Rehabilitation prediction device based on multi-source heterogeneous large model

Abstract

The application discloses a rehabilitation prediction device based on a multi-source heterogeneous large model, relates to the field of medical auxiliary devices, and solves the problems of low measurement accuracy of existing wearable devices and the lack of matching analysis software. The device is wearable, contains a data acquisition shell, a camera and other structures, and a shoulder strap can be detached to adapt to different body types. A piezoelectric film, a temperature sensor and other sensors are arranged in the shell, and a magnetometer, an accelerometer and other components are matched, so that multi-dimensional time sequence sensing data such as breathing, electrocardio, body temperature and motion state can be collected for 24 hours. After pre-processing, the multi-source heterogeneous time sequence data is input into a prediction model to obtain a prediction result. The device is matched with exclusive software, which can read patient cases, examination and diagnosis data, combine monitoring and prediction data, generate a rehabilitation prescription report through a built-in large language model, and wirelessly transmit relevant data and the report to a terminal such as a mobile phone or a computer, so that the rehabilitation direction can be adjusted in time, and a complete rehabilitation guidance system is formed.

Description

Technical Field

[0001] This invention relates to the field of medical assistive devices, specifically a rehabilitation prediction device based on a multi-source heterogeneous large model. Background Technology

[0002] Currently, with the increasing health awareness of people, the level of attention to physical health is also increasing. However, due to the limited capacity of the current medical system and the relatively limited professionalism and medical level of different medical systems in the treatment of diseases, patients are unable to receive timely and effective medical treatment in medical institutions. On the other hand, in daily life, users and medical institutions are also unable to grasp users' physical health information in a timely and effective manner, and therefore cannot provide users with targeted and effective medical assistance, plans and health management advice, thus delaying the diagnosis and treatment of diseases. In the current technology of wearable devices, commonly used wearable devices for monitoring electrocardiogram (ECG) signals mainly include wristbands, watches, heart rate belts, and rings, which primarily measure physiological indicators such as pulse rate, heart rate, blood oxygen, blood pressure, and temperature. Although wearable devices have many advantages, such as convenience, their size and operational limitations restrict the provision of complex interactive functions, resulting in relatively limited functional expansion. Moreover, because different wearable devices use different sensors for measurement, including optical sensors, bioelectric potential energy sensors, and bioimpedance sensors, and different sensors are used with different calculation models, the measurement results can vary greatly between wearable devices worn in the same location. This limits the accuracy and precision of the measurements, failing to meet the precision requirements of medical devices. To address these issues, this patent proposes a rehabilitation prediction device based on a multi-source heterogeneous large model. Summary of the Invention

[0003] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a rehabilitation prediction device based on a multi-source heterogeneous large model. This solves the problem that in existing technologies, different sensors combined with different computational models result in large differences in measurement results between wearable devices worn in the same location, leading to limited measurement accuracy and precision.

[0004] (II) Technical Solution To achieve the above objectives, the present invention is implemented through the following technical solution: a rehabilitation prediction device based on a multi-source heterogeneous large model, comprising a data acquisition housing and a camera, wherein a restraint strap is fixedly connected to both ends of the data acquisition housing, and a connector is fixedly connected to both ends of the upper part of the restraint strap, wherein a limit groove is opened inside the connector, a limit block is installed inside the limit groove, and a shoulder strap is fixedly connected to the top of the limit block. The connector has slots at both ends on the outer side. A spring is fixedly connected to one end of the inner side of the slot. A push plate is fixedly connected to the other end of the spring. A positioning post is fixedly connected to the other end of the push plate. The positioning post is movably connected inside the limiting slot. A positioning hole is opened through the inside of the limiting block. The positioning post and the positioning hole correspond to each other. A piezoelectric thin film sensor is provided at the inner end of the data acquisition housing, and temperature sensors are provided on both sides away from the piezoelectric thin film sensor. ECG lead electrodes are provided on the shoulder strap, and a connecting shell is installed at the outer end of the data acquisition housing.

[0005] Preferably, a data acquisition device is installed inside the data acquisition housing, and a magnetometer is installed inside the connecting housing.

[0006] Preferably, an accelerometer is installed on one side of the magnetometer, and an integrated gyroscope is installed on one side of the accelerometer. The magnetometer, accelerometer, and integrated gyroscope are all electrically connected to the data acquisition unit.

[0007] Preferably, a timing sensor is electrically connected to one side of the data acquisition unit, and a wireless transmission module is electrically connected to one side of the timing sensor.

[0008] Preferably, the camera is mounted on the wearable watch case, and an image processing algorithm module is electrically connected inside the camera. The other end of the image processing algorithm module is also electrically connected to the wireless transmission module.

