Visualizing a pulse method

Abstract

A kind of visualizing pulse-taking method, comprising the following steps: step S1, the pulse condition acquisition part of acquisition terminal is connected to display terminal by wireless transmission, wearable part is worn on one wrist, pulse-taking mouth is close to radial artery, cunkuanchizi pulse position is in pulse-taking mouth, palm is upward, infrared lamp module is placed at the bottom of wrist, and infrared lamp module is started;Step S2, the other hand holds adjustment acquisition seat, the bottom infrared camera of adjustment acquisition seat is irradiated pulse-taking mouth, and the display terminal is observed in real time the developing position of radial artery;Step S3, the position of pulse-taking mouth is adjusted, so that radial artery is located in the center of pulse-taking mouth, pulse condition acquisition part is connected on pulse-taking frame assembly, and pulse condition acquisition can be started.The present application has the advantages of precise and rapid positioning of radial artery position by visual method, convenient to use, etc.

Description

Technical Field

[0001] This invention relates to the visualization and acquisition of pulse patterns, specifically a visualization method for pulse diagnosis that uses automated processing of infrared camera images to locate the radial artery in real time. Background Technology

[0002] Pulse signals can reflect a wealth of physiological and pathological information. Traditional Chinese medicine pulse diagnosis mainly involves pressing the radial artery of the arm with the fingers, applying three levels of pressure—superficial, moderate, and deep—to collect pulse signals in the "cun," "guan," and "chi" areas. This method is characterized by its speed and non-invasiveness. With the development of biomedical detection technology, the development of pulse diagnosis devices has gradually become a focus of attention, especially wearable pulse diagnosis devices, which have the advantage of real-time pulse signal monitoring and have become a hot topic in the field of pulse diagnosis device research in recent years.

[0003] In existing technologies, the technical solutions for radial artery localization and detection mainly involve IPC classification number A61B5 / 02. Most existing pulse acquisition and diagnostic systems are single-point acquisition and single-point reproduction systems. However, as those skilled in the art know, during a pulse pulsation, the pulse actually spreads outwards from the radial artery as the midline, exhibiting a trend of amplitude decreasing from high to low. Therefore, during acquisition, the pulse should be detected over as wide a range as possible to obtain the optimal acquisition point. However, it is almost impossible to obtain the optimal acquisition location through single-point remote acquisition, thus inevitably leading to misdiagnosis in remote reproduction. To achieve pulse acquisition and reproduction, the following technical solutions have been disclosed in the prior art.

[0004] The invention patent application with publication number CN107898445A discloses a wearable intelligent pulse diagnosis device, including an adaptive wristband and a probe group. The probe group consists of four independently movable probes and a longitudinal linear motion drive unit, which can drive the tip of each probe to automatically locate the cun, guan, chi and chi positions along the X, Y and Z directions, automatically apply pressure, and stably monitor and record pulse signals under different pressure stimuli.

[0005] However, this technical solution has the following drawbacks: the solution only determines the approximate location of the radial artery based on the pulse amplitude after linear arrangement of the finger pressure. Due to the inability to achieve visualization and the differences in the physiological structure of the radial artery (blood vessel) between different patients, it is difficult to accurately locate the optimal pulse position.

[0006] Furthermore, patent applications CN113440113A and CN114224308A also disclose similar schemes for locating the radial artery by acquiring the pulse amplitude signal after pressure. Patent application CN112842292A also discloses a scheme for detecting cerebral arteries using a combination of an airbag wristband and an array sensor. Patent application CN114569087b further discloses a scheme for acquiring radial artery pulsation information using a combination of a non-penetrating visible light and infrared camera and an array sensor.

[0007] The invention patent application with publication number CN114041758A discloses a method for non-contact localization of the radial artery based on deep learning. However, it is obvious that this solution relies on a large amount of manually labeled data, and can only roughly locate the "area where the radial artery is located" rather than accurately locate the main vessel of the radial artery through video data. Therefore, its training results cannot be used for interaction with remote pulse diagnosis medical care terminal (reproduction terminal).

[0008] Utility model patent CN217090710U discloses a wristband for home-based elderly status recognition based on pulse detection. It uses infrared emitting lights and photodiodes respectively located in the upper and lower rings to detect the infrared light transmitted through the wrist pulse in real time. While this solution differs from single-sided adhesive infrared detection, it only measures heart rate by detecting pulse fluctuations and is not used to locate the radial artery.

[0009] In the existing technology, the technical solutions for radial artery signal reproduction mainly involve IPC classification number A61B5 / 02:

[0010] The invention patent applications published under CN108742546A, CN112690764A, CN113367669A, CN114864630A, and CN115721271A all employ three single-point pulse sensors in their disclosed technical solutions. Because there are only three "tactile feedback points," it is impossible to completely reproduce the array sensor data of the acquisition terminal through a point-to-point method. Therefore, it is impossible to measure pulse width and pulse length, and it is also impossible to accurately measure pulse rhythm (the interval and contour of the pulse).

[0011] Furthermore, the technical solution disclosed in the invention patent application with publication number CN115721271A uses a combination of finger-type flexible electrode array and electrical stimulation element in the reproduction module. However, it is obvious that compared with the physical reproduction of directly simulating blood vessel fluctuations, the electrical stimulation reproduction effect is poor, and the technical inspiration of point-to-point reproduction between the acquisition terminal and the reproduction terminal is not disclosed.

[0012] Existing pulse diagnosis instruments mostly adopt the "cunkou pulse taking" method, which converts the pulsation signal of the pulse into an electronic signal, thereby realizing the informatization and intelligentization of pulse diagnosis.

[0013] Patent application CN106137147b discloses a device and method for acquiring pulse data based on human-computer interaction. It primarily determines whether the pulse signal collected by the sensor element at the center position is the strongest pulse signal. If so, it instructs the user to apply corresponding pressure and acquires the pulse waveform corresponding to the pressure. If not, it instructs the user to move the sensor element at the center position towards the strongest pulse signal position, until the sensor element at the center position reaches the strongest pulse signal position. The strongest pulse signal is then converted into pulse data and displayed to the user. This method mainly obtains the optimal pulse value at the "guan" position. However, in clinical pulse diagnosis, the sensitivity of the fingers is higher than that of the sensor, and the pressure applied by the fingers at the cun, guan, and chi positions varies, resulting in low accuracy of this technical solution.

