An intrusive tactile sensor and application method

Abstract

The present invention provides an interventional tactile sensor and an application method, comprising a measuring head, a catheter, a main unit and a displacement unit, wherein the catheter connects the measuring head and the main unit, and based on the total internal reflection principle of a waveguide, utilizes the law that the optical power transmitted inside the waveguide changes with the size of the external contact load during the "organic" contact process to solve the core problems required in interventional tactile diagnosis, such as electrical passivity, miniaturization, large sensing angle, and adjustable resolution, and can maintain measurement consistency for different azimuths; the application method trains a deep learning model algorithm based on a palpation data set of a medical database, and the rich data set enables the algorithm to achieve accurate and efficient disease diagnosis and staging prediction, so as to cope with complex and diverse soft tissue environments in the body, reduce missed diagnosis and misdiagnosis of endoscopes, thereby achieving "early detection and early treatment" of diseases and improving the cure rate of diseases.

Description

Technical Field

[0001] The present invention relates to the technical field of endoscopes, and in particular to an interventional tactile sensor and an application method thereof. Background Art

[0002] According to data released by the World Health Organization's International Agency for Research on Cancer (IARC) in 2021, colorectal cancer has become one of the cancers with the most new cases in the world, and its mortality rate ranks second only to lung cancer. Colorectal cancer and other digestive tract cancers are mucosal lesions, and the cure rate is extremely high if discovered early, so early screening is of great significance.

[0003] Traditional cancer screening methods are mainly divided into three types: imaging screening, endoscopic screening and tumor marker screening. Currently, the main screening method for digestive system diseases is endoscopic examination. A digestive endoscope is an optical instrument that is inserted into the body through the natural cavity of the human body from outside the body. Physicians can use it to observe the location and range of lesions in the internal cavity of organs, and can use interventional surgical instruments such as forceps and scissors to operate on the endoscope or remove tissue for living tissue examination (referred to as biopsy). It is the "gold standard" for screening and diagnosis of digestive tract lesions.

[0004] However, due to the complex internal environment of the human body and the small and flat nature of early cancer or polyps, it is difficult for doctors to identify and judge in real time, resulting in a high rate of clinical misdiagnosis. For a long time, since endoscopes are only used for "looking" but not "measuring", they can only diagnose diseases qualitatively, but cannot determine the stage of the disease by quantitative means.

[0005] Endoscopic biopsy can only examine a small amount of pathological tissue that has been removed, but cannot fully examine the entire pathological tissue. Biopsy obtains test samples by damaging living tissue. During the test, it may cause pain at the incision site and symptoms such as incision bleeding. Some patients may also have infection problems. In addition, because medical devices that are repeatedly used in minimally invasive or interventional surgeries often have complex internal structures and are difficult to thoroughly clean and disinfect, they are prone to residual microorganisms, secretions, and blood, causing cross-infection problems between different patients.

[0006] Existing studies have shown that lesions in human tissue are often closely related to changes in the hardness of its surface. For example, the hardness of common lesions such as lymphoma, breast cancer and liver cancer is significantly greater than that of normal tissue. Therefore, the development of tactile sensors for measuring soft tissue contact mechanics will help to better understand the pathology and pathogenesis of various diseases, provide new methods for early detection of diseases, and provide important basis for the fight against and treatment of diseases.

[0007] In order to realize the most important mechanical sensitive unit in the tactile sensor, the current sensors use the principles of resistance, piezoresistive, piezoelectric, capacitive, electromagnetic, etc. to convert mechanical information into output electrical signals. However, the sensors with different working principles and materials mentioned above have certain defects in sensitivity, working environment or processing circuits. For example, there are often disturbances between different sensitive nodes of capacitive tactile sensors, which require special reading methods and circuits; electromagnetic tactile sensors have high requirements for the detection environment, which reduces the ease of integration and limits their application occasions.

