Data processing method, device and system based on finger motion feature analysis

Abstract

The invention relates to a data processing method, device and system based on finger motion feature analysis. The method comprises the following steps: receiving acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of a specific finger; combining pre-input clinical data of the user with the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value, and inputting a pre-trained cervical spondylotic myelopathy identification model to obtain a screening result; and inputting the screening result into a pre-trained cervical spondylotic myelopathy diagnosis model to obtain a diagnosis result, and the method has the advantages that multiple kinematics characteristics are combined, and diagnosis is accurately and rapidly assisted.

Description

Technical Field

[0001] The present invention relates to the field of medical technology applications, and in particular to a data processing method, device and system based on finger motion feature analysis. Background Art

[0002] In the prior art, auxiliary diagnosis of cervical spondylotic myelopathy is usually completed through a quick finger grip test. Since the blood supply to the cervical thoracic spinal cord is weak, only the upper thoracic spinal cord has a communicating branch to supply blood. Therefore, regardless of the compression segment, cervical spondylotic myelopathy causes the anterior horns of the spinal cord in the C7 and C8 segments to be most susceptible to ischemic damage, resulting in finger grip weakness and stiffness. After the spinal cord is compressed for a long time, it will cause atrophy of the intrinsic muscles of the hand, which is called myelopathy hand. A combination of factors causes the frequency of finger grip to decrease, so clinical diagnosis of cervical spondylotic myelopathy is often performed by the number of quick grips of the test subject within 10 seconds. However, in addition to the grip frequency, many important kinematic features are lost. Some testers also use video capture to record the finger movement characteristics of the subject during the 10-second grip test, but this method loses the acceleration characteristics of the finger movement and has high requirements for equipment and venue, and cannot achieve the purpose of quick testing.

[0003] Existing technology relies on cervical spine magnetic resonance imaging to diagnose cervical spondylotic myelopathy, which greatly increases medical costs and wastes social resources. Furthermore, existing technology can easily misdiagnose amyotrophic lateral sclerosis (the most common motor neuron disease) as cervical spondylotic myelopathy when detecting cervical spondylotic myelopathy. ALS manifests as split hands in its early stages, and affects the hand muscles more radially than ulnarly, often manifesting as significant atrophy of the thenar eminence but less atrophy of the hypothenar and ulnar flexor muscles. Due to the impact on upper motor neurons, there is slow and stiff finger movement, which can be easily confused with cervical spondylotic myelopathy. Therefore, ALS can be misdiagnosed as cervical spondylotic myelopathy.

[0004] Currently, in the initial screening technology for cervical spondylotic myelopathy, the testing method itself ignores more important kinematic characteristics, resulting in low efficiency and poor accuracy in diagnosing cervical spondylotic myelopathy by relying on the traditional 10-second grip test. No effective solution has been proposed yet. Summary of the Invention

[0005] The purpose of the present invention is to address the deficiencies in the existing technology and provide a data processing method, device and system based on finger motion feature analysis to solve the technical problem that in the relevant initial screening technology for cervical spondylotic myelopathy, the test method itself ignores more important kinematic characteristics, resulting in low efficiency and poor accuracy in diagnosing cervical spondylotic myelopathy by relying on the traditional 10-second grip test.

[0006] To achieve the above object, the technical solution adopted by the present invention is:

[0007] The present invention provides a data processing system based on finger motion feature analysis, comprising: a wearable device and a host computer, wherein the wearable device is worn on a specific finger of a user, and the wearable device comprises: an accelerometer, a gyroscope, an angle sensor and a communication module, wherein the accelerometer is used to collect the instantaneous linear acceleration of the specific finger on three vertical axes; the gyroscope is used to collect the instantaneous angular velocity of the specific finger in three vertical planes; the angle sensor is used to collect the instantaneous rotation angle value of the specific finger in three vertical planes; the communication module is respectively connected to the accelerometer, the gyroscope and the communication module. The instrument is connected to the angle sensor and is used to send the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​to the host computer; the host computer is connected to the wearable device and is used to receive the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​of a specific finger; the pre-input user's clinical data is combined with the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values, and input into a pre-trained spinal cord type cervical spondylotic myelopathy identification model to obtain a screening result, and the screening result is input into a pre-trained spinal cord type cervical spondylotic myelopathy diagnosis model to obtain a diagnosis result.

[0008] Optionally, the wearable device also includes: a circuit board, a hardware synchronization circuit and a cache device, wherein the circuit board adopts multi-layer wiring, and the positions of the accelerometer, gyroscope, angle sensor and communication module are set in a preset layout; the hardware synchronization circuit is a master-slave architecture, the accelerometer is set as the master sensor, the gyroscope and angle sensor are set as slave sensors, the master sensor controls the sampling frequency and timing of the slave sensors, and is used to control the accelerometer, gyroscope and angle sensor to collect data simultaneously; the cache device is located in the wearable device, and adopts a ring buffer structure, which is used to write data into the ring buffer in sequence. When the ring buffer is stored to the maximum value, the data written at the Nth time overwrites the data written at the 1st time; during the data transmission process, if the network fails or is interrupted, the collected data is first stored in the ring buffer; if the network is restored, the data in the ring buffer is transmitted to the host computer through the communication module in sequence.

[0009] Optionally, the communication module is also used to send the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​to the host computer through wireless low-power communication technology according to a preset transmission strategy; wherein the preset transmission strategy includes: compressing the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​through a lossy compression algorithm, and adjusting the transmission rate according to the network status.

[0010] Further, optionally, a cervical spondylotic myelopathy identification model and a cervical spondylotic myelopathy diagnosis model are configured in the host computer, wherein, in the model training stage of the cervical spondylotic myelopathy identification model and the cervical spondylotic myelopathy diagnosis model, the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data are input into the cervical spondylotic myelopathy identification model for training, and when the sensitivity reaches a first threshold, it is determined that the cervical spondylotic myelopathy identification model has converged; after excluding the similar disease sample data, the target cervical spondylotic myelopathy sample data and normal sample data are input into the cervical spondylotic myelopathy diagnosis model for training, and when the specificity reaches a second threshold, it is determined that the cervical spondylotic myelopathy diagnosis model has converged; wherein the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data all include: acquisition time, clinical data, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of a specific finger;

[0011] Wherein, in the model training stage, the acquisition time, clinical data, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data are normalized to obtain the normalized target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data; the normalized target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data are binary classified to obtain first classification data and second classification data, and the first classification data and the second classification data are input into the cervical spondylotic myelopathy identification model for training, and when the sensitivity reaches the first threshold, it is determined that the cervical spondylotic myelopathy identification model has converged; wherein, the first classification data is the normalized similar disease sample data, and the second classification data is the normalized target cervical spondylotic myelopathy sample data and normal sample data; the normalized target cervical spondylotic myelopathy sample data and the normalized normal sample data are input into the cervical spondylotic myelopathy diagnosis model for training, and when the specificity reaches the second threshold, it is determined that the cervical spondylotic myelopathy diagnosis model has converged;

[0012] During the model application phase of the cervical spondylotic myelopathy identification model and the cervical spondylotic myelopathy diagnosis model, the user's clinical data is combined with the acquisition time, instantaneous linear acceleration, instantaneous angular velocity, and instantaneous rotation angle values, and input into the cervical spondylotic myelopathy identification model to screen similar diseases. The data after similar diseases are eliminated is obtained and determined as the screening result. The screening result is input into the cervical spondylotic myelopathy diagnosis model to obtain the diagnosis result.

[0013] Among them, in the model application stage, the acquisition time, clinical data, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of the specific finger are normalized to obtain a normalized data group, and the normalized data group is input into the spinal cervical spondylotic myelopathy identification model to screen similar diseases, and the data after eliminating similar diseases is obtained, and the data is determined as the screening result; the screening result is input into the spinal cervical spondylotic myelopathy diagnosis model to obtain the diagnosis result.

[0014] Optionally, the host computer is further used to receive the acquisition time, the instantaneous linear acceleration and instantaneous angular velocity of a specific finger, and calculate the number of grips within the specific time based on the instantaneous angular velocity, combine the clinical data with one or a combination of the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity, and the number of grips within the specific time, input the data into a pre-trained cervical spondylotic myelopathy identification model to obtain a screening result, and input the screening result into a pre-trained cervical spondylotic myelopathy diagnosis model to obtain a diagnosis result;

[0015] Calculating the number of grips within a specific time period based on the instantaneous angular velocity includes: obtaining an average grip period by dividing the test time by the number of trough values ​​in the time series of the instantaneous angular velocity; and obtaining the number of grips within the specific time period by dividing the specific time by the average grip period.