[0014] Patent application CN105534490A discloses a finger-pressure pulse diagnosis device and its control method. The method includes: Step S1, using a pulse diagnosis device with multiple finger sleeve sensors to collect the superficial, middle, and deep voltage output signals of the three finger positions (cun, guan, chi) of the subject; Step S2, the pulse diagnosis device transmits the superficial, middle, and deep voltage output signals to a computer; Step S3, the computer converts the superficial, middle, and deep voltage output signals into superficial, middle, and deep pulse-taking pressure values, compares these pressure values ​​with a preset superficial, middle, and deep pulse-taking pressure range, and performs corresponding operations on the pressure values ​​based on the comparison results; Step S4, the computer saves the comparison results and the operation results obtained after operating on the superficial, middle, and deep pulse-taking pressure values. This technical solution focuses on superficial, middle, and deep pulse-taking at the three positions, but the optimal pulse-taking pressure for each position is unclear. Summary of the Invention

[0015] The purpose of this invention is to provide a visual pulse diagnosis method that fundamentally solves the aforementioned problems. It offers advantages such as ease of use, accurate radial artery localization, and high pulse pattern reproduction precision. It better simulates clinical pulse diagnosis and improves the accuracy of pulse diagnosis using pulse diagnostic instruments.

[0016] To achieve the above objectives, the present invention provides the following technical solution: This visual pulse diagnosis method includes the following steps:

[0017] Step S1: Connect the pulse acquisition unit of the acquisition terminal to the display terminal via wireless transmission, wear the wearable device on one wrist, place the pulse retrieval point close to the radial artery, place the cun, guan, and chi pulse positions in the pulse retrieval point, with the palm facing up, place the infrared lamp module at the bottom of the wrist, and turn on the infrared lamp module.

[0018] Step S2: Hold the adjustment acquisition seat with the other hand, irradiate the pulse acquisition port with the infrared camera at its bottom, and observe the imaging position of the radial artery in real time through the display terminal;

[0019] Step S3: Adjust the position of the pulse taking point so that the radial artery is located in the center of the pulse taking point, and attach the pulse image collection unit to the pulse taking frame assembly to start pulse image collection.

[0020] Furthermore, step S2 specifically includes:

[0021] Step P1: Acquire raw infrared images in real time;

[0022] Step P2, cut out the rectangular area where the radial artery is located:

[0023] Step P3: Apply a Laplacian filter to the rectangular region to sharpen the edges. The Laplacian operator (Equation 1) uses a four-neighbor convolution kernel (Equation 2).

[0024] Equation (1) Equation (2)

[0025] Step P4: Binarize the edge-processed image and locate the radial artery by extracting bones;

[0026] Step P5: Mark the outer contour of the radial artery using the Grabcut algorithm:

[0027] Step P6: Overlay the image with the marked outer contour onto the rectangular area where the pulse point of the original image is located, and output the integrated image to the display terminal.

[0028] Furthermore, step P2 specifically includes:

[0029] Step P21, Locating the radial and ulnar sides: The ulnar side of the pulse taking frame component is pre-set with specific positioning points for correcting the orientation of the original image. The specific positioning points are located by the Canny algorithm and the Hough transform algorithm.

[0030] Step P22: After identifying the edge using the edge detection algorithm, the rectangular area of ​​the pulse taking frame component is obtained;

[0031] Step P23: Based on the location of the rectangular region and the specific positioning point, a calibrated horizontal line is obtained. After rotation correction, the image of the rectangular region where the radial artery is located is obtained after cropping based on the location of the specific positioning point.

[0032] Furthermore, step P4 specifically includes:

[0033] Step P41: Binarize the image after edge sharpening;

[0034] Step P42: Use Zhang-Suen's Algorithm to extract the skeleton from the binarized image and obtain multiple connected regions;

[0035] Step P43: Use the Depth-First Search (DFS) algorithm on each connected region to find the longest path Li in each connected region;

[0036] Step P44: Set the pruning threshold La to 10mm~30mm, compare Li and La of each connected region, exclude the region where the longest path is less than the pruning threshold, and locate the longest path among the remaining paths as the radial artery.

[0037] Furthermore, step P5 specifically includes:

[0038] The radial artery path obtained in step P4 is used to generate multiple tiny rectangles, with the number of rectangles being n. Each tiny rectangle takes three adjacent pixels of the current pixel. Each rectangle is input into the Grabcut algorithm to obtain the selected region S(i). This yields S(1)∪S(2)∪...∪S(n)=S, and the radial artery vessel edge is marked according to the region S.

[0039] The beneficial effects of this invention are as follows: In terms of the overall technical solution, this invention can achieve pulse acquisition and reproduction through point-to-point cooperation between a flexible array sensor and a finger mold component with multi-point reproduction function. With the acquisition terminal pre-locating the radial artery using a visualization method, the reproduction terminal simulates the pulse-finding technique of a doctor in an actual pulse diagnosis scenario by actively pressing, effectively avoiding misdiagnosis and thus achieving high-precision pulse position and pulse pattern acquisition and reproduction.

[0040] In terms of specific structure, unlike the pulse-finding process of existing acquisition terminals, this acquisition terminal uses a visual pre-positioning method to directly locate the acquisition finger on the radial artery. One side of the pulse-finding frame component is placed against the base of the palm as the origin point for pulse finding. Real-time visualization of the radial artery is used to locate it at the center of the pulse-finding point for secondary positioning. This avoids the problem of needing to remotely control the XYZ spatial coordinates of the acquisition finger in existing pulse-finding processes, effectively improving the accuracy of pulse finding. When manually finding the pulse, the reproduction terminal only needs to remotely control the pressure of the acquisition finger on the Cun, Guan, and Chi acupoints to complete the pulse finding. When automatically finding the pulse, a certain amplitude, such as 25g, is applied to the acquisition finger component corresponding to the Cun, Guan, and Chi acupoints until the peak amplitude of that point is obtained. Simultaneously, through multi-point high-precision reproduction simulation, the problem of existing acquisition terminals requiring multiple repetitions of positioning in the XZ direction and difficulty in accurate positioning in the Y direction for single-point reproduction is solved. This avoids misdiagnosis caused by inaccurate positioning during single-point acquisition.