[0008] In contrast, optical tactile sensors use light signals to measure mechanical information, and have advantages in electrical passivity, anti-electromagnetic interference, chemical inertness, light weight, and small size. Among them, fiber optic tactile sensors have been put into practical use in minimally invasive or interventional surgeries due to their advantages of easy miniaturization, high resolution, and strong environmental adaptability. However, since their force-sensitive directions are generally concentrated in the axial direction, affected by the physical properties and layout of the optical fiber, it is difficult to continuously increase the sensing angle and make the angular resolution isotropic. Therefore, in the field of interventional tactile sensors, there is still a need to further develop disposable medical sensors that can assist endoscopes in accurately and quickly quantifying diseases, and have the advantages of good biocompatibility and electrical passivity, large sensing angles, and isotropic high resolution. Summary of the invention

[0009] The technical problem to be solved by the present invention is to provide an intrusive tactile sensor.

[0010] Another technical problem to be solved by the present invention is to provide an application method of the above-mentioned intrusive tactile sensor.

[0011] In order to solve the above technical problems, the technical solution of the present invention is:

[0012] An invasive tactile sensor comprises a probe head, a catheter, a host and a displacement part, wherein the catheter connects the probe head and the host,

[0013] The measuring head comprises a GRIN lens (self-focusing lens), a filter element, a waveguide substrate, a waveguide core layer and an elastic element with a hemispherical top. The front end of the GRIN lens is fixedly connected to the filter element and the elastic element. The outer side of the filter element is provided with a waveguide substrate and a waveguide core layer in sequence. The waveguide core layer is embedded in the elastic element. The inner surface of the elastic element is provided with a plurality of elastic contacts. An air gap (about several microns) is left between the tip of the elastic contact and the waveguide core layer. The plurality of elastic contacts have the same or different heights and different bottom diameters.

[0014] The conduit part includes a core, a cladding, a coating (protective) layer and a fiber interface of the optical fiber. One end of the core is connected to the GRIN lens, and the other end is connected to the fiber interface. The outside of the core is provided with a cladding, and the outside of the cladding and more than half of the side area of ​​the GRIN lens are covered with a coating layer. The fiber interface is threadedly connected to the host part.

[0015] The host part includes an optical fiber circulator, a laser light source, a photodetector and a microprocessor (MCU) inside, wherein the microprocessor (MCU) and the optical fiber circulator are respectively connected to the laser light source and the photodetector circuit, and the packaging shell of the host part is connected to the optical fiber interface of the catheter part;

[0016] The displacement part is located below the main machine part, and a mechanical transmission device is arranged inside the shell of the displacement part, and a displacement switch is arranged outside the shell, wherein the mechanical transmission device includes an active part, a driving pawl, a ratchet, a swinging mechanism, a slider, a driving part, and a first spring, a second spring and a third spring fixed on the shell with a reset function, the active part is in contact with the displacement switch, and the displacement switch pushes the active part to rotate clockwise when it rotates counterclockwise, the active part is fixedly connected to the first spring, one end of the driving pawl is fixedly connected to the active part, and the other end is in contact with the ratchet and fixedly connected to the third spring at an adjacent position, the middle of the swinging mechanism uses a bracket fixed on the shell as a fulcrum, one end is in contact with the ratchet and fixedly connected to the second spring at an adjacent position, and the other end is connected to the slider through a slide groove, the slider is fixedly connected to the driving part, the driving part is fixedly connected to the main machine part, and the main machine part is slidably connected to the displacement part through a sliding guide rail.

[0017] The above-mentioned interventional tactile sensor has a guide block of a sliding guide rail on the packaging shell of the host part, and a guide groove of the sliding guide rail on the shell of the displacement part. The host part is slidably connected with the guide groove of the sliding guide rail on the displacement part through the guide block of the sliding guide rail, so that the driving member can drive the host part to perform reciprocating linear motion along the sliding guide rail, and then drive the catheter part connected to the host part to also perform reciprocating linear motion.