[0016] Optionally, the host computer is further used to obtain target cervical spondylotic myelopathy sample data and normal sample data, and calculate the acceleration average, TopN acceleration average, maximum acceleration value, angular velocity average, TopN angular velocity average, maximum angular velocity value and the number of grips within a specific time based on the target cervical spondylotic myelopathy sample data and normal sample data respectively; and generate an ROC curve based on the acceleration average, TopN acceleration average, maximum acceleration value, angular velocity average, TopN angular velocity average, maximum angular velocity value and the number of grips within a specific time based on the target cervical spondylotic myelopathy sample data and normal sample data respectively. , calculate the area under the ROC curve and the cutoff value, and generate a diagnostic result based on the cutoff value; or, the host computer is also used to determine the user's fatigue by calculating the ratio of the average acceleration of adjacent time periods to the average acceleration, and the ratio of the angular velocity to the angular velocity of adjacent time periods; the host computer is also used to judge the direction of force according to the positive and negative values ​​of the angular velocity, determine the finger flexion acceleration and / or finger extension acceleration, and determine the extreme values, fatigue and strength reserves of the finger flexion acceleration and finger extension acceleration, and visualize the extreme values, fatigue and strength reserves of the finger flexion acceleration and finger extension acceleration so that they can be displayed to the user, and perform classification and judgment through deep learning.

[0017] The present invention provides a wearable device, which is applied to a data processing system based on finger motion feature analysis. The wearable device is worn on a specific finger of a user and includes: an accelerometer, a gyroscope, an angle sensor and a communication module, wherein the accelerometer is used to collect the instantaneous linear acceleration of the specific finger on three vertical axes; the gyroscope is used to collect the instantaneous angular velocity of the specific finger in three vertical planes; the angle sensor is used to collect the instantaneous rotation angle value of the specific finger in three vertical planes; the communication module is connected to the accelerometer, the gyroscope and the angle sensor respectively, and is used to send the instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value to a host computer; wherein the wearable device also includes: a circuit board and a hardware synchronization circuit, which In the circuit board, multi-layer wiring is used, and the positions of the accelerometer, gyroscope, angle sensor and communication module are set in a preset layout; the hardware synchronization circuit is a master-slave architecture, the accelerometer is set as the master sensor, the gyroscope and angle sensor are set as slave sensors, and the master sensor controls the sampling frequency and timing of the slave sensors, which is used to control the accelerometer, gyroscope and angle sensor to collect data simultaneously; the communication module is also used to send the instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​to the host computer through wireless low-power communication technology according to a preset transmission strategy; wherein the preset transmission strategy includes: compressing the instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​through a lossy compression algorithm, and adjusting the transmission rate according to the network status.

[0018] The present invention provides a data processing method based on finger motion feature analysis, which is applied to a data processing system based on finger motion feature analysis, comprising: receiving an acquisition time, an instantaneous linear acceleration, an instantaneous angular velocity, and an instantaneous rotation angle value of a specific finger; combining pre-input clinical data of a user with the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity, and the instantaneous rotation angle value, and inputting the data into a pre-trained cervical spondylotic myelopathy identification model to obtain a screening result; and inputting the screening result into a pre-trained cervical spondylotic myelopathy diagnosis model to obtain a diagnosis result.

[0019] Optionally, the method further includes: configuring a cervical spondylotic myelopathy identification model and a cervical spondylotic myelopathy diagnosis model in a host computer; in the model training stage of the cervical spondylotic myelopathy identification model and the cervical spondylotic myelopathy diagnosis model, inputting target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data into the cervical spondylotic myelopathy identification model for training, and when the sensitivity reaches a first threshold, determining that the cervical spondylotic myelopathy identification model has converged; after excluding similar disease sample data, inputting target cervical spondylotic myelopathy sample data and normal sample data into the cervical spondylotic myelopathy diagnosis model for training, and when the specificity reaches a second threshold, determining that the cervical spondylotic myelopathy diagnosis model has converged; wherein, the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data all include: acquisition time, clinical data, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of a specific finger; wherein, in the model training stage, the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data are input into the cervical spondylotic myelopathy diagnosis model for training, and when the specificity reaches a second threshold, determining that the cervical spondylotic myelopathy diagnosis model has converged. The acquisition time, clinical data, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of the specific finger in the sample data are normalized to obtain normalized target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data; the normalized target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data are binary classified to obtain first classification data and second classification data, the first classification data and the second classification data are input into the cervical spondylotic myelopathy identification model for training, and when the sensitivity reaches a first threshold, it is determined that the cervical spondylotic myelopathy identification model has converged; wherein the first classification data is the normalized similar disease sample data, and the second classification data is the normalized target cervical spondylotic myelopathy sample data and normal sample data; the normalized target cervical spondylotic myelopathy sample data and the normalized normal sample data are input into the cervical spondylotic myelopathy diagnosis model for training, and when the specificity reaches a second threshold, it is determined that the cervical spondylotic myelopathy diagnosis model has converged;

[0020] During the model application stage of the spinal cervical spondylotic myelopathy identification model and the spinal cervical spondylotic myelopathy diagnosis model, the user's clinical data is combined with the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values, and input into the spinal cervical spondylotic myelopathy identification model for similar disease screening to obtain data after excluding similar diseases, and the data is determined as the screening result; the screening result is input into the spinal cervical spondylotic myelopathy diagnosis model to obtain the diagnosis result; wherein, during the model application stage, the acquisition time, clinical data, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​of a specific finger are normalized to obtain a normalized data group, and the normalized data group is input into the spinal cervical spondylotic myelopathy identification model for similar disease screening to obtain data after excluding similar diseases, and the data is determined as the screening result; the screening result is input into the spinal cervical spondylotic myelopathy diagnosis model to obtain the diagnosis result.

[0021] Optionally, the method further includes: acquiring target cervical spondylotic myelopathy sample data and normal sample data; calculating the acceleration average, TopN acceleration average, maximum acceleration value, angular velocity average, TopN angular velocity average, maximum angular velocity value, and the number of grips within a specific time based on the target cervical spondylotic myelopathy sample data and the normal sample data, respectively; generating an ROC curve based on the acceleration average, TopN acceleration average, maximum acceleration value, angular velocity average, TopN angular velocity average, maximum angular velocity value, and the number of grips within a specific time based on the target cervical spondylotic myelopathy sample data and the normal sample data; calculating the area under the ROC curve and the cutoff value; and generating a diagnosis result based on the cutoff value;

[0022] or,

[0023] The user's fatigue level is determined by calculating the ratio of the average acceleration to the average acceleration of adjacent time periods, and the ratio of the angular velocity to the angular velocity of adjacent time periods; the direction of force is judged according to the positive or negative value of the angular velocity, the finger flexion acceleration and / or finger extension acceleration is determined, and the extreme values, fatigue and strength reserves of the finger flexion acceleration and finger extension acceleration are determined, and the extreme values, fatigue and strength reserves of the finger flexion acceleration and finger extension acceleration are visualized so that they can be presented to the user and classified and judged through deep learning.

[0024] The present invention adopts the above technical solution, by receiving the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of a specific finger; combining the pre-input user's clinical data with the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value, and inputting a pre-trained spinal cord type cervical spondylotic myelopathy identification model to obtain a screening result; inputting the screening result into a pre-trained spinal cord type cervical spondylotic myelopathy diagnosis model to obtain a diagnosis result. Compared with the existing technology, the present invention has the following technical effects: combining multiple kinematic characteristics to accurately and quickly assist in diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is a schematic diagram of a data processing system based on finger motion feature analysis according to a first embodiment of the present invention;

[0026] Figure 2 is a schematic diagram of model training in a data processing system based on finger motion feature analysis according to Embodiment 1 of the present invention;

[0027] Figure 3 4 is a flow chart of a data processing method based on finger motion feature analysis according to embodiment 3 of the present invention. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is described and illustrated below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely used to explain this application and are not intended to limit this application. Based on the embodiments provided in this application, all other embodiments obtained by those of ordinary skill in the art without making any creative efforts are within the scope of protection of this application.

[0029] Obviously, the drawings described below are merely examples or embodiments of the present application. Those skilled in the art can, without inventive effort, apply the present application to other similar scenarios based on these drawings. Furthermore, it is also understood that, although the effort involved in such a development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, changes in design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as an insufficiency of the content disclosed in this application.

[0030] References to "embodiments" in this application mean that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it refer to independent or alternative embodiments that are mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described in this application may be combined with other embodiments unless there is a conflict.

[0031] Unless otherwise defined, the technical or scientific terms used in this application should have the ordinary meaning understood by a person of ordinary skill in the technical field to which this application belongs. The words "a", "an", "a", "the" and the like used in this application do not indicate a limit on quantity and may indicate the singular or plural. The terms "include", "comprise", "have" and any variations thereof used in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or units (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units that are inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The words "multiple" / "several" used in this application refer to two or more. "And / or" describes the association relationship between associated objects, indicating that three relationships can exist. For example, "A and / or B" can mean: A exists alone, A and B exist at the same time, or B exists alone. The character " / " generally indicates that the objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific order for the objects.

[0032] Example 1

[0033] An exemplary embodiment of the present invention is as follows Figure 1 As shown, Figure 1 2 is a schematic diagram of a data processing system based on finger motion feature analysis according to a first embodiment of the present invention. The data processing system based on finger motion feature analysis provided by the embodiment of the present application includes:

[0034] Wearable device 12 and host computer 14, wherein,

[0035] The wearable device 12 is worn on a specific finger of the user. The wearable device 12 includes: an accelerometer 121, a gyroscope 122, an angle sensor 123 and a communication module 124, wherein the accelerometer 121 is used to collect the instantaneous linear acceleration of the specific finger on three vertical axes; the gyroscope 122 is used to collect the instantaneous angular velocity of the specific finger in three vertical planes; the angle sensor 123 is used to collect the instantaneous rotation angle value of the specific finger in three vertical planes; the communication module 124 is respectively connected to the accelerometer 121, the gyroscope 122 and the angle sensor 123. The sensor 123 is connected to send the instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​to the host computer 14; the host computer 14 is connected to the wearable device 12, and is used to receive the acquisition time, the instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​of a specific finger; the pre-input user's clinical data is combined with the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values, and input into a pre-trained spinal cord type cervical spondylotic myelopathy identification model to obtain a screening result, and the screening result is input into a pre-trained spinal cord type cervical spondylotic myelopathy diagnosis model to obtain a diagnosis result.