[0041] Visualized pulse diagnosis utilizes the principle that hemoglobin absorbs infrared light more strongly than other tissues. A specific wavelength of near-infrared light is projected onto the skin surface via an infrared lamp module at the bottom of the wrist, and the infrared image of the skin is captured by the photosensitive components of an infrared camera. While infrared vascular imaging methods provide relatively clear images of veins, they are not as clear for arteries. This method only identifies the radial artery, located between the flexor carpi radialis and brachioradialis muscles, for pulse diagnosis. A rectangular physical device is used to shield most of the interference from veins. After processing the acquired raw infrared images, the outline of the radial artery is displayed clearly and in real-time on a display terminal. A pre-set positioning program provides a notification on the display terminal when the radial artery is located within the specific area of ​​the pulse diagnosis site.

[0042] The method utilizes a rectangular window on the wristband to combine with images for pulse acquisition, reducing the scope and difficulty of pulse acquisition; a circular positioning device is used to assist in image calculation and correct the pulse acquisition image, facilitating pulse acquisition from various angles.

[0043] The mechanical finger, employing a rack and pinion structure and upper and lower limit components, can move independently up and down. Furthermore, it utilizes pressure feedback from an array of pressure sensors to achieve the pulse-taking techniques of traditional Chinese medicine, focusing on the superficial, middle, and deep pulse points.

[0044] This document describes the pulse cycle recognition of a multi-point array pulse sensor; the differentiation between positive and negative waveforms of the pulse signal from the multi-point array pulse sensor; the baseline drift removal algorithm for positive and negative waveforms; the method of real-time pulse output using a sliding window mode; and the ability of the multi-point sensor to collect all key sensory information, including eight essential parameters such as pulse position, pulse force, tension, length, frequency, rhythm, thickness, and fluency, as well as unique characteristics.

[0045] The replication terminal employs a unique mapping between its multi-array points and multi-voice coil motor components. The gateway board receives amplitude-time waveform data from the acquisition terminal with low latency. This data is then processed by the host computer and sent to the driver boards of each voice coil motor component. Each driver board outputs an operating current to the coil, which in turn drives the cover to move the valve stem of a reciprocating hydraulic cylinder along a linear guide. This causes the hydraulic oil to rhythmically impact the bionic skin of the finger mold component. Ultimately, the fluctuations at the corresponding points of the acquisition terminal's array sensors are projected onto the bionic skin of the finger mold component in real time, in a unique mapping. Simultaneously, a pressure sensor at the bottom of the finger mold base collects the doctor's pressure data, thus adapting to differences in pulse diagnosis results caused by variations in pulse-taking force among different schools of traditional Chinese medicine, effectively improving compatibility.

[0046] Furthermore, to achieve the highest possible accuracy while maintaining a certain number of array points, the array points are arranged in an alternating pattern on the bottom of the bionic skin. Specifically, the array points of the second row are placed at intervals between the array points of the first row, the array points of the third row are placed at intervals between the array points of the second row (which may coincide with the horizontal coordinates of the first row), and so on. This invention mainly uses four sets of array points arranged in an alternating pattern as an example. Through the above arrangement, the coverage area is increased when using the same number of array points. Attached Figure Description

[0047] Figure 1 This is a schematic diagram illustrating the structure of the terminal in use as described in this invention.

[0048] Figure 2 This is a schematic diagram of the structure of the terminal for replicating the present invention (pipelines are not shown).

[0049] Figure 3 for Figure 2 Schematic diagram of the middle voice coil motor assembly (I).

[0050] Figure 4 for Figure 2 Schematic diagram of the middle voice coil motor assembly II.

[0051] Figure 4a for Figure 4 A partially enlarged schematic diagram of the quick-connect structure.

[0052] Figure 5 for Figure 2 A schematic diagram of the middle finger mold component.

[0053] Figure 6 for Figure 5 A perspective view of the middle finger mold base.

[0054] Figure 6a This is a perspective structural diagram of another type of fingerprint holder according to the present invention.

[0055] Figure 6b for Figure 6 A schematic diagram of the isometric side view structure of the bionic skin.

[0056] Figure 7 for Figure 6 A schematic diagram along direction A shows one of the array configurations of the present invention.

[0057] Figure 7a for Figure 7 A cross-sectional structural diagram.

[0058] Figure 7b for Figure 7 A schematic diagram of the pulse reproduction result of one of the fingerprint components.

[0059] Figure 7c This is a schematic diagram of the wave waveform collected by one of the acquisition components.

[0060] Figure 7d for Figure 7 A schematic diagram of the hydraulic principle of one type of fingerprint assembly.

[0061] Figure 8 This is a structural diagram of the terminal used to collect data according to the present invention.

[0062] Figure 9 This is a schematic diagram of the data acquisition terminal of the present invention.

[0063] Figure 10 This is a schematic diagram of the exploded structure of the data acquisition terminal of the present invention (the strap structure is different).

[0064] Figure 11 for Figure 10 A schematic diagram of the structure of the mechanical finger assembly.

[0065] Figure 12 To visualize the raw infrared image taken during pulse diagnosis.

[0066] Figure 12a This is the infrared image after rotation correction.

[0067] Figure 12b for Figure 12a Images after first and second edge sharpening following positioning and cropping.

[0068] Figure 12c for Figure 12b The image after binarization.

[0069] Figure 12d for Figure 12c Image after bone extraction.

[0070] Figure 12e for Figure 12d A diagram illustrating the longest path in the diagram.

[0071] Figure 12f This is a diagram showing the positions of the three pixels adjacent to the current pixel in a tiny rectangle.

[0072] Figure 12g This is a schematic diagram of the region where the radial artery is located.

[0073] Figure 12h for Figure 12g Restore to Figure 12 The result is illustrated in the diagram.

[0074] Figure 13 This is a schematic flowchart of the pulse-taking method of the present invention.

[0075] Figure 14 This is a pulse potential diagram of an array sensor.

[0076] Figure 15 This is a three-dimensional schematic diagram of the time-domain and frequency-domain analysis of the acquisition terminal of the present invention.