[0018] In the above-mentioned invasive tactile sensor, the waveguide substrate, waveguide core layer and the air layer (waveguide covering layer) between the waveguide core layer and the elastic contact of the measuring head constitute the basic structure of the waveguide. The GRIN lens is used as a relay lens, which can expand the output light of the optical fiber (including the core and cladding) into parallel light of a larger area, so that part of the light becomes the input light of the waveguide, and can also converge the output light of the waveguide so that it all enters the optical fiber. The near-infrared light with a wavelength of 850nm generated by the laser light source is transmitted into the measuring head through the optical fiber circulator, and enters the waveguide after being collimated by the GRIN lens; when there is no contact load, the elastic body does not contact the waveguide, and the light that meets the waveguide conditions will be stably transmitted in the waveguide in the form of total internal reflection; the output light of the waveguide enters the optical fiber through the GRIN lens to form a "ring light circuit"; finally, the output light of the optical fiber is transmitted to the photodetector through the optical fiber circulator. The above is the direction of the light path when there is no contact load.

[0019] Preferably, in the above-mentioned interventional tactile sensor, the elastomeric element is prepared by the following method: a mold of the elastomeric element is made by a 3D printing process, the mold comprising two parts, mold A and mold B, a micro-nano structure is printed on the inner surface of mold B, and then a silicone rubber prepolymer solution is injected into the mold and vacuum degassed, and the elastomeric element is separated from the mold after curing to obtain a elastomeric element having an elastic contact array of different sizes on its inner surface and a micro-nano structure surface with super-hydrophobic properties on its outer surface.

[0020] Preferably, in the above-mentioned intrusive tactile sensor, the elastomeric element is made of silicone rubber, and its elastic modulus changes with the ratio of the main agent and the curing agent, thereby achieving flexible adjustment of the sensor range and resolution.

[0021] Preferably, in the above-mentioned intrusive tactile sensor, the elastic contacts are conical in shape, and the elastic contacts are annularly dispersed inside the elastomeric element, and the bottom diameter of the elastic contacts dispersed in the outer circle is larger than the bottom diameter of the elastic contacts dispersed in the inner circle.

[0022] When there is a contact load, the elastic contact on the inner surface of the elastomeric element contacts the core layer of the waveguide, causing the latter's waveguide conditions to be destroyed and generating radiated light, and the larger the contact area between the two, the more leakage light is generated, thereby establishing a "force-light" conversion relationship. The light near the center of the hemispherical shell-shaped waveguide is dense and the light at the edge is sparse, which will cause the light intensity in the waveguide to show a distribution pattern that decreases from the center to the edge according to a certain rule. Therefore, the present invention optimizes the arrangement of the cone-shaped elastic contact array on the inner surface of the elastomeric element, setting them to have the same height and a distribution pattern in which the bottom diameter increases step by step from the center to the edge, so as to compensate for the uneven distribution of light intensity in the waveguide, thereby achieving isotropic measurement resolution for contact loads of the same size within the 180° sensing range of the measuring head.

[0023] Preferably, in the above-mentioned invasive tactile sensor, the measuring head is prepared by the following method: first, a filter element with high absorption characteristics for near-infrared light with a wavelength of 850nm is used as a substrate; then a waveguide substrate and a waveguide core layer are sequentially deposited on the upper surface of the filter element by a vacuum evaporation process; the filter element and the elastomer element are sequentially bonded to the GRIN lens by medical UV glue to complete the fixation of the components inside the measuring head.

[0024] Preferably, in the above-mentioned invasive tactile sensor, the main body is externally packaged with medical metal materials, and a digital display screen is also provided on the surface thereof, which is connected to the microprocessor circuit.

[0025] Preferably, in the above-mentioned intrusive tactile sensor, the host unit further includes a Bluetooth transmission module and a wireless charging module, and the Bluetooth transmission module and the wireless charging module are respectively connected to the microprocessor circuit.

[0026] Preferably, in the above-mentioned invasive tactile sensor, the laser light source is near-infrared light with a wavelength of 850nm, which is transmitted through optical fiber, collimated by a GRIN lens, enters the waveguide core layer from one side of the waveguide, and then exits from the other side of the waveguide. After being converged by the GRIN lens, it enters the optical fiber core for transmission and is finally received by the photodetector.