[0009] Preferably, the other end of the wireless transmission module is connected to the display device using WIFI and wireless remote technology, and the rehabilitation prescription report generated by the dedicated software can be synchronized to the display device for display.

[0010] Working principle: In use, the restraint strap 4 can be aligned with the wearer's waist. Then, select the appropriate shoulder strap 2 according to the body size, align the limiting blocks 12 at both ends of the shoulder strap 2 with the limiting groove 10, and then pull the push plate 8 outward. Under the action of the spring 7, the push plate 8 can pull the positioning column 6 inward, so that the positioning column 6 is retracted into the slot 9. At this time, the user inserts the limiting block 12 directly into the limiting groove 10, releases the push plate 8 at both ends, and under the action of the spring 7, the push plate 8 and the positioning column 6 can be driven to quickly reset and engage with the positioning hole 11 inside the limiting block 12, thus completing the installation of the shoulder strap 2. This allows the entire rehabilitation prediction device to fit the human body for real-time monitoring. During monitoring, the data acquisition housing 1 comes into contact with the user's skin, aligning the piezoelectric film sensor 14 with the location of the lungs. The temperature sensor 15 collects the user's body temperature, and the piezoelectric film sensor 14 collects data on the user's lung breathing. The wireless transmission module 21 transmits the data collected by each sensor to the caregiver's mobile phone, cloud, PC, or other display devices. The caregiver can understand the basic condition of the user's respiratory diseases and physiological symptoms such as lung sounds based on the data collected by the piezoelectric film sensor 14. The caregiver can understand the user's body temperature changes and duration based on the data collected by the temperature sensor 21. The integrated gyroscope 18, accelerometer 18, and magnetometer 17 inside the housing 5 provide sensing data through a built-in computing library. The study uses device posture data to record sensor data of the patient model on different motion trajectories, as well as changes in body posture during specific tasks. Through accelerometer 18 and integrated gyroscope 19, the study can monitor the patient's motion status in real time, including motion trajectory, gait, posture, and activity intensity, and monitor physiological parameters such as heart rate, respiratory rate, and body temperature. Visual tracking technology monitors the patient's motion through camera 23 and image processing algorithm module 22. The development of high-resolution cameras can capture subtle motion details to ensure tracking accuracy. Through structured light cameras, three-dimensional spatial information is obtained to help analyze the patient's posture and movements. Image processing algorithm module 22 (such as edge detection and feature extraction) is used to identify and analyze the patient's movements. The data collected above can facilitate doctors' diagnosis during online consultations. During data collection, the data acquisition device 16 detects multiple indicators of the wearer and transmits the data to the time-series sensor 20 for multi-source preprocessing to generate multi-source heterogeneous time-series data. This multi-source heterogeneous time-series data is then input via the wireless transmission module 21 into a multi-dimensional, multi-modal collaborative prediction model established using a deterministic learning method and dynamic environment, generating multiple prediction results. These prediction results are compared with the user's historical data. Simultaneously, all the monitoring data and prediction comparison results are transmitted via the wireless transmission module 21 to the accompanying dedicated software. This software can read the patient's medical records, examination data, and diagnostic data, and, combined with the received device monitoring data, performs comprehensive analysis using a built-in large language model to generate a personalized rehabilitation prescription report. The monitoring comparison data and rehabilitation prescription report can be directly sent to a mobile phone or computer. By using the monitoring data and rehabilitation prescription report, the subsequent rehabilitation direction can be adjusted in a timely manner, realizing the rehabilitation prediction effect of the entire multi-source heterogeneous large model.

[0011] (III) Beneficial Effects This invention provides a rehabilitation prediction device based on a multi-source heterogeneous large model, which has the following beneficial effects: 1. This invention makes the device wearable, with multiple monitoring modules integrated on the wearable device. The detachable shoulder strap design allows for use by wearers of different body shapes. Multiple internal sensors work together to acquire multiple time-series sensor data. These multiple time-series sensor data are preprocessed from multiple sources to generate multi-source heterogeneous time-series data. The cooperation between multiple sensors can continuously acquire real-time data of the user's physiological characteristics such as respiration, electrocardiogram, and body temperature 24 hours a day. The data obtained by comparing multiple prediction results with the user's historical data can be directly sent to a mobile phone or computer. By monitoring the data, the subsequent rehabilitation direction can be adjusted in a timely manner. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the overall invention.

[0013] Figure 2 This is a schematic diagram of the shoulder strap installation according to the present invention.

[0014] Figure 3 This is a rear view of the present invention; Figure 4 This is a framework diagram of the present invention; Figure 5 This is a flowchart of the present invention.

