-modal prompt coding representation between the time series characteristics of the patient's blood flow and the semantic representation of the abnormal vibration pattern of the affected part. In this way, the unique health pattern of each patient can be identified based on their unique characteristics and dynamic changes during the rehabilitation process, and the evaluation criteria can be optimized accordingly. In addition, this method can also provide a more comprehensive understanding of the patient's state changes, and can also provide more accurate evaluation and support for patients with complex rehabilitation processes or special situations that are difficult to predict with established rules.

[0121] In an embodiment of the present application, the early warning module 140 is used to generate an early warning signal based on the fracture rehabilitation monitoring results. It should be understood that during the fracture rehabilitation process, the patient may not be able to accurately perceive the subtle changes in his or her condition at all times, and the medical team cannot manually monitor the data of each patient in real time. By automatically generating an early warning signal, once the monitoring system analyzes and finds that the patient has abnormal rehabilitation conditions, an early warning can be triggered in time. In this way, the medical staff responsible for the patient's rehabilitation treatment can be informed of the patient's possible risks at the first time, and take corresponding measures in time to avoid missing the best time for intervention due to delayed information transmission. And from the perspective of medical management, early warning signals help optimize the allocation of medical resources. Through accurate and timely early warnings, the medical team can reasonably allocate medical resources according to the actual needs of the patient, avoid frequent indiscriminate inspections and interventions on all patients, and then focus more energy and resources on patients who really need attention and treatment.

[0122] In the embodiment of the present application, the result display module 150 is used to display the fracture rehabilitation monitoring results and the warning signal. It should be understood that displaying complex monitoring results and warning signals in a visual way can convert data into intuitive and easy-to-understand graphics and text prompts, which is convenient for medical personnel to quickly obtain key information. For example, by displaying warning signals in eye-catching colors or icons (such as flashing red to indicate serious abnormalities), relevant personnel can see at a glance whether the patient's rehabilitation status is normal and whether there are problems that need attention, so there is no need to spend a lot of time and energy to interpret the original data, thereby improving information processing efficiency. Moreover, the displayed monitoring results and warning signals can be saved as data records, which is convenient for subsequent data tracing and analysis. At different stages of patient rehabilitation, medical personnel can review previous monitoring data and warning situations, analyze the changing trends in the rehabilitation process, summarize and analyze and make relevant rehabilitation plan adjustments, thereby promoting the patient's fracture rehabilitation process.

[0123] In summary, the fracture patient rehabilitation intelligent monitoring system 100 based on the embodiment of the present application is explained, which first collects the patient's blood flow and affected area vibration data through the blood flow sensor and the vibration sensor, and uploads these data to the cloud server for storage, and then retrieves these data from the cloud server for analysis to monitor the patient's fracture rehabilitation, and generates an early warning signal based on the obtained monitoring results, and finally displays the monitoring results and early warning signals. In this way, it is possible to more accurately monitor the patient's fracture rehabilitation status.

[0124] In an embodiment of the present application, a method for intelligently monitoring the rehabilitation of fracture patients is also provided, which is used to execute the intelligent monitoring system for the rehabilitation of fracture patients as described above. Figure 4 FIG. 1 is a flow chart of a method for intelligently monitoring the rehabilitation of fracture patients according to an embodiment of the present application. Figure 4As shown, in the intelligent monitoring method for the rehabilitation of fracture patients, the method includes: S110, respectively collecting a blood flow time series data set and an affected part vibration signal of a fracture patient object through a blood flow sensor and a vibration sensor; S120, transmitting the data collected by the patient data acquisition module to a cloud server for storage; S130, retrieving the stored blood flow time series data set and the affected part vibration signal from the cloud server and analyzing them to obtain a fracture rehabilitation monitoring result; S140, generating an early warning signal based on the fracture rehabilitation monitoring result; S150, displaying the fracture rehabilitation monitoring result and the early warning signal, wherein step S1 130, including: performing time series coding based on the patient's blood flow on the blood flow time series data set to obtain the patient's blood flow time series characteristics; performing kurtosis value distribution calculation and vibration abnormality pattern feature extraction on the affected part vibration signal to obtain the affected part vibration abnormality pattern semantic representation; performing cross-modal prompt significant joint coding on the patient's blood flow time series characteristics and the affected part vibration abnormality pattern semantic representation to obtain the patient's blood flow time series-vibration abnormality semantic cross-modal prompt significant joint coding features, and based on the patient's blood flow time series-vibration abnormality semantic cross-modal prompt significant joint coding features, obtain the fracture rehabilitation monitoring results.