[0027] The application method of the above-mentioned invasive tactile sensor is to upload the palpation data obtained by the measuring head to the computer through the Bluetooth module of the host part, and draw the "force-displacement" curve by using the interpolation method; by preprocessing each set of palpation data, the force value at the "initial contact state" of the palpation is normalized to zero; then the deep learning algorithm is used to diagnose the disease: first, the fully connected neural network (FCNN) is used to extract the primary features of the "force-displacement" curve; then, the long short-term memory network (LSTM) is used to extract the spatiotemporal features of the diagnostic data in the medical records of the same patient at different times in the medical database; finally, the FCNN feature classification network is used to predict the output vector of the above-mentioned feature extraction network in two forms: the type and stage of the disease, and the prediction results are stored in the medical database.

[0028] Preferably, the application method of the above-mentioned interventional tactile sensor uses NoSOL database technology to connect data with the existing medical database, stores the palpation data obtained by the measuring head and the disease results diagnosed by medical experts through endoscope visual diagnosis in the medical database, and uses the attention mechanism to dynamically adjust the classification weights of the palpation data distribution attribute information and the medical expert knowledge and experience information, reclassifies the "abnormal samples" into different disease period categories, completes the establishment of the palpation data set, and uses it as the training set of the deep learning algorithm.

[0029] During use, the measuring head of the tactile sensor enters the human body through the "clamp channel" of the endoscope, uses the "bend" of the endoscope to control the rotation of the "tip" and uses the visual information of the objective lens to find the measurement object, which is used to solve the problem that the endoscope can only observe the surface erosion area of ​​the diseased tissue but cannot obtain the mechanical information of the tissue surface, and realize the combination of "observation" and "measurement". Based on the medical database, the measurement deviation of the tactile sensor and the mechanical differences of the soft tissues in different patients are corrected through the model algorithm, so as to accurately and quickly diagnose the disease stage quantitatively.

[0030] Beneficial effects:

[0031] The above-mentioned interventional tactile sensor is an interventional tactile sensor with large sensing angle and isotropic resolution. It is based on the total internal reflection principle of the waveguide and utilizes the law that the optical power transmitted inside the waveguide changes with the size of the external contact load during the "biological-mechanical" contact process to solve the core problems required in interventional tactile diagnosis, such as electrical passivity, miniaturization, large sensing angle, and adjustable resolution. It can maintain measurement consistency for different azimuths; in addition, by developing replaceable disposable palpation instruments, it can be used to eliminate the risk of cross-infection between different patients and prevent the occurrence of accidental medical accidents such as patients being infected with viral hepatitis or AIDS due to incomplete disinfection. Specifically:

[0032] 1. Elastic contacts with the same height but different bottom diameters are set on the inner surface of the elastomer element, which solves the problem of uneven light intensity distribution at the center and edge of the hemispherical shell waveguide, so that the measuring head can achieve isotropic resolution measurement effect within the 180° sensing range.

[0033] 2. A GRIN lens is used as a relay lens to collimate the light in the optical fiber and input it into the waveguide, and the output light in the waveguide is converged into the optical fiber, which solves the complex coupling problem between the hemispherical shell waveguide and the optical fiber. Medical UV glue is used to bond the elastomer and the filter element to the GRIN lens respectively, which avoids the complex process of optical-mechanical structure assembly and debugging, and enables the measuring head to be miniaturized for use in interventional palpation.

[0034] 3. The mold of the elastomeric element is manufactured using the 3D printing process, and the micro-nano structure is printed on the part that contacts the outer surface of the elastomeric element. The elastomeric micro-nano structure surface with super-hydrophobic properties can be obtained in batches to prevent body fluids on the surface of the soft tissue from adhering to the surface of the measuring head and affecting the repeatability of the measurement.

[0035] 4. The catheter part is threadedly connected to the main unit through an optical fiber interface, and is used together with the measuring head as a medical consumable. After a patient uses it, the doctor can replace it using the optical fiber interface, thereby preventing cross infection between patients due to incomplete disinfection.

[0036] 5. By taking advantage of the effect of the ratio of silicone rubber main agent and curing agent on the elastic modulus of the material, elastomers with different hardness were produced. Under the premise that the basic structure of the measuring head remains unchanged, the range and resolution of the tactile sensor can be adjusted to adapt to different types of soft tissue detection.

[0037] 6. The tactile sensor enters the body through the forceps channel of the endoscope for in-situ palpation. Without changing the existing endoscope structure, it solves the problem that the endoscope can only observe tissue lesions, provides doctors with important biomechanical diagnostic standards, and enables the endoscope to combine "observation" and "measurement".