[0036] Specifically, the wearable device 12 in the embodiment of the present application is a ring-shaped wearable device worn on the user's finger, which contains a variety of mechanical parameter sensors of different modes and directions for collecting acceleration, angular velocity and angle. By applying a customized deep learning artificial intelligence model to analyze the above acceleration, angular velocity and angle, an assessment and diagnosis result of whether the target patient has spinal cervical spondylotic myelopathy is automatically obtained. In particular, in the embodiment of the present application, before the accelerometer 121, the gyroscope 122, and the angle sensor 123 collect the instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​of a specific finger, the accelerometer 121, the gyroscope 122, and the angle sensor 123 are calibrated and / or reset to zero.

[0037] By sending the data to the host computer 14 (for example, a computer or mobile phone) via wireless communication (Wi-Fi or Bluetooth, etc.), the algorithm deployed on the host computer 14 processes and analyzes the received data, and presents the analysis results on the host computer and feeds them back to the doctor.

[0038] In the embodiment of the present application, the specific finger may be: the user's little finger.

[0039] In the embodiment of the present application, clinical data may include: user's age, gender, BMI and finger length;

[0040] Among them, the Body Mass Index (BMI) is an internationally used indicator to measure the degree of fatness and health of the human body. The calculation formula is:

[0041] BMI = weight (kg) / height (m2)2 ).

[0042] In the embodiment of the present application, the accelerometer 121 may be, but is not limited to, the Bosch BMI160 accelerometer. The BMI160 accelerometer is characterized by its small size, measuring only 2.5mm x 3.0mm x 0.8mm, making it ideal for integration into a device worn on the pinky finger without significantly hindering pinky movement. Furthermore, the BMI160 accelerometer features low power consumption, with current consumption as low as 35μA within a measurement range of ±16g, effectively extending the device's battery life. The BMI160 accelerometer also boasts high measurement accuracy, capable of precisely measuring changes in the pinky finger's acceleration in all directions, meeting the need for accurate monitoring of the pinky's motion status.

[0043] In the present embodiment, gyroscope 122 may be, but is not limited to, an L3G4200D gyroscope. The L3G4200D gyroscope's measurement range is flexibly configurable, reaching up to ±2000° / s, enabling accurate capture of angular velocity changes during rapid pinky finger movements. Furthermore, the L3G4200D gyroscope offers low noise levels and high stability, ensuring accurate and reliable angular velocity data even in complex motion situations, providing strong support for analyzing pinky finger motion postures.

[0044] It should be noted that the embodiments of the present application are described using only the L3G4200D gyroscope and its parameters as examples, with the data processing system based on finger motion feature analysis provided in the embodiments of the present application being the basis, and no specific limitation is made.

[0045] The angle sensor 123 in the present embodiment can be, but is not limited to, the micro-switch MMA7660FC angle sensor. The micro-switch MMA7660FC angle sensor can directly measure the tilt angle of an object. Its small size and light weight make it easy to integrate into a device worn on the pinky finger. With a measurement accuracy of ±0.5°, the micro-switch MMA7660FC angle sensor can precisely measure the angle changes of the pinky finger in different postures, providing accurate data for determining whether the pinky finger's angle of motion is abnormal.

[0046] The process of the wearable device 12 collecting the instantaneous linear acceleration, instantaneous angular velocity, and instantaneous rotation angle value of a specific finger includes:

[0047] Collect the instantaneous linear acceleration of the little finger on three vertical axes, the instantaneous angular velocity in three vertical planes, the instantaneous rotation angle value in three vertical planes, and the acquisition time. The square root of the sum of the accelerations on the three vertical axes is used as the spatial instantaneous acceleration. The calculation is as follows:

[0048]

[0049] Where a is the instantaneous linear acceleration on the three perpendicular axes, x is the acceleration on the x-axis, y is the acceleration on the y-axis, and z is the acceleration on the z-axis;

[0050] In the embodiment of the present application, the acquisition time at least includes: recording the moment corresponding to the peak angle, taking the time between adjacent peaks as the unit period of the finger flexion and extension movement, taking the average period as a feature and inputting the instantaneous linear acceleration on the three vertical axes, the instantaneous angular velocity in the three vertical planes, the instantaneous rotation angle value in the three vertical planes and the clinical data into the host computer 14.

[0051] It should be noted that the movement characteristics of the little finger (i.e., the instantaneous linear acceleration of the little finger on three vertical axes, the instantaneous angular velocity in three vertical planes, and the instantaneous rotation angle value in three vertical planes in the embodiment of the present application) are used as diagnostic characteristics of spinal cervical spondylotic myelopathy. The flexion and extension activities of the little finger represent the function of the most distal anterior horn of the cervical spinal cord. This area is most sensitive to the blood supply of the spinal cord. When the cervical spine is under severe compression, the blood supply to the cervical thoracic spinal cord is usually weak, resulting in ischemia of the anterior horn of the cervical spinal cord, which causes the corresponding little finger to move clumsily, weakly, and with a reduced range. Therefore, the data processing system based on finger movement characteristic analysis provided in the embodiment of the present application uses the movement characteristics of the little finger to diagnose spinal cervical spondylotic myelopathy with higher accuracy than the existing technology.

[0052] In addition, in terms of signal processing and noise reduction, the collected sensor signals are susceptible to various noise interferences, such as the body's own physiological electrical signals, environmental electromagnetic interference, etc. It is necessary to adopt effective filtering and signal enhancement technologies to remove noise and improve signal quality. The embodiment of the present application uses wavelet transform filtering to remove high-frequency noise, and uses an adaptive filtering algorithm to reduce interference caused by individual differences and environmental changes, ensuring that the extracted kinematic parameters accurately reflect the actual movement state of the little finger. In addition, it is also necessary to deal with the baseline drift problem of the signal to ensure the stability and reliability of the data.

[0053] Optionally, the wearable device 12 also includes: a circuit board, a hardware synchronization circuit and a cache device, wherein the circuit board adopts multi-layer wiring, and the positions of the accelerometer 121, the gyroscope 122, the angle sensor 123 and the communication module are set in a preset layout; the hardware synchronization circuit is a master-slave architecture, the accelerometer 121 is set as the master sensor, the gyroscope 122 and the angle sensor 123 are slave sensors, the master sensor controls the sampling frequency and timing of the slave sensors, and is used to control the accelerometer 121, the gyroscope 122 and the angle sensor 123 to collect data simultaneously; the cache device is located in the wearable device, and adopts a ring buffer structure, which is used to write data into the ring buffer in sequence. When the ring buffer is stored to the maximum value, the data written at the Nth time overwrites the data written at the 1st time; during the data transmission process, if the network fails or is interrupted, the collected data is first stored in the ring buffer; if the network is restored, the data in the ring buffer is transmitted to the host computer 14 in sequence through the communication module.

[0054] Specifically, in terms of hardware integration, the embodiment of the present application tightly integrates the accelerometer 121, the gyroscope 122, and the angle sensor 123 into a tiny module by designing a customized printed circuit board (PCB), that is, the circuit board in the embodiment of the present application. Through reasonable layout, the distance between the sensors is reduced, the signal transmission delay is reduced, and it is ensured that the accelerometer 121, the gyroscope 122, and the angle sensor 123 can synchronously collect the little finger motion data. The multi-layer wiring technology of the PCB is utilized to reduce the electromagnetic interference between the sensors and improve the stability and accuracy of data acquisition. At the same time, a power interface and a data transmission interface are reserved on the PCB to facilitate connection with external devices.

[0055] In the embodiment of the present application, a master-slave architecture is used for data acquisition during synchronous data acquisition. The accelerometer 121 is used as the master sensor to control the sampling frequency and timing of the gyroscope 122 and the angle sensor 123. Through a hardware synchronization circuit, the accelerometer 121, gyroscope 122, and angle sensor 123 collect data at the same time, ensuring that the acquired kinematic parameters accurately reflect the motion state of the little finger at the same moment. At the software level, corresponding drivers are written to precisely control and coordinate the data acquisition of each sensor to ensure the synchronization and integrity of the data.

[0056] In the cache device in the embodiment of the present application, since the data transmission from the wearable device 12 to the host computer 14 may be affected by factors such as network conditions, in order to prevent data loss, a cache area is set inside the wearable device 12 (that is, a ring buffer in the cache device in the embodiment of the present application). A ring buffer structure is used to write data into the buffer in sequence. When the buffer is full, the new data will overwrite the earliest data (that is, the data written at the Nth moment in the embodiment of the present application overwrites the data written at the 1st moment). During the data transmission process, if the network fails or is interrupted, the collected data is first stored in the ring buffer. After the network is restored, the data in the buffer is transmitted to the host computer in sequence to ensure the integrity of the data.