[0125] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned fracture patient rehabilitation intelligent monitoring method have been referred to above. Figures 1 to 3 The invention has been introduced in detail in the description of the intelligent monitoring system for rehabilitation of fracture patients, and therefore, its repeated description will be omitted.

[0126] In summary, the intelligent monitoring method for fracture patient rehabilitation based on the embodiment of the present application is explained, which first collects the patient's blood flow and affected area vibration data through a blood flow sensor and a vibration sensor, and uploads these data to a cloud server for storage, and then retrieves these data from the cloud server for analysis to monitor the patient's fracture rehabilitation, and generates an early warning signal based on the obtained monitoring results, and finally displays the monitoring results and early warning signals. In this way, it is possible to more accurately monitor the patient's fracture rehabilitation status.

Claims

1. An intelligent monitoring system for the rehabilitation of fracture patients, characterized in that: include: A patient data acquisition module, used to respectively acquire a blood flow time series data set and an affected part vibration signal of a fracture patient object through a blood flow sensor and a vibration sensor; A patient data transmission and storage module, used to transmit the data collected by the patient data collection module to a cloud server for storage; A patient data analysis module, used for retrieving the stored blood flow time series data set and the affected part vibration signal from the cloud server and analyzing them to obtain fracture rehabilitation monitoring results; An early warning module, used for generating an early warning signal based on the fracture rehabilitation monitoring result; A result display module is used to display the fracture rehabilitation monitoring result and the warning signal; wherein the patient data analysis module includes: A patient blood flow encoding unit, used for performing time series encoding based on the patient blood flow on the blood flow time series data set to obtain a patient object blood flow time series feature; An affected part vibration feature encoding unit, used for calculating the kurtosis value distribution and extracting the vibration abnormal pattern feature of the affected part vibration signal to obtain a semantic representation of the affected part vibration abnormal pattern; The affected part feature joint encoding unit is used to perform cross-modal prompt significant joint encoding on the patient object's blood flow timing characteristics and the affected part vibration abnormality pattern semantic representation to obtain the patient object's blood flow timing-vibration abnormality semantic cross-modal prompt significant joint encoding features, and based on the patient object's blood flow timing-vibration abnormality semantic cross-modal prompt significant joint encoding features, obtain the fracture rehabilitation monitoring results.

2. The intelligent monitoring system for fracture patient rehabilitation according to claim 1, characterized in that: The patient blood flow encoding unit is used to: pass the blood flow time series data set through a patient blood flow time series feature encoder based on a one-dimensional causal convolutional neural network model to obtain a patient object blood flow time series feature vector as the patient object blood flow time series feature.

3. The intelligent monitoring system for fracture patient rehabilitation according to claim 2, characterized in that: The affected part vibration characteristic encoding unit comprises: The affected part vibration signal kurtosis value distribution calculation subunit is used to calculate the kurtosis value distribution diagram of the affected part vibration signal to obtain the affected part vibration frequency kurtosis value distribution diagram; The affected part vibration abnormal pattern feature extraction subunit is used to pass the affected part vibration frequency kurtosis value distribution diagram through the vibration abnormal pattern feature extractor based on the FCN model to obtain the affected part vibration abnormal pattern semantic representation diagram as the affected part vibration abnormal pattern semantic representation.

4. The intelligent monitoring system for fracture patient rehabilitation according to claim 3, characterized in that: The affected part feature joint encoding unit comprises: A blood flow autocorrelation calculation subunit, used to calculate the blood flow autocorrelation characteristics of the blood flow time series feature vector of the patient object; A blood flow-vibration local cross-modal interaction subunit is used to perform local cross-modal prompt interaction on the affected part vibration abnormality pattern semantic representation map based on the blood flow autocorrelation feature to obtain a patient blood flow time series-vibration abnormality semantic cross-modal prompt significant joint coding feature map as the patient blood flow time series-vibration abnormality semantic cross-modal prompt significant joint coding feature; The fracture rehabilitation monitoring result generating subunit is used to obtain the fracture rehabilitation monitoring result based on the patient's blood flow timing-vibration abnormal semantic cross-modal prompt significant joint coding feature map, and the fracture rehabilitation monitoring result is used to indicate whether there is an abnormal fracture rehabilitation situation.

5. The intelligent monitoring system for fracture patient rehabilitation according to claim 4, characterized in that: The blood flow autocorrelation calculation subunit is used to calculate the product between the patient object blood flow time series feature vector and the transposed vector of the patient object blood flow time series feature vector to obtain the patient object blood flow autocorrelation coding matrix as the blood flow autocorrelation feature.