[0038] 7. Use the existing medical database to establish a palpation data set to provide a rich training set for the deep learning model algorithm, and use the attention mechanism to reasonably allocate the classification weights of the palpation data distribution attribute information and the medical expert knowledge and experience information, thereby increasing the reliability of the training set.

[0039] 8. The deep learning model algorithm is trained based on the palpation data set of the medical database. The rich data set enables the algorithm to achieve accurate and efficient disease diagnosis and staging prediction to cope with the complex and diverse soft tissue environment in the body, reduce endoscopic missed diagnosis and misdiagnosis, and thus achieve "early detection and early treatment" of the disease and improve the cure rate of the disease; in addition, the palpation data is "standardized" through the algorithm to solve the error caused by the uncertainty of the "initial state" of palpation; each palpation result is stored in the medical database, which not only increases the material for pathological research, but also enriches the training data set of the deep learning model, which has a positive feedback effect on the accuracy of disease diagnosis and staging prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a schematic diagram of the structure of a tactile sensor;

[0041] Figure 2 It is a schematic diagram of the connection between the catheter part, the displacement part and the host part;

[0042] Figure 3 This is a schematic diagram of the measuring head structure;

[0043] Figure 4 This is a schematic diagram of the working principle of the tactile sensor;

[0044] Figure 5 This is the workflow diagram of intelligent palpation;

[0045] Figure 6 This is a schematic diagram of the "force-light" conversion principle for measuring the head;

[0046] Figure 7 Schematic diagram of mechanical measurement for measuring the isotropic resolution of the head;

[0047] Figure 8 To measure the "integrated" process roadmap of the head;

[0048] Fig. 9 This is a schematic diagram of how to use the tactile sensor;

[0049] Fig.10 is a schematic diagram of the mechanical structure of the displacement part;

[0050] Fig.11 Schematic diagram of the measurement process of the tactile sensor;

[0051] Fig.12 A schematic diagram of a multi-position measurement method for measuring the head;

[0052] Fig.13 This is the algorithm flow chart of intelligent palpation.

[0053] In the figure: 1-probe head; 2-catheter; 3-host; 4-displacement; 5-optical fiber interface; 6-displacement switch; 7-GRIN lens; 8-filter element; 9-waveguide substrate; 10-waveguide core layer; 11-elastic element; 12-elastic contact on the inner surface of the elastomer; 13-optical fiber core; 14-optical fiber cladding; 15-coating layer; 16-optical fiber circulator; 17-laser light source with a wavelength of 850nm; 18-photodetector; 19-endoscope; 20-endoscope Angle control knob of the bending part of the endoscope; 21-entrance of the clamp channel of the endoscope; 22-insertion tube of the endoscope; 23-bending part of the endoscope; 24-exit of the clamp channel of the endoscope; 25-objective lens of the endoscope; 26-light guide mirror of the endoscope; 27-front end of the endoscope; 28-active part; 29-driving pawl; 30-ratchet; 31-swinging mechanism; 32-slider; 33-driving part; 34-first spring; 35-second spring; 36-third spring; 37-sliding guide rail. DETAILED DESCRIPTION

[0054] The intrusive tactile sensor and application method of the present invention are described below with reference to embodiments and drawings.

[0055] Example 1

[0056] 1. Design and manufacturing method of measuring head

[0057] like Figure 1 As shown, an interventional tactile sensor comprises a probe head 1, a catheter 2, a host 3 and a displacement 4. Since the clamp channel sizes of different models of flexible endoscopes are different, such as Fig. 9 As shown, this embodiment takes a gastrointestinal endoscope with a clamp channel diameter of 3.0 mm as an example, and designs the catheter portion of the tactile sensor to have a diameter of 2.5 mm to match the clamp channel size.