[0057] Among them, the use of compact binary format to store data can not only save storage space but also speed up data reading, so as to efficiently store large amounts of kinematic data and facilitate subsequent deep learning algorithm calls and analysis.

[0058] It should be noted that the circuit board, hardware synchronization circuit and cache device in the embodiment of the present application are Figure 1 Not shown in the figure.

[0059] Optionally, the communication module 124 is also used to send the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​to the host computer 14 through wireless low-power communication technology according to a preset transmission strategy; wherein the preset transmission strategy includes: compressing the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​through a lossy compression algorithm, and adjusting the transmission rate according to the network status.

[0060] Specifically, the wireless low-power communication technology in the embodiment of the present application can select but is not limited to Bluetooth Low Energy (BLE) technology for data transmission. BLE technology is designed for short-range, low-power device communication, operates in the 2.4GHz frequency band, and can meet the needs of the wearable device 12 being worn on the little finger. The transmission distance can reach about 10 meters in an open environment, which can meet daily use scenarios. For example, when the user is indoors, he connects to the host computer through BLE technology. BLE technology has low power consumption, meets the endurance requirements of the wearable device 12, and can effectively reduce the energy consumption of the device when transmitting data. In addition, BLE technology has a certain anti-interference ability, and uses adaptive frequency hopping technology to automatically switch between 79 channels, reducing interference from other devices in the same frequency band and ensuring the stability of data transmission.

[0061] In the transmission optimization strategy (i.e., the preset transmission strategy in the embodiment of the present application), data compression technology is used to improve transmission efficiency. The collected acquisition time, acceleration, angular velocity and angle (i.e., the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value in the embodiment of the present application) are compressed using a lossy compression algorithm such as Discrete Cosine Transform (DCT). DCT can effectively remove redundant information in the data, greatly reduce the amount of data and increase the transmission speed while ensuring the key features of the data. During the transmission process, the transmission rate is dynamically adjusted according to the network conditions. If the signal strength is good and the interference is less, the transmission rate is increased to transmit data quickly; if the signal is weak or the interference is large, the transmission rate is reduced to ensure the stability of data transmission and avoid data loss (i.e., the transmission rate is adjusted according to the network status in the embodiment of the present application).

[0062] Optionally, a cervical spondylotic myelopathy identification model and a cervical spondylotic myelopathy diagnosis model are configured in the host computer 14, wherein, in the model training stage of the cervical spondylotic myelopathy identification model and the cervical spondylotic myelopathy diagnosis model, the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data are input into the cervical spondylotic myelopathy identification model for training, and when the sensitivity reaches a first threshold, it is determined that the cervical spondylotic myelopathy identification model has converged; after excluding the similar disease sample data, the target cervical spondylotic myelopathy sample data and normal sample data are input into the cervical spondylotic myelopathy diagnosis model for training, and when the specificity reaches a second threshold, it is determined that the cervical spondylotic myelopathy diagnosis model has converged. Model convergence; wherein, the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data all include: acquisition time, clinical data, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of a specific finger; in the model application stage of the cervical spondylotic myelopathy identification model and the cervical spondylotic myelopathy diagnosis model, the user's clinical data is combined with the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value, and input into the cervical spondylotic myelopathy identification model for similar disease screening, and the data after excluding similar diseases is obtained, and the data is determined as the screening result; the screening result is input into the cervical spondylotic myelopathy diagnosis model to obtain the diagnosis result.

[0063] Among them, the similar disease sample data in the embodiment of the present application can be cervical spine related identification disease data, that is, disease data that is easily misdiagnosed as spinal cervical spondylotic myelopathy.

[0064] Furthermore, optionally, the host computer 14 is also used to normalize the acquisition time, clinical data, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data in the model training stage to obtain the normalized target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data; perform binary classification on the normalized target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data to obtain first classification data and second classification data, input the first classification data and the second classification data into the cervical spondylotic myelopathy identification model for training, and determine that the cervical spondylotic myelopathy identification model has converged when the sensitivity reaches a first threshold value; wherein the first classification data is the normalized data. The first classification data is the normalized similar disease sample data, and the second classification data is the normalized target cervical spondylotic myelopathy sample data and normal sample data; the normalized target cervical spondylotic myelopathy sample data and the normalized normal sample data are input into the cervical spondylotic myelopathy diagnosis model for training, and when the specificity reaches the second threshold, the cervical spondylotic myelopathy diagnosis model is determined to have converged; in the model application stage, the acquisition time, clinical data, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of a specific finger are normalized to obtain a normalized data group, and the normalized data group is input into the cervical spondylotic myelopathy identification model for similar disease screening, and the data after excluding similar diseases is obtained, and the data is determined as the screening result; the screening result is input into the cervical spondylotic myelopathy diagnosis model to obtain the diagnosis result.

[0065] Specifically, such as Figure 2 As shown, Figure 2 This is a schematic diagram of model training in a data processing system based on finger motion feature analysis according to embodiment 1 of the present invention. The data processing system based on finger motion feature analysis provided in the embodiment of the present application uses a wearable device 12 to collect multiple kinematic features of the user's little finger (i.e., the instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of the specific finger in the embodiment of the present application), and selects the best model through an algorithm. First, other little finger nerve dysfunction diseases that are easily confused with spinal cervical spondylotic myelopathy are excluded, and finally it is determined whether it is spinal cervical spondylotic myelopathy. The data processing system based on finger motion feature analysis provided in the embodiment of the present application solves the problem of being too complicated and not suitable for wearing, and repeated multiple tests in the prior art. In addition, the data processing system based on finger motion feature analysis provided in the embodiment of the present application eliminates differential diagnosis diseases through a cascade model, which can improve the accuracy of diagnosis of spinal cervical spondylotic myelopathy.

[0066] Among them, Figure 2 As shown, the model application in the data processing system based on finger motion feature analysis provided by the embodiment of the present application includes two stages: training and testing;

[0067] Phase 1: Training phase:

[0068] It is worn by diagnosed patients or normal subjects to collect instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​and train the algorithm model together with the clinical data such as age, gender, finger length, etc. of the subjects.

[0069] Specifically, in the embodiment of the present application, the testers are grouped into normal people (i.e., the population corresponding to the normal sample data in the embodiment of the present application), patients with cervical spondylotic myelopathy (i.e., the population corresponding to the target cervical spondylotic myelopathy sample data in the embodiment of the present application), and patients with differential diagnosis diseases (i.e., similar disease sample data in the embodiment of the present application). Data of 500 people each (i.e., the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data in the embodiment of the present application) are taken for deep learning to establish a cervical spondylotic myelopathy identification model and a cervical spondylotic myelopathy diagnosis model. Through model screening, the model with the highest sensitivity is taken as the cervical spondylotic myelopathy identification model, and the model with the highest specificity is taken as the cervical spondylotic myelopathy diagnosis model. The two models are tested on external data to obtain a cervical spondylotic myelopathy identification model and a cervical spondylotic myelopathy diagnosis model.

[0070] Among them, in the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data, the clinical data is added to the instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value in the time series in the form of a constant value, wherein the instantaneous linear acceleration is in three vertical directions, the instantaneous angular velocity is on three vertical planes and the instantaneous rotation angle value is on three vertical planes; after the clinical data is added to the instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value in the time series in the form of a constant value, normalization is performed, and the normalized target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data are normalized. Disease sample data, similar disease sample data and normal sample data, the normalized target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data are binary-classified to obtain first classification data and second classification data, the first classification data and the second classification data are input into the cervical spondylotic myelopathy identification model for training, and when the sensitivity reaches a first threshold, it is determined that the cervical spondylotic myelopathy identification model has converged; wherein the first classification data is the normalized similar disease sample data, and the second classification data is the normalized target cervical spondylotic myelopathy sample data and normal sample data.

[0071] Take a preferred example for illustration: select 500 patients with cervical spondylotic myelopathy (i.e., the target cervical spondylotic myelopathy samples in the embodiment of the present application), select 500 patients with motor neuron disease (i.e., similar disease samples in the embodiment of the present application), and select 500 normal people (i.e., normal samples in the embodiment of the present application).

[0072] First, a cervical spondylotic myelopathy identification model was established. The instantaneous linear acceleration, instantaneous angular velocity, and instantaneous rotation angle of the little finger of patients with motor neuron disease, cervical spondylotic myelopathy, and healthy individuals, as well as clinical data, were used for supervised deep learning.

[0073] Second, through binary classification, the data of patients with motor neuron disease, patients with cervical spondylotic myelopathy, and normal people are divided into first-category data and second-category data, wherein the first-category data includes: data of patients with motor neuron disease, that is, normalized similar disease sample data; the second-category data includes: data of patients with cervical spondylotic myelopathy and normal people, that is, normalized target cervical spondylotic myelopathy sample data and normal sample data;

[0074] Inputting the first classification data and the second classification data into the cervical spondylotic myelopathy identification model for training, finding a model with the highest sensitivity (i.e., the first threshold in the embodiment of the present application) for diagnosing and differentiating the disease, and determining it as the converged cervical spondylotic myelopathy identification model;

[0075] Third, through binary classification, the normalized target cervical spondylotic myelopathy sample data and normal sample data are input into the cervical spondylotic myelopathy diagnosis model for training, and the model with the highest specificity for diagnosing cervical spondylotic myelopathy (i.e., the second threshold in the embodiment of the present application) is found to determine the convergence of the cervical spondylotic myelopathy diagnosis model.