[0077] Figure 16 This is a schematic diagram of the overall architecture of the system of the present invention. Detailed Implementation

[0078] The following combination Figures 1-16 The specific content of the present invention will be described in detail through specific embodiments.

[0079] In the overall design of the reproduction terminal 1, a single pressure reproduction point cannot provide enough information to accurately reproduce the pressure range in the pulse. By combining multiple pressure reproduction points together and arranging them in an alternating manner, the sense of gaps between the pressure reproduction points can be eliminated, thereby obtaining pulse data with higher resolution and more accurate positioning, and avoiding misdiagnosis.

[0080] like Figures 1-2 As shown, the reproduction terminal 1 uses a drawer-type mounting bracket 15 as the overall structural support, combined with a streamlined outer shell as the main structure. Due to the use of a voice coil motor assembly 13 array, a drawer-type structure is adopted for ease of daily use, facilitating quick location of the corresponding voice coil motor assembly 13 during inspection and enabling rapid assembly and disassembly. To enable visual communication and consultation between doctors and patients, commonly used components in the field, such as a camera 111, touchscreen 112, microphone 113, power button 114, and speaker 115, are also installed on the panel assembly 11. To expand functionality, a data interface (e.g., a reproduction terminal USB interface, not shown in the figure) is provided on the back of the shell for system debugging or data import / export, and a power interface and main switch (not shown in the figure) are also provided. To ensure stable operation of the hardware, a heat dissipation vent (not shown in the figure) is also provided on the back.

[0081] To achieve overall machine control, the electrical control component 14 mainly adopts a gateway board 141 for remote communication, an AC-DC power adapter board 142 for power supply to the whole machine, and a host computer motherboard 143 for loading the main control system.

[0082] To facilitate high-precision pulse reproduction, a finger mold assembly 12 with multi-point control is positioned below the touchscreen 112. Several voice coil motor assemblies 13, matching the number of reproduction points, are mounted on a drawer-type mounting bracket 15. A hydraulic oil tank 16, serving as a fluid source, is also mounted on the drawer-type mounting bracket 15 to support the use of the voice coil motor assemblies 13. The hydraulic oil tank 16, the voice coil motor assemblies 13, and the finger mold assembly 12 are arranged sequentially from top to bottom. Preferably, the number of layers on the drawer-type mounting bracket 15 containing the voice coil motor assemblies 13 corresponds to the height of the connecting end 125 on the finger mold base 121. Figure 7d As shown, the hydraulic pipeline system mainly includes a hydraulic oil tank 16, several solenoid valve groups 161 (not shown in the figure), a reciprocating hydraulic cylinder 135, several single-point reproducing oil chambers 129, and corresponding connecting pipelines.

[0083] like Figure 6 , Figure 6a , Figure 6b , Figure 7 , Figure 7a As shown, the bionic skin 123 made of flexible material (e.g., oil-resistant rubber) is sealed and fixed (e.g., bonded) by the protruding connecting parts 1211 at the four corners of the top of the finger mold base 121 made of rigid material (e.g., aluminum alloy). During the fixing process, air is discharged through the exhaust port 1231, so that the bionic skin 123 is completely attached to the top of the finger mold base 121. The single-point reproduction oil cavity 129 in the bionic skin 123 corresponds one-to-one with the protrusion (not marked in the figure) at the end of the flow channel 127 at the top of the finger mold base 121. The single-point reproduction oil cavity 129 can just cover and seal the outlet end of each flow channel 127 to prevent hydraulic oil leakage.

[0084] Under the unified control of the host computer motherboard 143, the voice coil motor assembly 13 periodically performs a large stroke (not a pulse-reproducing stroke), causing the air in the pipeline to move to the solenoid valve assembly 161 at the highest point of the pipeline under the action of buoyancy. At this time, the corresponding solenoid valve assembly 161 is opened, returning the air to the hydraulic oil tank 16, while simultaneously replenishing the hydraulic oil to the corresponding pipeline, ultimately achieving periodic automatic oil replenishment and venting. The breather valve 163 is a one-way valve that realizes the exchange of gas and hydraulic oil, maintains the internal pressure balance of the hydraulic oil tank 16, avoids excessive air pressure in the hydraulic oil tank 16, and prevents external impurities from entering. When the liquid level sensor 162 detects that the liquid level is too low, it sends a signal to the host computer to manually replenish the hydraulic oil tank 16.

[0085] like Figures 3-4 , Figure 4aAs shown, the voice coil motor assembly 13 includes components commonly found in existing technologies, such as a base 139, a permanent magnet 133, a yoke (not shown), a coil 131, a cover 137, and a linear potentiometer 138. This invention, based on existing linearly operating voice coil motors, adds a reciprocating hydraulic cylinder 135 to one side of the base 139 and installs a three-way quick-connect fitting 136 for connecting to the hydraulic oil tank 16. Specifically, the cover 137 has a pair of spaced-apart C-shaped recesses on the side corresponding to the reciprocating hydraulic cylinder 135, serving as a quick-connect structure 1371. During installation, the rear of the main body of the reciprocating hydraulic cylinder 135 is first fixed to the base 139 with bolts. Then, the valve stem 1351 is inserted into the quick-connect structure 1371 without contacting its inner wall, thus limiting the valve plate 1352 within the quick-connect gap 1372 of the quick-connect structure 1371. Since the valve stem 1351 and the quick-release structure 1371 do not directly contact each other, space is reserved for installation or machining errors (form and position tolerance deviations caused by installation or manufacturing precision). This allows the valve stem 1351 to bear only axial force when it moves, thereby avoiding cylinder leakage and liquid leakage caused by the valve stem 1351 bearing additional radial force, thus improving the operational stability of the equipment.

[0086] This invention also improves the drive module of the voice coil motor assembly 13. The linear guide 134 ensures the dimensional and positional tolerances of each component; the drive board 132 reduces remote interference from high-frequency analog signals, and the CAN bus is individually calibrated and controlled for easy maintenance; the reciprocating hydraulic cylinder 135 achieves precise, high-frequency motion, converting linear motion into the pulsation of blood vessels, with multi-point combinations simulating a complete pulse pattern. Specifically, the drive board 132 integrates position control, current control, and PID control algorithms to achieve more precise motion control.