[0058] like Figure 3 As shown, the appearance of the measuring head 1 is a spherical top cylinder, which can be divided into two parts according to its use. One part is a mechanical sensitive element, that is, an elastic element 11 with a hemispherical top; the other part is a "force-light" conversion element, including a GRIN lens 7, a filter element 8, a waveguide substrate 9 and a waveguide core layer 10. The filter element 8 has good light absorption performance for light with a wavelength of 850nm. The core function of the measuring head 1 is to perform "force-light" conversion, wherein the principle of "force-light" conversion is as follows:

[0059] The optical waveguide consists of two outer covering layers and a substrate of low refractive index media and an inner core layer of relatively high refractive index media, and its refractive index distribution is n1>n2≥n3, where n1, n2 and n3 are the refractive indices of the core layer, substrate and air respectively.

[0060] The stable transmission of light in a waveguide requires two basic conditions to be met, namely

[0061] n2k0<β<n1k0 (I)

[0062] κd=mπ+δ 12 +δ 13 (II)

[0063] Wherein, formula (I) is the waveguide condition, in which β is the propagation constant, k0 is the wave number of the light wave in vacuum, and only the light wave whose propagation constant β satisfies formula (I) can propagate in the waveguide; formula (II) is the dispersion equation, when the light wave is stably transmitted in the waveguide, its optical path difference must satisfy the coherence enhancement condition, in which is the x component of the core wave vector, d is the core thickness of the waveguide, δ 12 and δ 13 are the total reflection phase shifts of light at the interfaces between the core and substrate and between the core and the cladding respectively. m is called the mode number, and its value is a non-negative integer of 0, 1, 2, etc., where each number represents a waveguide mode.

[0064] like Figure 6 As shown in (a), when the elastomer is not in contact with the waveguide, that is, when no pressure event occurs, the light satisfying equations (I) and (II) propagates in the waveguide and forms a guided mode through coherent enhancement; while those light waves with a propagation constant of 0≤β≤n2k0 will be refracted outward to form a radiation wave.

[0065] like Figure 6 As shown in (b) in the figure, when the elastic body contacts the waveguide, that is, when a pressure event occurs, since the refractive index of the elastic body is greater than the refractive index of the waveguide substrate, that is, n4>n2, ​​the waveguide condition of the light at the contact position between the two will no longer be formula (II), but

[0066] n4k0<β<n1k0 (III)

[0067] Therefore, those light rays with propagation constants of n2k0≤β≤n4k0 will escape from the waveguide and radiate outward. If the contact area between the elastic body contact and the waveguide continues to increase under the action of external load, the waveguide will radiate more energy outward, causing the optical power transmitted inside the waveguide to continuously decrease, thus establishing a functional relationship between the optical power change ΔP transmitted inside the waveguide and the contact load ΔF, completing the conversion from force to light.

[0068] According to the “force-light” conversion principle, the functional relationship between the variation of the optical power transmitted in the waveguide ΔP and the contact load ΔF is established with the contact area between the elastic contact and the waveguide as the medium. Figure 6As shown in (d), it is a schematic diagram of the light intensity distribution inside the hemispherical shell waveguide. Since the light inside the hemispherical shell waveguide is input and output through the edge, the light intensity in the waveguide shows a distribution trend that gradually increases from the edge to the center. This means that even if the contact area is the same, if the elastomeric contact and the hemispherical shell waveguide contact positions are different, the waveguide may also generate radiation energy of different magnitudes: when on the same cross-section of the hemispherical shell waveguide, the radiation light energy that the elastomeric contact contacts with it is the same, otherwise the light energy radiated by the two contacts is different, that is, the probe exhibits anisotropic measurement results for different sensing angles.

[0069] In order to solve this problem, this embodiment designs an array of elastic contacts 12 with the same height and a bottom diameter that decreases from the edge to the center. Figure 6 As shown in (b) and (c) in the figure, when the same pressure event occurs, the elastic body contact in Figure (c) has a larger bottom diameter than that in Figure (b), so that the waveguide generates more light energy loss under the same waveguide conditions (Formula (III)). Therefore, the present invention compensates for the problem caused by the uneven light intensity distribution inside the hemispherical shell waveguide by changing the bottom diameter of the contact, so that the probe exhibits isotropic mechanical resolution within a large sensing angle range of 180°, as shown in FIG. Figure 7 shown.