[0076] Phase II: Testing Phase (i.e., the model application phase in the embodiment of the present application):

[0077] In clinical practice, a patient whose disease is unknown or not wears a wearable device 12 and performs hand movement tasks according to the doctor's instructions. During the process, the wearable device 12 records the instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values, and simultaneously sends the instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​wirelessly to the host computer 14. The doctor inputs the patient's age, gender, finger length and other clinical information into the host computer 12, which is equipped with trained algorithm software (i.e., the trained and converged cervical spondylotic myelopathy identification model and cervical spondylotic myelopathy diagnosis model in the embodiment of the present application). The algorithm software analyzes and processes the instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​and clinical data, and finally obtains the evaluation and diagnosis results of cervical spondylotic myelopathy (i.e., the diagnosis results in the embodiment of the present application), and presents them to the doctor for reference through the software.

[0078] That is, the clinical data, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle values ​​of a specific finger are normalized to obtain a normalized data group, the normalized data group is input into the spinal cervical spondylotic myelopathy identification model to screen similar diseases, and the data after eliminating similar diseases is obtained, and the data is determined as the screening result; the screening result is input into the spinal cervical spondylotic myelopathy diagnosis model to obtain the diagnosis result.

[0079] In summary, a preferred example is used for explanation:

[0080] In the first step, the patient sits at a table, places the longitudinal axis of the wearable device 12 parallel to the little finger, and fixes the wearable device 12 on the little finger with a strap to check whether it is firmly fixed.

[0081] The second step is to open the applet and set the acceleration to zero.

[0082] The third step is to hold the test hand vertically upward, and adjust the movement angle to zero before testing. A pad can be used to secure the forearm and wrist, and the patient can be asked to warm up by repeatedly making a fist and straightening the fingers for 10 seconds.

[0083] The fourth step is to start the test. Within 15 seconds, the subject must quickly and forcefully repeat the process of clenching and straightening their fingers. The motion parameters during the test are collected. The patient's test time, acceleration, angular velocity, and angle values ​​along the three axes are obtained. The spatial integrated acceleration value and motion period are calculated. These characteristics, along with the patient's age, gender, and finger length, are input into the program to determine whether the patient has cervical spondylotic myelopathy. All these characteristics are then input into the program to determine whether the patient has cervical spondylotic myelopathy.

[0084] In the process of model selection, the host computer in the embodiment of the present application selects a long short-term memory network (LSTM) as the basic model, given that the little finger kinematic data has time series characteristics. LSTM can effectively process long-term dependencies in time series through a gating mechanism, and has significant advantages for analyzing the changing trend of the little finger movement over time. In order to further improve the performance of the model, a hybrid model is constructed in combination with a convolutional neural network (CNN). CNN has powerful feature extraction capabilities and can automatically extract spatial features from the little finger movement data. Combined with LSTM, it can simultaneously capture the spatiotemporal characteristics of the data and improve the ability to recognize features related to cervical spondylotic myelopathy.

[0085] In terms of model structure design, the input layer arranges the instantaneous linear acceleration, instantaneous angular velocity, and instantaneous rotation angle values ​​in a time series to form multidimensional time series data as input. For example, the acceleration, angular velocity, and angle data over a period of time (e.g., 10 seconds, depending on the specific experimental task) are sampled at a frequency of 100 sampling points per second to form an input tensor of shape (1000, 3), where 1000 represents the number of sampling points and 3 represents the three parameter dimensions of acceleration, angular velocity, and angle.

[0086] Convolutional layers: 2-3 convolutional layers are set up at the front end of the model. Each convolutional layer uses convolution kernels of different sizes, such as 3x1 and 5x1, to perform convolution operations on the input data. The convolutional layer extracts local features from the data, such as the acceleration trend of the pinky finger over a short period of time or sudden changes in angle. The convolutional layer is followed by a pooling layer, which uses maximum pooling or average pooling to downsample the convolutional feature map, reducing the data dimension and computational complexity while retaining important features.

[0087] LSTM layer: Connected after the convolutional layer, it consists of 2-3 layers of LSTM units. The hidden layer size of each LSTM unit can be adjusted based on actual conditions, such as 128 or 256. The LSTM layer is responsible for learning long-term dependencies in the data and capturing the changing patterns of pinky movement over a long period of time. For example, it can analyze the overall trend and periodic changes of pinky movement over a period of time.

[0088] Fully connected layer: After the LSTM layer, connect one or two fully connected layers. The fully connected layer integrates the feature vectors output by the LSTM layer and maps them to the final classification space. In the fully connected layer, the ReLU activation function is used to increase the model's nonlinear expression capabilities and improve the model's ability to classify complex features.

[0089] Output layer: For a diagnosis task involving cervical spondylotic myelopathy (a binary classification of 0 or 1), the output layer of the deep learning model contains one node and uses a sigmoid activation function. The sigmoid function maps the output of a neuron to a value between 0 and 1, directly representing the probability that a sample belongs to a particular category. Output values ​​close to 1 indicate a high probability that the model predicts the sample has cervical spondylotic myelopathy; values ​​close to 0 indicate a low probability.

[0090] For multi-classification problems (for example, distinguishing between different severity levels of cervical spondylotic myelopathy, such as mild, moderate, and severe), the output layer needs to contain more nodes, the same number of which corresponds to the number of categories. The Softmax function is an ideal choice. The model's output is converted into a probability distribution corresponding to each category, where the sum of all category probabilities is 1. Using the Softmax function, the model can clearly indicate the probability that each sample belongs to a different category, allowing doctors to make diagnostic decisions based on the category with the highest probability.

[0091] The clinical data in the embodiments of the present application, for example, the patient's age, gender, finger length and other information may also play an important role in the diagnosis of cervical spondylotic myelopathy, and the clinical data need to be incorporated into the algorithm model. The above-mentioned algorithm model is used to extract features of instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value to obtain a feature representation of sensor data. At the same time, a fully connected network is designed for the patient's static information to extract features and obtain a feature representation of static information. Before the classification layer of the model, the clinical data and instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value are spliced ​​or other fusion operations (such as addition, multiplication, etc.) are performed, and then final classification is performed.

[0092] In model training:

[0093] First, it is necessary to collect labeled data that is appropriate for the classification target. For example, for a two-classification task, it is necessary to collect a large amount of sensor data collected by patients with cervical spondylotic myelopathy and normal control subjects wearing wearable devices while performing the same task. The label corresponding to the data is whether the subject is sick or not. The data and the label are provided to the network together, and the network prediction value is calculated through forward propagation. The error function is calculated against the label, and the gradient of all parameters in the network is calculated using backpropagation. The parameters are updated using gradient descent to achieve the purpose of training. In practice, it is also necessary to comprehensively consider issues such as data preprocessing, overfitting, underfitting, etc. in order to achieve the best training effect, and iteratively update the model structure design based on the training and test results.

[0094] In terms of model deployment, the trained model parameters are fixed and deployed in the host computer software. During use, the subject performs the experimental task according to the doctor's instructions. The host computer obtains real-time sensor data from the device during the experiment. After preprocessing, the data is input into the algorithm model for analysis, resulting in an assessment and diagnosis of cervical spondylotic myelopathy.

[0095] It should be noted that the spinal cervical spondylotic myelopathy identification model and the spinal cervical spondylotic myelopathy diagnosis model in the embodiments of the present application adopt a cascade model. First, the spinal cervical spondylotic myelopathy identification model is trained to exclude easily confused diseases. Further, the spinal cervical spondylotic myelopathy diagnosis model is used to screen out more serious spinal cervical spondylotic myelopathy to avoid misdiagnosis of peripheral neuropathy or motor neuron disease as spinal cervical spondylotic myelopathy, prevent the aggravation of the disease that may be caused by surgery, and screen out serious cervical spondylotic myelopathy to avoid irreversible damage to the spinal cord function of the cervical spondylotic myelopathy that may be caused by delayed surgery.

[0096] It should be noted that the embodiment of the present application constructs a high-quality training data set by collecting a large amount of little finger kinematic data from patients with cervical spondylotic myelopathy and healthy people. The data should cover samples of different disease severity, different age groups and different genders to enhance the generalization ability of the model. During the training process, appropriate loss functions, optimization algorithms and hyperparameters are selected, such as using the cross entropy loss function and the Adam optimization algorithm, and by continuously adjusting hyperparameters such as the learning rate and the number of iterations, the model achieves optimal performance. At the same time, data enhancement techniques are used, such as performing transformations such as translation and rotation on the data, to expand the training data and prevent the model from overfitting.

[0097] In addition, in terms of diagnostic accuracy evaluation and calibration, the embodiments of the present application use an independent test data set to strictly evaluate the trained model, and use indicators such as accuracy, recall rate, and F1 value to measure the diagnostic accuracy of the model for cervical spondylotic myelopathy. If the model performance does not meet expectations, it is necessary to analyze the reasons and adjust the model structure or training parameters. In addition, due to individual differences and the complexity of the actual use environment, it is necessary to calibrate the device and algorithm regularly to ensure that it can work stably and accurately in different populations and environments. For example, according to the differences in physical characteristics of people in different regions and different races, the model is calibrated in a targeted manner to improve the reliability of diagnosis.