6. The intelligent monitoring system for fracture patient rehabilitation according to claim 5, characterized in that: The blood flow-vibration local cross-modal interaction subunit comprises: A secondary subunit for decoupling the feature of the abnormal vibration pattern of the affected part is used to perform feature decoupling along the channel dimension on the semantic representation graph of the abnormal vibration pattern of the affected part to obtain a set of semantic local feature matrices of the abnormal vibration pattern of the affected part; A patient object blood flow feature linear transformation secondary subunit, used for linearly transforming the patient object blood flow autocorrelation coding matrix to obtain a patient object blood flow semantic query coding matrix and a patient object blood flow semantic value coding matrix; A secondary subunit for generating blood flow-vibration cross-modal semantic features is used to perform Transformer-based cross-modal prompt information encoding on each affected part vibration abnormality pattern semantic local feature matrix in the set of the patient object blood flow semantic query encoding matrix, the patient object blood flow semantic value encoding matrix, and the affected part vibration abnormality pattern semantic local feature matrix to obtain a set of blood flow-vibration semantic cross-modal prompt information semantic encoding matrices; The blood flow-vibration feature weighted attention interaction secondary subunit is used to perform gated weighted attention interaction optimization on the set of semantic coding matrices of the blood flow-vibration semantic cross-modal prompt information to obtain the significant joint coding feature map of the patient's blood flow timing-vibration abnormal semantic cross-modal prompt.

7. The intelligent monitoring system for fracture patient rehabilitation according to claim 6, characterized in that: The blood flow-vibration feature weight attention interaction secondary subunit is used to: Performing decoding gating processing on the set of blood flow-vibration semantic cross-modal prompt information semantic encoding matrices to obtain a set of blood flow-vibration cross-modal semantic interaction attention weights; Multiplying the blood flow autocorrelation coding matrix of the patient object by each affected part vibration abnormality pattern semantic local feature matrix in the set of affected part vibration abnormality pattern semantic local feature matrices by position point to obtain a set of blood flow-vibration cross-modal interaction local feature matrices; The set of the blood flow-vibration cross-modal semantic interaction attention weights and the set of the blood flow-vibration cross-modal interaction local feature matrices are cross-modally jointly encoded to obtain the patient subject's blood flow timing-vibration abnormal semantic cross-modal prompt significant joint encoding feature map.

8. The intelligent monitoring system for fracture patient rehabilitation according to claim 7, characterized in that: The fracture rehabilitation monitoring result generating subunit is used for: passing the patient's blood flow timing-vibration abnormal semantic cross-modal prompt significant joint encoding feature map through a classifier-based rehabilitation status monitor to obtain the fracture rehabilitation monitoring result, and the fracture rehabilitation monitoring result is used to indicate whether there is an abnormal fracture rehabilitation situation.

9. A method for intelligent monitoring of the rehabilitation of fracture patients, used to implement the intelligent monitoring system for the rehabilitation of fracture patients as claimed in claim 1, characterized in that: The intelligent monitoring method for fracture patient rehabilitation comprises: respectively collecting a blood flow time series data set and an affected part vibration signal of a fracture patient object through a blood flow sensor and a vibration sensor; transmitting the data collected by the patient data acquisition module to a cloud server for storage; retrieving the stored blood flow time series data set and the affected part vibration signal from the cloud server and analyzing them to obtain a fracture rehabilitation monitoring result; generating an early warning signal based on the fracture rehabilitation monitoring result; and displaying the fracture rehabilitation monitoring result and the early warning signal; The step of retrieving the stored blood flow time series data set and the affected part vibration signal from the cloud server and analyzing them to obtain the fracture rehabilitation monitoring result includes: Performing time series encoding based on the patient's blood flow on the blood flow time series data set to obtain a blood flow time series feature of the patient object; Calculating the distribution of kurtosis values ​​and extracting the features of abnormal vibration patterns on the affected part vibration signal to obtain semantic representation of the abnormal vibration pattern of the affected part; The patient's blood flow timing characteristics and the affected area vibration abnormality pattern semantic representation are cross-modal significantly jointly encoded to obtain the patient's blood flow timing-vibration abnormality semantic cross-modal significantly jointly encoded features, and based on the patient's blood flow timing-vibration abnormality semantic cross-modal significantly jointly encoded features, the fracture rehabilitation monitoring results are obtained.

10. The intelligent monitoring method for fracture patient rehabilitation according to claim 9, characterized in that: The blood flow time series data set is time-series encoded based on the patient's blood flow to obtain the patient's blood flow time series characteristics, including: the blood flow time series data set is passed through a patient's blood flow time series characteristic encoder based on a one-dimensional causal convolutional neural network model to obtain a patient's blood flow time series characteristic vector as the patient's blood flow time series characteristic.

Images