[0015] The components include: 1. Data acquisition housing; 2. Shoulder strap; 3. Connector; 4. Restraint strap; 5. Connecting shell; 6. Positioning post; 7. Spring; 8. Push plate; 9. Slot; 10. Limiting slot; 11. Positioning hole; 12. Limiting block; 13. ECG lead electrode pads; 14. Piezoelectric thin film sensor; 15. Temperature sensor; 16. Data acquisition unit; 17. Magnetometer; 18. Accelerometer; 19. Integrated gyroscope; 20. Timing sensor; 21. Wireless transmission module; 22. Image processing algorithm module; 23. Camera; 24. Display device. Detailed implementation methods; The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example: like Figure 1-5As shown, this embodiment of the invention provides a rehabilitation prediction device based on a multi-source heterogeneous large model, including a data acquisition housing 1 and a camera 23. Two binding straps 4 are fixedly connected to both ends of the data acquisition housing 1. Connectors 3 are fixedly connected to both ends of the binding straps 4. A limiting groove 10 is formed inside the connector 3, and a limiting block 12 is correspondingly installed inside the limiting groove 10. A shoulder strap 2 is fixedly connected to the top of the limiting block 12. The binding straps 4 and shoulder straps 2 work together to allow the wearer to wear the device closely. Simultaneously, the camera 23 is mounted on a wearable watch. The two wearable devices can complement each other and work together to perform multi-faceted rehabilitation monitoring of the wearer. This device is equipped with dedicated software that can read patient medical records, examination data, and diagnostic data. The software has a built-in large language model and can generate rehabilitation prescription reports based on the data collected by the device.

[0018] The outer ends of the connector 3 are provided with slots 9. A spring 7 is fixedly connected to one end of the inner side of the slot 9. A push plate 8 is fixedly connected to the other end of the spring 7. A positioning post 6 is fixedly connected to the other end of the push plate 8. The positioning post 6 is movably connected inside the limiting groove 10. A positioning hole 11 is provided through the inside of the limiting block 12. The positioning post 6 and the positioning hole 11 correspond to each other. The limiting blocks 12 at both ends of the shoulder strap 2 are aligned with the limiting groove 10. Then, the push plate 8 is pulled outward. Under the action of the spring 7, the push plate 8 can pull the positioning post 6 inward, so that the positioning post 6 is retracted into the inside of the slot 9. At this time, the user inserts the limiting block 12 directly into the inside of the limiting groove 10 and releases the push plates 8 at both ends. Under the action of the spring 7, the push plate 8 and the positioning post 6 can be driven to quickly reset and engage with the positioning hole 11 inside the limiting block 12 to complete the installation of the shoulder strap 2. This allows the entire rehabilitation prediction device to fit on the human body for real-time monitoring. A piezoelectric thin film sensor 14 is provided at the inner end of the data acquisition housing 1, and temperature sensors 15 are provided on both sides away from the piezoelectric thin film sensor 14. An electrocardiogram lead electrode 13 is provided on the shoulder strap 3, and a connecting shell 5 is installed at the outer end of the data acquisition housing 1. The piezoelectric thin film sensor 14 is used to collect data on the user's lung breathing status, and the electrocardiogram lead electrode 13 can collect data on the user's heart rate changes. The data acquisition housing 1 houses a data acquisition unit 16, and the connecting housing 5 houses a magnetometer 17. An accelerometer 18 is installed on one side of the magnetometer 17, and an integrated gyroscope 19 is installed on one side of the accelerometer 18. The magnetometer 17, accelerometer 18, and integrated gyroscope 19 are all electrically connected to the data acquisition unit 16. One side of the data acquisition unit 16 is electrically connected to a timing sensor 20, and the other side of the timing sensor 20 is electrically connected to a wireless transmission module 21. A camera 23 is mounted on the wearable watch casing, and an image processing algorithm module 22 is electrically connected inside the camera 23. The other end of the image processing algorithm module 22 is also electrically connected to the wireless transmission module 21. The other end of the wireless transmission module 21 is connected to a display device 24 via WIFI and wireless remote technology. The wireless transmission module 21 is used to transmit data collected by various sensors to the guardian's mobile phone, cloud, PC, and other display devices. Simultaneously, it can transmit monitoring data and prediction comparison results to the accompanying dedicated software. The guardian can understand the user's basic respiratory condition and physiological symptoms such as lung sounds based on the data collected by the piezoelectric film sensor 14. The guardian can also understand the user's body temperature changes and duration based on the data collected by the temperature sensor 21. For an extended period, the integrated gyroscope 18 and magnetometer 17 inside the shell 5 are connected to the device. The built-in computing library provides sensor posture data, records sensor data of the patient model on different motion trajectories, and changes in body posture during specific tasks. The study uses the accelerometer 18 and integrated gyroscope 19 to monitor the patient's motion status in real time, including motion trajectory, gait, posture and activity intensity, and monitors physiological parameters such as heart rate, respiratory rate and body temperature. The visual tracking technology monitors the patient's motion through the camera 23 and image processing algorithm module 22. The development of high-resolution cameras can capture subtle motion details to ensure the accuracy of tracking. The structured light camera obtains three-dimensional spatial information to help analyze the patient's posture and movements. The image processing algorithm module 22 (such as edge detection and feature extraction) is used to identify and analyze the patient's movements. The data collected above can facilitate the doctor's diagnosis during online consultations. The wireless transmission module 21 can be one or more of Bluetooth, 5G, Wi-Fi, or RF modules. It can also be a communication module in other open frequency bands. Wireless data transmission can be achieved through the wireless transmission module 21. The data acquisition device 16 detects multiple indicators of the wearer and transmits the data to the time sequence sensor 20 for multi-source preprocessing to generate multi-source heterogeneous time series data. Then, the multi-source heterogeneous time series data is input into the multi-dimensional multimodal collaborative prediction model established by deterministic learning method and dynamic environment through the wireless transmission module 21 to generate multiple prediction results. The multiple prediction results are compared with the user's historical data. The monitoring data and prediction comparison results are then transmitted to the accompanying dedicated software. After reading the patient's medical records, examination data, and diagnostic data, the software performs comprehensive analysis of all data through the built-in large language model to generate a personalized rehabilitation prescription report. The monitoring comparison data and rehabilitation prescription report can be directly sent to a mobile phone or computer. The monitoring data and rehabilitation prescription report can be used to adjust the subsequent rehabilitation direction in a timely manner, realizing the rehabilitation prediction effect of the entire multi-source heterogeneous large model.