[0098] Optionally, the host computer 14 is further configured to receive the acquisition time, the instantaneous linear acceleration and instantaneous angular velocity of a specific finger, calculate the number of grips within the specific time based on the instantaneous angular velocity, combine the clinical data with one or a combination of the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity, and the number of grips within the specific time, input the data into a pre-trained cervical spondylotic myelopathy identification model, obtain a screening result, and input the screening result into a pre-trained cervical spondylotic myelopathy diagnosis model to obtain a diagnosis result;

[0099] Calculating the number of grips within a specific time period based on the instantaneous angular velocity includes: obtaining an average grip period by dividing the test time by the number of trough values ​​in the time series of the instantaneous angular velocity; and obtaining the number of grips within the specific time period by dividing the specific time by the average grip period.

[0100] Specifically, the test time t is divided by the number of trough values ​​n in the time series of the instantaneous angular velocity in the motion plane to obtain the average gripping period t / n, and then 10 seconds (i.e., the specific time in the embodiment of the present application) is divided by the average gripping period to obtain the number of grips (or frequency) within the specific time of 10 seconds.

[0101] Since different ages, genders, BMIs and finger lengths will produce different accelerations, angular velocities, angles and the number of grips within a specific time; therefore, in order to ensure the accuracy of subsequent data processing in the embodiments of this application, the user's clinical data is combined with the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​to input a pre-trained spinal cord type cervical spondylotic myelopathy identification model.

[0102] Optionally, the host computer 14 is further used to obtain target cervical spondylotic myelopathy sample data and normal sample data, and calculate the acceleration average, TopN acceleration average, maximum acceleration value, angular velocity average, TopN angular velocity average, maximum angular velocity value, and the number of grips within a specific time based on the target cervical spondylotic myelopathy sample data and the normal sample data respectively; generate an ROC curve based on the acceleration average, TopN acceleration average, maximum acceleration value, angular velocity average, TopN angular velocity average, maximum angular velocity value, and the number of grips within a specific time based on the target cervical spondylotic myelopathy sample data and the normal sample data respectively, calculate the area under the ROC curve and the cutoff value, and generate a diagnosis result based on the cutoff value; or,

[0103] The host computer 14 is also used to judge the direction of force based on the positive and negative values ​​of the angular velocity, determine the finger flexion acceleration and / or finger extension acceleration, and determine the extreme values, fatigue and strength reserves of the finger flexion acceleration and finger extension acceleration, and visualize the extreme values, fatigue and strength reserves of the finger flexion acceleration and finger extension acceleration so that they can be presented to the user and classified and judged through deep learning.

[0104] Specifically, except Figure 2 According to the method shown, the host computer 14 in the embodiment of the present application uses the average value of acceleration, the top ten average values ​​(i.e., the average value of acceleration of TopN in the embodiment of the present application), the maximum acceleration value, the average value of angular velocity, the top ten average values ​​(i.e., the average value of angular velocity of TopN in the embodiment of the present application), and the maximum angular velocity value, and uses one or more of the average value of acceleration, the average value of acceleration of TopN, the maximum acceleration value, the average value of angular velocity, the average value of angular velocity of TopN and the maximum angular velocity value as a boundary value judgment for the diagnosis of cervical spondylosis under spinal microscopy.

[0105] Among them, 500 (or more than 500) patients with cervical spondylotic myelopathy and 500 healthy people of corresponding age groups were selected to test and extract the average acceleration, TopN acceleration average, maximum acceleration value, angular velocity average, TopN angular velocity average and maximum angular velocity values. Then, according to the ROC curve, the area under the ROC curve and the cutoff value were calculated, and whether it was cervical spondylotic myelopathy was judged according to the cutoff value.

[0106] It should be noted that the embodiments of this application only use 500 patients with cervical spondylotic myelopathy and 500 healthy people of corresponding age groups as examples for illustration, based on the data processing system based on finger motion feature analysis provided in the embodiments of this application, and the specific capture limitations are based on this.

[0107] In a preferred example, the host computer 14 determines the user's fatigue level by calculating the ratio of the average acceleration value to the average acceleration value in adjacent time periods, and the ratio of the angular velocity to the angular velocity in adjacent time periods. Specifically, the fatigue level of the patient (i.e., the user in the embodiment of the present application) can be determined by calculating the ratio of the average acceleration value in the first ten seconds to the average acceleration value in the next ten seconds, and the ratio of the angular velocity in the first ten seconds to the angular velocity in the next ten seconds. The first ten seconds and the last ten seconds are adjacent time periods in the embodiment of the present application.

[0108] In a preferred example, the upper computer 14 determines the direction of force based on the positive and negative values ​​of the angular velocity, determines the finger flexion acceleration and / or finger extension acceleration, and determines the extreme values ​​of the finger flexion acceleration and the finger extension acceleration as well as the fatigue level, and visualizes the extreme values ​​of the finger flexion acceleration and the finger extension acceleration as well as the fatigue level so as to display them to the user, and classifies and distinguishes them through deep learning. Since the muscle strength classification method in the prior art is based on whether it can resist gravity, that is, an acceleration of 1G is a level 3 muscle strength, the number of Gs can be used to more quantitatively evaluate the muscle strength; the acceleration of long fingers is generally large, and when the fingers are short, the activity angle is generally small. It can be seen that it is valuable to count the patient's finger length, height and weight, and the collected and calculated variables can be used for machine learning or deep learning of raw data.

[0109] It should be noted that the values ​​and sample quantities in the above examples are only described as preferred examples, and are based on the data processing system based on finger motion feature analysis provided in the embodiment of the present application, and are not specifically limited.

[0110] The data processing system based on finger motion feature analysis in the embodiment of the present application can collect multi-dimensional data, conduct deep learning on the relationship with time, and use algorithms as a tool to objectively determine spinal cord disease hands. It can also identify diseases that may also affect finger motor function, such as motor neuron disease; it can also facilitate users to decide whether to undergo surgical treatment, conduct preoperative monitoring or postoperative follow-up.

[0111] It should be noted that the spinal cord cervical spondylotic myelopathy identification model and the spinal cord cervical spondylotic myelopathy diagnosis model in the embodiments of the present application are only illustrated by deep learning as an example. In addition, they can also be implemented through machine learning, based on the data processing system based on finger motion feature analysis provided in the embodiments of the present application, without specific limitation.

[0112] The present invention adopts the above technical solution, by receiving the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of a specific finger; combining the pre-input user's clinical data with the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value, and inputting a pre-trained spinal cord type cervical spondylotic myelopathy identification model to obtain a screening result; inputting the screening result into a pre-trained spinal cord type cervical spondylotic myelopathy diagnosis model to obtain a diagnosis result. Compared with the existing technology, the present invention has the following technical effects: combining multiple kinematic characteristics to accurately and quickly assist in diagnosis.

[0113] Example 2

[0114] In an exemplary embodiment of the present invention, a wearable device provided in an embodiment of the present application is applied to the data processing system based on finger motion feature analysis in Example 1. The wearable device in the embodiment of the present application is worn on a specific finger of a user, and includes:

[0115] Accelerometer, gyroscope, angle sensor and communication module, wherein the accelerometer is used to collect the instantaneous linear acceleration of a specific finger on three vertical axes; the gyroscope is used to collect the instantaneous angular velocity of a specific finger in three vertical planes; the angle sensor is used to collect the instantaneous rotation angle value of a specific finger in three vertical planes; the communication module is connected to the accelerometer, gyroscope and angle sensor respectively, and is used to send the collection time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value to the host computer; wherein the wearable device also includes: a circuit board and a hardware synchronization circuit, wherein the circuit board adopts multi-layer wiring, and the accelerometer, gyroscope, The location of the angle sensor and communication module; the hardware synchronization circuit is a master-slave architecture, with the accelerometer set as the master sensor, the gyroscope and the angle sensor as slave sensors, and the master sensor controls the sampling frequency and timing of the slave sensors, and is used to control the accelerometer, gyroscope and angle sensor to collect data simultaneously; the communication module is also used to send the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​to the host computer through wireless low-power communication technology according to a preset transmission strategy; wherein the preset transmission strategy includes: compressing the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​through a lossy compression algorithm, and adjusting the transmission rate according to the network status.

[0116] Example 3

[0117] An exemplary embodiment of the present invention is as follows Figure 3 As shown, Figure 3 1 is a flow chart of a data processing method based on finger motion feature analysis according to a third embodiment of the present invention, which is applied to the data processing system based on finger motion feature analysis in embodiment 1. On the host computer side, the data processing method based on finger motion feature analysis provided by the embodiment of the present application includes:

[0118] Step S300, receiving the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value of the specific finger;

[0119] Step S302 , combining the pre-input user clinical data with the acquisition time, instantaneous linear acceleration, instantaneous angular velocity, and instantaneous rotation angle values, and inputting them into a pre-trained cervical spondylotic myelopathy identification model to obtain a screening result;

[0120] Step S304: input the screening result into the pre-trained cervical spondylotic myelopathy diagnosis model to obtain a diagnosis result.