[0070] like Figure 6 As shown in (a) and (b) in FIG. 1 , since the maximum contact area S between each elastic contact and the hemispherical shell waveguide is max is a fixed value. According to the principle of "force-light" conversion, the maximum light loss in the waveguide caused by each elastic contact is also a constant value. Therefore, the range of the tactile sensor of the present invention can be defined as: the contact area between the elastic contact and the waveguide is S max The pressure at that time is recorded as F max .

[0071] Since the mechanical properties of soft tissue surfaces in different parts are different, max When F is constant, for those biological tissues with harder surfaces, it is necessary to select an elastic body with a larger elastic modulus, that is, to increase F max For biological tissues with softer surfaces, when diagnosing diseases, it is necessary to select an elastomer with a smaller elastic modulus to increase the resolution of the sensor ΔF. min , thereby capturing its mechanical information more accurately.

[0072] In this embodiment, polydimethylsiloxane (PDMS) is selected as the material of the elastomer element 11. PDMS is a silicone rubber material commonly used in the biomedical field and has good biocompatibility. When the ratio of the main agent and the curing agent is different, PDMS will show different elastic moduli. When they are in contact with the same load, they will cause different deformations, so that the contact area between the elastic contact 12 and the waveguide core layer 10 is different and different degrees of light loss are generated, thereby obtaining a measuring head 1 with different mechanical ranges and resolutions. Commonly used ratios are: 5:1, 10:1, 15:1 and 20:1, and their elastic moduli decrease in turn. Therefore, according to the different properties of the measured object, PDMS materials with different ratios can be used to make the elastomer element 11 to obtain different ranges and resolutions.

[0073] Since human body fluids have a certain viscosity, a small amount of liquid may adhere to the outer surface of the probe during contact with it, resulting in a large error in the next measurement result. This embodiment overcomes this problem by improving the manufacturing process of the elastic element 11. Figure 8 As shown, first, a mold of the elastomer element 11 is made by a 3D printing process, the mold comprising two parts, mold A and mold B, and a micro-nano structure is printed on the inner surface of mold B, and then a silicone rubber prepolymer solution is injected into the mold and vacuum degassed, and the elastomer element 11 is obtained by separating from the mold after curing. There are arrays of elastic contacts 12 of different sizes on the inner surface of the elastomer element 11, and the outer surface is a micro-nano structure surface with super-hydrophobic properties, thereby overcoming the problem that the probe is easily adhered to by body fluids to ensure the repeatability of the measurement. In addition, in order to prevent the radiation light of the waveguide from interfering with the internal transmission light, this embodiment adds 1 to 2% dark red silicone paste material to the PDMS material, so that it has good light absorption performance at a wavelength of 850nm.

[0074] This embodiment refers to the material setting of SOI type waveguide, whose core layer, cover layer and substrate are silicon (Si), air and silicon dioxide (SiO2) in sequence. At a wavelength of 850nm, the refractive index of Si is 3.64 and the refractive index of SiO2 is 1.45.

[0075] Since the PDMS material with a main agent and a curing agent ratio of 15:1 is used to make the elastomer element (11) in this embodiment, its refractive index is 1.40, which is smaller than the refractive index of SiO2, a SOI type waveguide substrate, in order to satisfy the waveguide conditions of formula (I) and formula (II), n4>n3 (such as Figure 6(b)), it is necessary to replace the substrate material of the SOI waveguide and select a waveguide medium with a lower refractive index, MgF2, whose refractive index is 1.38, which can meet the design requirements. Therefore, the materials of the waveguide in this embodiment are: core layer (Si, n1 = 3.64), substrate (MgF2, n2 = 1.38), and cover layer is air (n3 = 1).