[0087] To improve accuracy, position control offers higher precision, enabling more accurate and repeatable pulse amplitude output positioning. Furthermore, current control provides higher control precision, eliminating the impact of sensor and system noise on control system performance; to improve response speed, it enables fast and smooth linear transitions. Additionally, position and current control can monitor mechanical load in real time while rapidly adjusting the current to maintain pulse curve overlap; to improve stability, linear motors, with their complete position and current control, better handle load variations and maintain stability, ensuring consistent motion pressure and amplitude under pulse-taking finger pressure; to reduce energy consumption, linear motors are generally more efficient due to their better handling of dynamic loads. Furthermore, position and current control can reduce motor power consumption during startup and steady-state operation, lowering motor operating temperature.

[0088] like Figure 5As shown, the finger mold assembly 12 consists of three strip-shaped structures located within the base 128, corresponding to the Cun, Guan, and Chi acupoints respectively. One perspective view of the finger mold assembly 12 is shown below. Figure 6 As shown, the finger mold assembly 12 includes a pressure sensor 124 disposed within a base 128, a strip-shaped finger mold base 121 located on the pressure sensor 124, several flow channels 127 communicating from the sidewalls of the finger mold base 121 to the top, a connecting end 125 located on the sidewall of the finger mold base 121 at the input end of the flow channel 127, an array of dots 126 located at the output end of the flow channel 127, and a bionic skin 123 (such as silicone) disposed on the top of the finger mold base 121 that mates with the output end of the flow channel 127. Each connecting end 125 is fixed with a quick-tightening elbow 122. The array of dots on the top of the finger mold base 121 corresponds one-to-one with the electrode points of the flexible array sensor 2142 of the acquisition terminal 2.

[0089] Meanwhile, to simulate wrist size as closely as possible, the size of the bionic skin 123 was limited; to avoid signal deviation when pressing down on the pressure sensor 124 due to excessive height of the finger mold base 121, the height of the finger mold base 121 was limited; to ensure that the flow channels 127 have consistent orientation within the finger mold base 121, consistent pipe diameters without interference, and consistent entry directions from the connection end 125, the installation direction of the quick-tightening elbow 122 was limited. Due to the minimum size limitation of the quick-tightening elbow 122, it was also necessary to ensure that it did not interfere during installation. If the quick-tightening elbow 122 were arranged vertically, to avoid interference, it would be impossible to install all quick-tightening elbows 122 simultaneously upwards or downwards, resulting in different impact forces on the bionic skin 123 under the same hydraulic conditions, leading to decreased reproduction accuracy. Therefore, all liquid inlet ends of the quick-tightening elbow 122 were arranged horizontally. To ensure consistent load on the pressure sensor 124, the weight of each finger mold base 121 must also be ensured. Under the premise that all the above conditions are met, it may be necessary to deform the structure of the fingerprint holder 121 to a certain extent. For example Figure 6a Another fingerprint holder structure is shown, in which the height of the fingerprint holder 121 is reduced and the mounting position of the connecting end 125 is protruded outward (i.e. widened), thereby avoiding interference between the quick-turn bends 122 of the middle and side fingerprint assemblies 12. In order to ensure that the weight of the middle fingerprint holder 121 is comparable to that of the side fingerprint holders 121, the height of the middle fingerprint holder 121 is reduced.

[0090] Through the above-mentioned modified design, an alternating wide and narrow installation method can be achieved, allowing for high-density installation of the quick-connect elbows 122 within a limited space, significantly reducing product volume and optimizing pipeline layout. By calculating volume and adjusting the center of gravity through drilling, the weight of the two replica finger mold bases 121 is made uniform, avoiding any impact on the pressure sensor 124 at the bottom of the finger mold and ensuring accurate feedback of the floating, sinking, and buoyancy forces.

[0091] The array points 126 of the reproduction terminal 1, the voice coil motor assembly 13, and the flexible array sensor 2142 of the acquisition terminal 2 are in a one-to-one correspondence to complete the pulse acquisition and reproduction process. Specifically, each array point 126 is driven by a completely independent voice coil motor assembly 13, and each voice coil motor assembly 13 is connected to a relatively independent hydraulic oil tank 16. Figure 2 The diagram shown is merely a schematic representation of the structure of terminal 1 and does not represent the actual number of components. When one of the acquisition components 214, the flexible array sensor 2142, acquires data as shown... Figure 7c After the waveform signal is shown, it is transmitted to the gateway board 141 via the wireless communication module (not shown in the figure) of the terminal device (such as a smartphone). After data decoding, it is sent by the host computer motherboard 143 to the drive board 132 of the corresponding voice coil motor assembly 13. The drive board 132 then outputs a drive current to the coil 131. Under the action of the drive current, the coil 131 drives the cover 137 to move along the linear guide rail 134, which in turn drives the valve stem 1351 of the reciprocating hydraulic cylinder 135 to move. When the valve stem 1351 extends or retracts, it pumps hydraulic oil into the finger mold assembly 12 with different forces, thereby generating undulations on the corresponding array points 126 of the bionic skin 123. After the other array points 126 move synchronously, they produce the following effect: Figure 7b The amplitude fluctuations are shown. Figure 7b The diagram shows an ideal application of dense array points 126. In actual applications, due to cost and overall weight considerations, when the number of array points 126 is small, it still conforms to the rule that the array points 126 in a certain row fluctuate within the range of zero to the maximum amplitude, while the array points 126 in other rows fluctuate within the range of zero to less than the maximum amplitude, and the maximum amplitude decreases from the middle row to the upper and lower sides.

[0092] like Figure 8 This is one embodiment of the present invention. The acquisition terminal 2 includes a pulse acquisition unit 21 and a wearable unit 22 that are slidably engaged. The wearable unit 22 is modified from the existing adjustable wristband 223 by adopting a double-sided tensioning structure. The pulse-taking frame assembly 225 and the infrared lamp module 224 are respectively arranged opposite to each other at the upper and lower ends of the adjustable wristband 223. The infrared lamp module 224 is housed within a slidable lamp housing 221. The pulse-taking frame assembly 225 has a pulse-taking port 2251 on the radial artery side. The pulse-taking frame assembly 225 slides with the acquisition seat 212 of the pulse acquisition unit 21 via a pair of linear slides 2252 arranged along the radial artery direction. The side of the pulse-taking frame assembly 225 near the palm also has an arc-shaped indentation 2254 for accommodating the base of the palm.