[0121] Optionally, the data processing method based on finger motion feature analysis provided in the embodiment of the present application also includes: configuring a cervical spondylotic myelopathy identification model and a cervical spondylotic myelopathy diagnosis model in the host computer; in the model training stage of the cervical spondylotic myelopathy identification model and the cervical spondylotic myelopathy diagnosis model, inputting the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data into the cervical spondylotic myelopathy identification model for training, and when the sensitivity reaches a first threshold, determining that the cervical spondylotic myelopathy identification model has converged; after excluding the similar disease sample data, inputting the target cervical spondylotic myelopathy sample data and normal sample data into the cervical spondylotic myelopathy diagnosis model for training, and when the specificity reaches a second threshold, determining that the cervical spondylotic myelopathy diagnosis model has converged; wherein, the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data all include: acquisition time, clinical data, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of a specific finger; wherein, in the model training stage, the target cervical spondylotic myelopathy sample data , normalize the acquisition time, clinical data, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of a specific finger in the similar disease sample data and the normal sample data to obtain the normalized target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data; classify the normalized target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data into two categories to obtain first classification data and second classification data, input the first classification data and the second classification data into the cervical spondylotic myelopathy identification model for training, and when the sensitivity reaches the first threshold, determine that the cervical spondylotic myelopathy identification model has converged; wherein the first classification data is the normalized similar disease sample data, and the second classification data is the normalized target cervical spondylotic myelopathy sample data and normal sample data; input the normalized target cervical spondylotic myelopathy sample data and the normalized normal sample data into the cervical spondylotic myelopathy diagnosis model for training, and when the specificity reaches the second threshold, determine that the cervical spondylotic myelopathy diagnosis model has converged;

[0122] During the model application stage of the spinal cervical spondylotic myelopathy identification model and the spinal cervical spondylotic myelopathy diagnosis model, the user's clinical data is combined with the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values, and input into the spinal cervical spondylotic myelopathy identification model for similar disease screening to obtain data after excluding similar diseases, and the data is determined as the screening result; the screening result is input into the spinal cervical spondylotic myelopathy diagnosis model to obtain the diagnosis result; wherein, during the model application stage, the acquisition time, clinical data, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle values ​​of a specific finger are normalized to obtain a normalized data group, and the normalized data group is input into the spinal cervical spondylotic myelopathy identification model for similar disease screening to obtain data after excluding similar diseases, and the data is determined as the screening result; the screening result is input into the spinal cervical spondylotic myelopathy diagnosis model to obtain the diagnosis result.

[0123] Optionally, the data processing method based on finger motion feature analysis provided in the embodiment of the present application further includes: calculating the acceleration average, the TopN acceleration average, the maximum acceleration value, the angular velocity average, the TopN angular velocity average, the maximum angular velocity value, and the number of grips within a specific time based on the target cervical spondylotic myelopathy sample data and the normal sample data, respectively; generating an ROC curve based on the acceleration average, the TopN acceleration average, the maximum acceleration value, the angular velocity average, the TopN angular velocity average, the maximum angular velocity value, and the number of grips within a specific time based on the target cervical spondylotic myelopathy sample data and the normal sample data; calculating the area under the ROC curve and the cutoff value; and generating a diagnosis result based on the cutoff value;

[0124] or,

[0125] The user's fatigue level is determined by calculating the ratio of the average acceleration to the average acceleration of adjacent time periods, and the ratio of the angular velocity to the angular velocity of adjacent time periods; the direction of force is judged according to the positive or negative value of the angular velocity, the finger flexion acceleration and / or finger extension acceleration is determined, and the extreme values, fatigue and strength reserves of the finger flexion acceleration and finger extension acceleration are determined, and the extreme values, fatigue and strength reserves of the finger flexion acceleration and finger extension acceleration are visualized so that they can be presented to the user and classified and judged through deep learning.

[0126] The present invention adopts the above technical solution, by receiving the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of a specific finger; combining the pre-input user's clinical data with the acquisition time, instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value, and inputting a pre-trained spinal cord type cervical spondylotic myelopathy identification model to obtain a screening result; inputting the screening result into a pre-trained spinal cord type cervical spondylotic myelopathy diagnosis model to obtain a diagnosis result. Compared with the existing technology, the present invention has the following technical effects: combining multiple kinematic characteristics to accurately and quickly assist in diagnosis.

[0127] The above description is only a preferred embodiment of the present invention and does not limit the implementation mode and protection scope of the present invention. For those skilled in the art, it should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the description and illustrations of the present invention should be included in the protection scope of the present invention.

Claims

1. A data processing system based on finger motion feature analysis, characterized in that: include: Wearable device and host computer, wherein, The wearable device is worn on a specific finger of a user, and the wearable device includes: an accelerometer, a gyroscope, an angle sensor and a communication module, wherein the accelerometer is used to collect the instantaneous linear acceleration of the specific finger on three vertical axes; the gyroscope is used to collect the instantaneous angular velocity of the specific finger in three vertical planes; the angle sensor is used to collect the instantaneous rotation angle value of the specific finger in three vertical planes; the communication module is connected to the accelerometer, the gyroscope and the angle sensor respectively, and is used to send the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value to the host computer; The host computer is connected to the wearable device and is used to receive the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value of the specific finger; the pre-input clinical data of the user is combined with the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value, and input into a pre-trained spinal cord type cervical spondylotic myelopathy identification model to obtain a screening result; the screening result is input into a pre-trained spinal cord type cervical spondylotic myelopathy diagnosis model to obtain a diagnosis result.

2. The data processing system based on finger motion feature analysis according to claim 1, characterized in that: The wearable device further includes: a circuit board, a hardware synchronization circuit and a cache device, wherein: The circuit board adopts multi-layer wiring to set the positions of the accelerometer, the gyroscope, the angle sensor and the communication module in a preset layout; The hardware synchronization circuit adopts a master-slave architecture, wherein the accelerometer is set as a master sensor, and the gyroscope and the angle sensor are slave sensors. The master sensor controls the sampling frequency and timing of the slave sensors, and is used to control the accelerometer, the gyroscope and the angle sensor to collect data simultaneously. The cache device is located in the wearable device and adopts a ring buffer structure, which is used to write data into the ring buffer in sequence. When the ring buffer stores the maximum value, the data written at the Nth moment overwrites the data written at the 1st moment; during the data transmission process, if the network fails or is interrupted, the collected data is first stored in the ring buffer; if the network is restored, the data in the ring buffer is transmitted to the host computer in sequence through the communication module.

3. The data processing system based on finger motion feature analysis according to claim 1 or 2, characterized in that: The communication module is further used to send the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value to the host computer according to a preset transmission strategy through wireless low-power communication technology; wherein the preset transmission strategy includes: compressing the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value through a lossy compression algorithm, and adjusting the transmission rate according to the network status.

4. The data processing system based on finger motion feature analysis according to claim 3, characterized in that: The host computer is configured with the cervical spondylotic myelopathy identification model and the cervical spondylotic myelopathy diagnosis model, wherein: During the model training stage of the cervical spondylotic myelopathy identification model and the cervical spondylotic myelopathy diagnosis model, the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data are input into the cervical spondylotic myelopathy identification model for training, and when the sensitivity reaches a first threshold, it is determined that the cervical spondylotic myelopathy identification model has converged; after excluding the similar disease sample data, the target cervical spondylotic myelopathy sample data and the normal sample data are input into the cervical spondylotic myelopathy diagnosis model for training, and when the specificity reaches a second threshold, it is determined that the cervical spondylotic myelopathy diagnosis model has converged; wherein the target cervical spondylotic myelopathy sample data, the similar disease sample data and the normal sample data all include: the acquisition time, the clinical data, the instantaneous linear acceleration of the specific finger, the instantaneous angular velocity and the instantaneous rotation angle value; Wherein, in the model training stage, the acquisition time, the clinical data, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value of the target cervical spondylotic myelopathy sample data, the similar disease sample data and the normal sample data are normalized to obtain the normalized target cervical spondylotic myelopathy sample data, the similar disease sample data and the normal sample data; the normalized target cervical spondylotic myelopathy sample data, the similar disease sample data and the normal sample data are binary classified to obtain first classification data and second classification data, and the first classification data is classified into the first classification data and the second classification data. The first classification data and the second classification data are input into the cervical spondylotic myelopathy identification model for training, and when the sensitivity reaches the first threshold, the cervical spondylotic myelopathy identification model is determined to have converged; wherein the first classification data is the normalized similar disease sample data, and the second classification data is the normalized target cervical spondylotic myelopathy sample data and the normal sample data; the normalized target cervical spondylotic myelopathy sample data and the normalized normal sample data are input into the cervical spondylotic myelopathy diagnosis model for training, and when the specificity reaches the second threshold, the cervical spondylotic myelopathy diagnosis model is determined to have converged; During the model application phase of the cervical spondylotic myelopathy identification model and the cervical spondylotic myelopathy diagnosis model, the user's clinical data is combined with the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity, and the instantaneous rotation angle value, and input into the cervical spondylotic myelopathy identification model to perform similar disease screening, thereby obtaining data after eliminating similar diseases, and determining the data as the screening result; and inputting the screening result into the cervical spondylotic myelopathy diagnosis model to obtain the diagnosis result; Among them, in the model application stage, the acquisition time, the clinical data, the instantaneous linear acceleration of the specific finger, the instantaneous angular velocity and the instantaneous rotation angle value are normalized to obtain a normalized data group, and the normalized data group is input into the spinal cervical spondylotic myelopathy identification model to screen similar diseases, and the data after eliminating similar diseases is obtained, and the data is determined as the screening result; the screening result is input into the spinal cervical spondylotic myelopathy diagnosis model to obtain the diagnosis result.