[0019] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A rehabilitation prediction device based on a multi-source heterogeneous large model, comprising a data acquisition housing (1) and a camera (23), characterized in that: The data acquisition housing (1) is fixedly connected to two ends with a restraint strap (4), and the upper ends of the restraint strap (4) are fixedly connected with a connector (3). A limit groove (10) is opened inside the connector (3), and a limit block (12) is installed inside the limit groove (10). A shoulder strap (2) is fixedly connected to the top of the limit block (12). The connector (3) has slots (9) at both ends on the outer side. A spring (7) is fixedly connected to one end of the inner side of the slot (9). A push plate (8) is fixedly connected to the other end of the spring (7). A positioning post (6) is fixedly connected to the other end of the push plate (8). The positioning post (6) is movably connected inside the limiting groove (10). A positioning hole (11) is opened through the inside of the limiting block (12). The positioning post (6) and the positioning hole (11) correspond to each other. The data acquisition housing (1) is provided with a piezoelectric thin film sensor (14) at its inner end, and temperature sensors (15) are provided on both sides away from the piezoelectric thin film sensor (14). The shoulder strap (3) is provided with an electrocardiogram lead electrode (13). The data acquisition housing (1) is provided with a connecting shell (5) at its outer end.

2. The rehabilitation prediction device based on a multi-source heterogeneous large model according to claim 1, characterized in that: The data acquisition housing (1) is equipped with a data acquisition device (16), and the connecting housing (5) is equipped with a magnetometer (17).

3. The rehabilitation prediction device based on a multi-source heterogeneous large model according to claim 2, characterized in that: An accelerometer (18) is installed on one side of the magnetometer (17), and an integrated gyroscope (19) is installed on one side of the accelerometer (18). The magnetometer (17), accelerometer (18) and integrated gyroscope (19) are all electrically connected to the data acquisition unit (16).

4. The rehabilitation prediction device based on a multi-source heterogeneous large model according to claim 2, characterized in that: One side of the data acquisition unit (16) is electrically connected to a timing sensor (20), and one side of the timing sensor (20) is electrically connected to a wireless transmission module (21).

5. The rehabilitation prediction device based on a multi-source heterogeneous large model according to claim 1, characterized in that: The camera (23) is mounted on the wearable watch case, and the camera (23) is electrically connected to an image processing algorithm module (22). The other end of the image processing algorithm module (22) is also electrically connected to the wireless transmission module (21).

6. The rehabilitation prediction device based on a multi-source heterogeneous large model according to claim 4, characterized in that: The other end of the wireless transmission module (21) is connected to the display device (24) using WIFI and wireless remote technology. The rehabilitation prescription report generated by the dedicated software can be synchronized to the display device (24) for display.

Images