[0093] like Figures 9-11In another embodiment of the invention shown, the wearable part 22 employs a single-sided tensioning structure, with the acquisition seat 212 and the pulse-taking frame assembly 225 engaged on the self-locking protrusion 2253 via a self-locking member 216. The self-locking member 216 can employ a structure found in the prior art, such as a commercially available spring-loaded buckle. Through the interaction of the spring and the elastic buckle (not shown), the self-locking member 216 switches between locked and unlocked operating states. This allows the pulse acquisition part 21 to be quickly engaged with or removed from the pulse-taking frame assembly 225.

[0094] The pulse acquisition unit 21 includes an acquisition base 212 as the main support structure, a protective shell 211, a main control board 219 as the main control unit, a (lithium) battery 215, three sets of acquisition finger assemblies 214 arranged side by side for acquiring pulse signals, an infrared camera 218 for cooperating with the infrared lamp module 224, a self-locking member 216 for cooperating with the self-locking protrusion 2253, a rack 217 for cooperating with the gear 2141 of the acquisition finger assembly 214, and a pair of limit modules 213 with paddles 2131 for limiting the upper and lower travel ends of the acquisition finger assembly 214. The main control board 219 is equipped with a main control button 2191, a light guide post 2192, and a USB interface (not marked in the figure) for charging or data transmission.

[0095] Multiple light guide columns 2192 serve as status indicator lights, or multi-colored single indicator lights can be used to display information such as current pressure status, data transmission status, and battery level. For example, a high-frequency flashing indicator light indicates that data is being transmitted; a low-frequency flashing indicator light indicates an error such as abnormal data acquisition, network connection failure, or Bluetooth connection failure; a solid red indicator light indicates low battery; and a solid green indicator light indicates good status. These status indicators are also reflected in the mobile app.

[0096] The acquisition finger assembly 214 includes a motor assembly 2145, a gear 2141 disposed at its output end, a motor bracket 2144 for cooperating with adjacent acquisition finger assemblies 214, a flexible array sensor 2142 disposed at the bottom, and a silicone finger mold 2143. The silicone finger mold 2143 covers the flexible array sensor 2142.

[0097] In use, the radial artery is first located using a visual positioning method, and then the pulse acquisition unit 21 is fixed to the wearable unit 22. (See the usage method below for detailed steps.) The acquisition base 212 has a rack 217 that corresponds to the number of acquisition finger components 214. When the motor assembly 2145 drives the gear 2141 to rotate, the acquisition finger components 214 can move vertically along the rack 217, thereby placing the silicone finger mold 2143 at the bottom against the patient's wrist. The pulse signal of different intensities is fed back to the flexible array sensor 2142 according to the pressure applied. When the doctor of the reproduction terminal 1 presses down on the bionic skin 123, the pressure sensor 124 located at the bottom of the finger mold base 121 receives the pressure signal and sends it to the main control board 219 through the gateway board 141. The main control board 219 drives the motor assembly 2145, thereby causing the acquisition finger components 214 to produce an action corresponding to the force applied by the reproduction terminal 1, in order to simulate the face-to-face pulse diagnosis scenario with high precision.

[0098] This pulse-taking technique is mainly used for automatic pulse taking by the finger assembly 214. It involves progressively increasing pressure at the cun, guan, and chi pulse positions until the maximum amplitude range obtainable within the array sensor is reached. For example... Figure 13 As shown, the specific method of taking a pulse includes the following steps:

[0099] Step S1, Locate and cover the radial artery: Locate the radial artery pulsation range by vision or finger touch, place the acquisition finger assembly 214 of the three parts of cun, guan, and chi above the radial artery, so that each flexible array sensor 2142 covers the pulsation range of the radial artery (this embodiment takes 4*6 staggered array sensors as an example).

[0100] Step S2, Obtain the optimal pulse-taking pressure:

[0101] Step S21, Generate pulse potential diagram: The amplitude signal detected by the flexible array sensor 2142 is combined with time domain analysis and frequency domain analysis to generate a two-dimensional (time-amplitude) pulse potential diagram. Figure 14 ) and 3D vector graphics (time-amplitude-frequency) Figure 15 The flexible array sensor 2142 measures the pulse pattern changes of the three positions of Cun, Guan, and Chi at a pressure of 25~250g; the pulse is taken at the three positions of Cun, Guan, and Chi simultaneously according to the gradually increasing pressure value, so as to imitate the pulse taking method of clinical doctors.

[0102] Step S22, determine the optimal pulse-taking pressure: the pulse-taking pressure value with the highest amplitude and no baseline drift from five consecutive pulse waveforms at each of the cun, guan, and chi positions; specifically: gradually increase the pressure value in a multi-cycle equal increment until the maximum pulse amplitude range is obtained in a certain cycle; the pressure increase is 15~35g, the cycle duration is 3~8s, and the measurement cycle is two or more.

[0103] In step S3, each flexible array sensor 2142 takes pulses synchronously or asynchronously at its optimal pulse taking pressure value. The pulse taking time is 20~40s. The optimal pulse taking pattern is based on the highest amplitude of five consecutive pulse patterns with no baseline drift and no deviation.

[0104] Simultaneous pulse taking simulates the overall superficial, middle and deep pulse taking of the three pulse positions (cun, guan, chi) and the total pressure of the three positions under the optimal pulse taking pressure value;

[0105] Asynchronous pulse taking simulates the overall superficial, middle and deep pulse taking of the three pulse positions (cun, guan, chi) and single pulse taking of the three positions under the optimal pulse taking pressure value;

[0106] Step S4, Generate the optimal pulse pattern: Record the pulse amplitude trend changes of the overall optimal pulse pattern in step S2 under different pressures to obtain a three-dimensional pulse pattern, and record the optimal pulse pattern of the cun, guan, and chi positions in step S3.