5. The data processing system based on finger motion feature analysis according to claim 1, characterized in that: The host computer is further configured to receive the acquisition time, the instantaneous linear acceleration and the instantaneous angular velocity of the specific finger, calculate the number of grips within the specific time based on the instantaneous angular velocity, combine the clinical data with one or a combination of the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity, and the number of grips within the specific time, input the data into a pre-trained cervical spondylotic myelopathy identification model to obtain a screening result, and input the screening result into a pre-trained cervical spondylotic myelopathy diagnosis model to obtain a diagnosis result; The calculation of the number of grips within a specific time based on the instantaneous angular velocity includes: obtaining an average grip period by dividing the test time by the number of trough values ​​in the time series of the instantaneous angular velocity; and obtaining the number of grips within the specific time by dividing the specific time by the average grip period.

6. The data processing system based on finger motion feature analysis according to claim 1 or 5, characterized in that: The host computer is further used to obtain target cervical spondylotic myelopathy sample data and normal sample data, and calculate the acceleration average, TopN acceleration average, maximum acceleration value, angular velocity average, TopN angular velocity average, maximum angular velocity value, and the number of grips within a specific time based on the target cervical spondylotic myelopathy sample data and the normal sample data respectively; generate an ROC curve based on the acceleration average, TopN acceleration average, maximum acceleration value, angular velocity average, TopN angular velocity average, maximum angular velocity value, and the number of grips within a specific time based on the target cervical spondylotic myelopathy sample data and the normal sample data respectively, calculate the area under the ROC curve and the cutoff value, and generate the diagnosis result based on the cutoff value; or, The host computer is further configured to determine the user's fatigue level by calculating a ratio of an average value of acceleration to an average value of acceleration in adjacent time periods, and a ratio of an angular velocity to an angular velocity in adjacent time periods; The host computer is also used to judge the direction of force based on the positive and negative values ​​of the angular velocity, determine the finger flexion acceleration and / or finger extension acceleration, and determine the extreme values, fatigue and strength reserves of the finger flexion acceleration and the finger extension acceleration, and visualize the extreme values ​​of the finger flexion acceleration and the finger extension acceleration, the fatigue and the strength reserves so as to show them to the user and perform classification and judgment through deep learning.

7. A wearable device, characterized in that: A data processing system based on finger motion feature analysis is applied to a wearable device worn on a specific finger of a user, comprising: Accelerometer, gyroscope, angle sensor and communication module, among which, The accelerometer is used to collect the instantaneous linear acceleration of the specific finger on three perpendicular axes; The gyroscope is used to collect the instantaneous angular velocity of the specific finger in three perpendicular planes; The angle sensor is used to collect instantaneous rotation angle values ​​of the specific finger in three vertical planes; The communication module is connected to the accelerometer, the gyroscope and the angle sensor respectively, and is used to send the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value to the host computer; The wearable device further comprises: a circuit board, a hardware synchronization circuit and a cache device, wherein the circuit board adopts multi-layer wiring to set the positions of the accelerometer, the gyroscope, the angle sensor and the communication module in a preset layout; the hardware synchronization circuit is a master-slave architecture, the accelerometer is set as the master sensor, the gyroscope and the angle sensor are slave sensors, the master sensor controls the sampling frequency and timing of the slave sensors, and is used to control the accelerometer, the gyroscope and the angle sensor to collect data simultaneously; the cache device is located in the wearable device and adopts a ring buffer structure to write data into the ring buffer in sequence. When the ring buffer stores the maximum value, the data written at the Nth time overwrites the data written at the 1st time; during the data transmission process, if the network fails or is interrupted, the collected data is first stored in the ring buffer; if the network is restored, the data in the ring buffer is transmitted to the host computer in sequence through the communication module; The communication module is further used to send the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value to the host computer according to a preset transmission strategy through wireless low-power communication technology; wherein the preset transmission strategy includes: compressing the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value through a lossy compression algorithm, and adjusting the transmission rate according to the network status.

8. A data processing method based on finger motion feature analysis, characterized in that: The data processing system used for finger motion feature analysis includes: Receive the acquisition time, the instantaneous linear acceleration, instantaneous angular velocity and instantaneous rotation angle value of a specific finger; Combining the pre-input clinical data of the user with the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value, and inputting the pre-trained cervical spondylotic myelopathy identification model to obtain a screening result; The screening results are input into a pre-trained cervical spondylotic myelopathy diagnosis model to obtain a diagnosis result.

9. The data processing method based on finger motion feature analysis according to claim 8, characterized in that: The method further comprises: Configuring the cervical spondylotic myelopathy identification model and the cervical spondylotic myelopathy diagnosis model in a host computer; In the model training stage of the cervical spondylotic myelopathy identification model and the cervical spondylotic myelopathy diagnosis model, the target cervical spondylotic myelopathy sample data, similar disease sample data and normal sample data are input into the cervical spondylotic myelopathy identification model for training. When the sensitivity reaches a first threshold, it is determined that the cervical spondylotic myelopathy identification model has converged; after excluding the similar disease sample data, the target cervical spondylotic myelopathy sample data and the normal sample data are input into the cervical spondylotic myelopathy diagnosis model for training. When the specificity reaches a second threshold, it is determined that the cervical spondylotic myelopathy diagnosis model has converged; wherein the target cervical spondylotic myelopathy sample data, the similar disease sample data and the normal sample data all include: the acquisition time, the clinical data, the instantaneous linear acceleration of the specific finger, the instantaneous angular velocity and the instantaneous rotation angle value; wherein, in the model training stage, the acquisition time, the clinical data, the instantaneous linear acceleration of the specific finger, the instantaneous angular velocity and the instantaneous rotation angle value in the target cervical spondylotic myelopathy sample data, the similar disease sample data and the normal sample data are input into the cervical spondylotic myelopathy diagnosis model for training. Normalizing the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value to obtain the normalized target cervical spondylotic myelopathy sample data, the similar disease sample data and the normal sample data; performing binary classification on the normalized target cervical spondylotic myelopathy sample data, the similar disease sample data and the normal sample data to obtain first classification data and second classification data, inputting the first classification data and the second classification data into the cervical spondylotic myelopathy identification model for training, and determining that the cervical spondylotic myelopathy identification model has converged when the sensitivity reaches the first threshold; wherein the first classification data is the normalized similar disease sample data, and the second classification data is the normalized target cervical spondylotic myelopathy sample data and the normal sample data; inputting the normalized target cervical spondylotic myelopathy sample data and the normal sample data into the cervical spondylotic myelopathy diagnosis model for training, and determining that the cervical spondylotic myelopathy diagnosis model has converged when the specificity reaches the second threshold; During the model application stage of the spinal cervical spondylotic myelopathy identification model and the spinal cervical spondylotic myelopathy diagnosis model, the user's clinical data is combined with the acquisition time, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value, and input into the spinal cervical spondylotic myelopathy identification model for similar disease screening to obtain data after excluding similar diseases, and the data is determined as the screening result; the screening result is input into the spinal cervical spondylotic myelopathy diagnosis model to obtain the diagnosis result; wherein, during the model application stage, the acquisition time, the clinical data, the instantaneous linear acceleration, the instantaneous angular velocity and the instantaneous rotation angle value of the specific finger are normalized to obtain a normalized data group, the normalized data group is input into the spinal cervical spondylotic myelopathy identification model for similar disease screening to obtain data after excluding similar diseases, and the data is determined as the screening result; the screening result is input into the spinal cervical spondylotic myelopathy diagnosis model to obtain the diagnosis result.

10. The data processing method based on finger motion feature analysis according to claim 8, characterized in that: The method further comprises: Obtain target cervical spondylotic myelopathy sample data and normal sample data; calculate the acceleration average, TopN acceleration average, maximum acceleration value, angular velocity average, TopN angular velocity average, maximum angular velocity value, and the number of grips within a specific time based on the target cervical spondylotic myelopathy sample data and the normal sample data, respectively; generate an ROC curve based on the acceleration average, TopN acceleration average, maximum acceleration value, angular velocity average, TopN angular velocity average, maximum angular velocity value, and the number of grips within a specific time of the target cervical spondylotic myelopathy sample data and the normal sample data; calculate the area under the ROC curve and a cutoff value; and generate the diagnosis result based on the cutoff value; or, The user's fatigue level is determined by calculating the ratio of the average acceleration of adjacent time periods to the average acceleration, and the ratio of the angular velocity to the angular velocity of adjacent time periods; the direction of force is judged according to the positive and negative values ​​of the angular velocity, the finger flexion acceleration and / or finger extension acceleration is determined, and the extreme values ​​of the finger flexion acceleration and the finger extension acceleration, the fatigue level and the strength reserve are determined, and the extreme values ​​of the finger flexion acceleration and the finger extension acceleration, the fatigue level and the strength reserve are visualized so as to be presented to the user and classified and identified through deep learning.

Images