[0107] The change curves of pulse amplitude at different pressure values ​​were recorded to determine the pulse position, pulse strength, and pulse momentum among the pulse elements. Three optimal pulse momentum diagrams were recorded respectively, and the pulse length, pulse width, fluency, and tension were analyzed from two-dimensional and three-dimensional perspectives.

[0108] In summary, this embodiment not only includes simultaneous pulse taking at the cun, guan, and chi positions to identify the pulse location, strength, and momentum, but also extracts the optimal pulse momentum diagram under the best pulse taking pressure at each position, thereby analyzing and measuring other factors of the pulse. Its collection method is closer to clinical practice, and the operation time is shorter, which is tolerable for patients. It is simple, portable, and suitable for clinical use.

[0109] Step S1: Connect the pulse acquisition unit 21 of the acquisition terminal 2 to a display terminal, such as a smartphone, tablet, or display module integrated on the pulse acquisition unit 21, via Bluetooth. For cost reduction, the pulse acquisition unit 21 in the embodiments shown in the accompanying drawings does not integrate a display module. Wear the wearable unit 22 on one wrist, place the pulse retrieval port 2251 near the radial artery, position the cun, guan, and chi pulses in the pulse retrieval port 2251 with the palm facing upwards, place the infrared lamp module 224 at the bottom of the wrist, and turn on the infrared lamp module 224 using a mobile app.

[0110] Step S2: Hold the adjustment acquisition seat 22 with the other hand, irradiate the pulse acquisition port 2251 with the infrared camera 218 at its bottom, and observe the imaging position of the radial artery in real time through the display terminal;

[0111] Step S3: Based on the image on the display terminal, rotate the arc-shaped concave 2254 of the pulse-taking frame assembly 225 along the base of the palm while close to the palm to adjust the position of the pulse-taking port 2251 so that the radial artery is located in the center of the pulse-taking port 2251. The mobile phone or the acquisition base 22 will notify the user via voice, indicator light or message that the pulse-taking port 2251 is in the correct position. The pulse acquisition unit 21 is locked onto the pulse-taking frame assembly 225 through the self-locking cooperation of the self-locking protrusion 2253 and the self-locking member 216, or the pulse acquisition unit 21 is slidably locked onto the pulse-taking frame assembly 225 through the linear slide 2252, and pulse acquisition can then begin.

[0112] The method for detecting the radial artery in step S2 includes the following steps:

[0113] Step P1, Acquire raw image: Illuminate the wrist skin located within the pulse point 2251 using infrared camera 218 to acquire raw infrared image in real time. Figure 12 (The annotations in the image are for identification and explanation only; the corresponding annotations did not exist when the image was captured.)

[0114] Step P2, image correction:

[0115] Step P21, locating the radial and ulnar sides: The ulnar side of the pulse-taking frame assembly 225 is pre-set with specific positioning points (two circular points in this embodiment) to correct the orientation of the original image, improving positioning accuracy and speed. Other shapes can also be used, or the number of positioning points can be increased or decreased, as long as it achieves rapid positioning of the pulse-taking port 2251. Specific positioning points are located using edge detection algorithms such as the Canny algorithm and the Hough transform algorithm.

[0116] Step P22, locate the pulse point 2251: continue to use the edge detection algorithm to identify the rectangular region;

[0117] Step P23, Rotation Correction: Based on the rectangular position and the calibrated horizontal line obtained from the specific positioning points, determine the rotation angle of the image, perform rotation correction, and then crop the rectangular area obtained in step P22. For example, set conditions in the program to rotate the image until the center line connecting the specific positioning points is rotated to the horizontal, and calculate the position of the pulse orifice 2251 based on the length of the line connecting the positioning points;

[0118] Step P3, clarifies the boundary of the rectangular region ROI: Two edge sharpening processes are performed using a Laplace filter, yielding the following results: Figure 12b (superior), Figure 12b (Below), the Laplace operator (Equation 1) uses a four-neighborhood convolution kernel (Equation 2);

[0119] Equation (1) Equation (2)

[0120] Step P4, Mark the arteries:

[0121] Step P41, Binarization: For Figure 12b (Below) After binarization, we obtain Figure 12c ;

[0122] Step P42, Skeleton Extraction: Using Zhang-Suen's Algorithm... Figure 12c Extracting bones Figure 12d This yields multiple connected regions A~D;

[0123] Step P43, locate the longest path: Figure 12d The longest path L(i) in each connected region is found using the depth-first search (DFS) algorithm.

[0124] Step P44, Locate the radial artery: Compare Li with La based on the preset length threshold La < 2.73 cm. Figure 12d Pruning is performed on regions where the longest path is less than a threshold. Figure 12d (in the marked area), determine Figure 12d Region A in the diagram represents the radial artery;

[0125] Normally, the radial artery is 2.73cm to 4.10cm long. Although the length of the radial artery varies from person to person, for safety, a length threshold of La=20mm is used to determine whether it is a radial artery.

[0126] Step P5, Mark the radial artery: Figure 12d The path is generated by creating multiple tiny rectangles, with n rectangles in total. Each tiny rectangle is formed by taking the three adjacent pixels of the current pixel. Figure 12f As shown. Place each rectangle in Figure 12d The input above is given to the Grabcut algorithm to obtain the selected region S(i). This will result in S(1)∪S(2)∪...∪S(n)=S. After marking S, we get... Figure 12g ;

[0127] Step P6, Image Restoration: Figure 12g Restore to Figure 12 The data is processed and output to the display terminal.

[0128] like Figure 16As shown, the pulse acquisition and reproduction system includes a reproduction terminal 1, an acquisition terminal 2, and a server terminal. The acquisition terminal 2 and the reproduction terminal 1 can establish a data transmission channel directly through a C / S architecture. The acquisition terminal 2 (patient end) acts as the server and sends pulse data to the reproduction terminal 1 (client) as the client in real time. The reproduction terminal 1 reproduces the pulse data by using several reproduction points that can convert waveform signals into fluctuations, and then corresponds one-to-one with the flexible array sensor 2142 of the acquisition terminal 2, thus completing the pulse reproduction.

[0129] To establish a dataset for the intelligent pulse imaging system, data samples of different pulse characteristics at different ages need to be collected. At this point, a data transmission channel is established using a server terminal as an intermediary, and repetitive pulse data of specific durations is extracted and uploaded to the server in real time. The pulse signals collected by acquisition terminal 2 are acquired by a flexible array sensor 2142. The data acquired at each acquisition point is two-dimensional data composed of time and pulse amplitude, or three-dimensional data composed of time, pulse amplitude, and acquisition frequency. Acquisition terminal 2 sends the pulse data to a smart device via Bluetooth, which then transmits it to the server terminal via a smartphone or tablet. The pulse data and / or diagnostic results from the server terminal are then downloaded to reproduction terminal 1. Figure 15 As shown, on the intelligent device of the reproduction terminal 1, the two-dimensional data composed of time and pulse amplitude, or the three-dimensional data composed of time, pulse amplitude and acquisition frequency, is restored into a more intuitive graphic result, thereby realizing the real-time signal acquisition of pulse or directly obtaining pulse diagnosis results without the need for a doctor's diagnosis.

[0130] Specifically, the server terminal includes:

[0131] The communication module is used to receive pulse data uploaded by the acquisition terminal 2 and / or send diagnostic results to the reproduction terminal 1;

[0132] The database module is used to store two-dimensional or three-dimensional pulse data uploaded by the acquisition terminal 2;

[0133] The deep learning model module is used to load the pulse diagnosis model and return the diagnosis results from the pulse data uploaded by the acquisition terminal 2.

[0134] The pulse data annotation module is used for supervised training of the pulse diagnosis model;

[0135] The model training module is used to process the labeled data into a pulse diagnosis model.

[0136] Figure reference numerals: 1 Reproduction terminal, 11 Panel assembly, 111 Camera, 112 Touchscreen, 113 Microphone, 114 Power button, 115 Speaker, 12 Fingerprint assembly, 121 Fingerprint base, 1211 Connector, 122 Quick-connect elbow, 123 Bionic skin, 1231 Exhaust port, 124 Pressure sensor, 125 Connector end, 126 Array point, 127 Flow channel, 128 Base, 129 Single-point reproduction oil chamber, 13 Voice coil motor assembly, 131 Coil, 132 Driver board, 133 Permanent magnet, 134 Linear guide rail, 135 Reciprocating hydraulic cylinder, 1351 Valve stem, 1352 Valve plate, 136 Three-way quick-connect fitting, 137 Cover, 1371 Quick-release structure, 1372 Quick-release gap, 138 Linear potentiometer, 139 base, 14 electrical control components, 141 gateway board, 142 AC-DC power adapter board, 143 host computer motherboard, 15 drawer-type mounting bracket, 16 hydraulic oil tank, 161 solenoid valve assembly, 162 level sensor, 163 breather valve;

[0137] 2 Acquisition terminal, 21 Pulse acquisition unit, 211 Protective shell, 212 Acquisition base, 213 Limiting module, 2131 Paddle, 214 Acquisition finger assembly, 2141 Gear, 2142 Flexible array sensor, 2143 Silicone finger mold, 2144 Motor bracket, 2145 Motor assembly, 215 Battery, 216 Self-locking component, 217 Rack, 218 Infrared camera, 219 Main control board, 2191 Main control button, 2192 Light guide column, 22 Wearable part, 221 Sliding light box, 223 Adjustable wristband, 224 Infrared light module, 225 Pulse retrieval frame assembly, 2251 Pulse retrieval port, 2252 Linear slide, 2253 Self-locking protrusion, 2254 Arc-shaped concave.

Claims

1. A visual pulse diagnosis method, characterized in that, Includes the following steps: Step S1: Connect the pulse acquisition unit (21) of the acquisition terminal (2) to the display terminal via wireless transmission, wear the wearable unit (22) on one wrist, place the pulse taking point (2251) close to the radial artery, place the Cun, Guan, and Chi pulse positions in the pulse taking point (2251), with the palm facing up, place the infrared lamp module (224) at the bottom of the wrist, and turn on the infrared lamp module (224). Step S2: Hold the adjustment acquisition seat (212) with the other hand, irradiate the pulse taking port (2251) with the infrared camera (218) at its bottom, and observe the imaging position of the radial artery in real time through the display terminal; Step P1: Acquire raw infrared images in real time; Step P2, cut out the rectangular area where the radial artery is located: Step P21, Locating the radial and ulnar sides: The ulnar side of the pulse taking frame component (225) is pre-set with specific positioning points for correcting the orientation of the original image. The specific positioning points are located by the Canny algorithm and the Hough transform algorithm. Step P22: After identification using the edge detection algorithm, the rectangular region of the pulse taking frame component (225) is obtained; Step P23: Obtain the calibrated horizontal line based on the location of the rectangular region and the specific positioning point. After rotation correction, crop the image based on the location of the specific positioning point to obtain the rectangular region image where the radial artery is located. Step P3: Apply a Laplacian filter to the rectangular region to sharpen the edges. The Laplacian operator uses a four-neighbor convolution kernel (Equation 2). Equation (2); Step P4: Binarize the edge-processed image and locate the radial artery by extracting bones; Step P41: Binarize the image after edge sharpening; Step P42: Use Zhang-Suen's Algorithm to extract the skeleton from the binarized image and obtain multiple connected regions; Step P43: Use the Depth-First Search (DFS) algorithm on each connected region to find the longest path Li in each connected region; Step P44: Set the pruning threshold La to 10mm~30mm, compare Li and La of each connected region, exclude the region where the longest path is less than the pruning threshold, and locate the longest path in the remaining path as the radial artery. Step P5: Mark the outer contour of the radial artery using the Grabcut algorithm: The radial artery path obtained in step P4 is used to generate multiple tiny rectangles, with n rectangles in total. Each tiny rectangle is formed by taking the three adjacent pixels of the current pixel. Each rectangle is input into the Grabcut algorithm to obtain the selected region S(i), resulting in S(1)∪S(2)∪...∪S(n)=S. The radial artery vessel edge is then marked according to the region S. Step P6: Cover the rectangular area where the pulse point (2251) of the original image is located with the image after marking the outer contour, and output the integrated image to the display terminal. Step S3: Adjust the position of the pulse taking point (2251) so that the radial artery is located in the center of the pulse taking point (2251), and attach the pulse image collection part (21) to the pulse taking frame assembly (225) to start pulse image collection.

Images