[0076] When the light in the waveguide is totally reflected, it is not absolutely reflected back to the core layer at the interface, but it must penetrate a certain depth into the substrate or cover layer and propagate a certain distance along the interface before emitting along the direction of the reflected light. This depth is called the penetration depth, and the light wave propagating along the interface is called the "evanescent wave". In addition to participating in the propagation of the guided mode field, the evanescent wave also carries a certain amount of energy, which makes the effective thickness d of the guided mode field e The geometric thickness d of the core layer increases, that is,

[0077]

[0078] Where 1 / p and 1 / q are the penetration depths of light in the substrate and cover layer, respectively; p and q are the attenuation constants in the substrate and cover layer along the x direction, respectively, that is,

[0079]

[0080]

[0081] Wherein, k0 = 2π / λ. Taking into account the application requirements and processing technology of the waveguide, the thickness of the waveguide substrate is finally designed to be 20 μm and the thickness of the cover layer is 200 μm. So far, this embodiment has completed the design of the characteristic parameters d, n1, n2, n3 and the transmission wavelength λ of the waveguide.

[0082] like Figure 4 As shown, the GRIN lens 7 is used as a relay lens, which is a cylindrical optical lens. The difference from the ordinary lens is that the distribution of the refractive index of the self-focusing lens material gradually decreases along the radial direction, which can cause the light transmitted along the axial direction to produce continuous refraction, thereby achieving smooth and continuous convergence of the output light. This embodiment uses a GRIN lens with a pitch of 0.25P, which can collimate the output light of the optical fiber and expand it to a larger area, so that part of the light becomes the waveguide input light; it can also converge the output light of the waveguide so that it all enters the optical fiber, thereby forming a "ring light loop". The use of the GRIN lens, combined with the unique design structure of the measuring head, solves the complex coupling problem between the hemispherical shell waveguide and the optical fiber.

[0083] like Figure 3As shown, the radial dimension of the measuring head of this embodiment is 2 mm, that is, the outer surface radius of the elastomer element and the bottom diameter of the GRIN lens are both 2 mm. The manufacturing method of the measuring head 2 is as follows: First, a filter element 8 is used as a substrate. The filter element 8 has a high absorption characteristic for near-infrared light with a wavelength of 850 nm, thereby avoiding interference from stray light such as radiation light of the waveguide; then, a vacuum evaporation process is used to deposit a waveguide substrate 9 and a waveguide core layer 10 on the upper surface of the element 8 in sequence; the filter element 8 and the elastomer 11 are bonded to the GRIN lens in sequence using medical UV glue, and the fixing of the various components inside the measuring head can be completed without a mechanical sleeve, avoiding the complicated process of optical-mechanical structure assembly and debugging, such as Figure 8 shown.

[0084] 2. Design of the catheter, main unit and displacement unit

[0085] like Figure 2 As shown, the catheter part 2 connects the measuring head part 1 and the main body part 3, and is the medium for the optical signal interaction between the two. The optical fiber is composed of a core 13 and a cladding 14. The core 13 is provided with a cladding 14 outside. The coating layer 15 of the catheter part 2 is made of medical rubber material, which covers more than half of the side area of ​​the optical fiber and the GRIN lens 7 of the measuring head part 1, and is fixed to the measuring head part 1 by medical UV glue. The coating layer 15 not only protects the internal optical fiber, but also strengthens the connection between the measuring head part 1 and the catheter part 2. In this embodiment, the optical fiber in the catheter part 2 uses a single-mode optical fiber. The other end of the catheter part 2 has an optical fiber interface 5, which is threadedly connected to the packaging shell of the main body part 3 to facilitate the flexible replacement of the measuring head part 1 and the catheter part 2, and avoid cross infection between different patients. After the molds of the measuring head part 1 and the catheter part 2 are made, they can be mass-produced with low mass production costs. They can be used as disposable medical consumables and replaced with the main body part 3 through the optical fiber interface 5, thereby preventing cross infection problems between patients caused by incomplete disinfection.

[0086] like Figure 4As shown, the host part 3 mainly includes a microprocessor (MCU), a laser light source 17, a photodetector 18, a Bluetooth transmission module, a wireless charging module and other circuit modules (such as a displacement module), and the microprocessor (MCU) is respectively connected to the laser light source 17, the photodetector 18, the Bluetooth transmission module, the wireless charging module and other circuit modules (such as a displacement module). Among them, the laser light source adopts a laser diode with a wavelength of 850nm, and the photodetector adopts a photodiode. The Bluetooth transmission module transmits mechanical data to the computer in real time; the wireless charging module can eliminate the potential risk of electric shock to the doctor due to the presence of liquid on his hands during wired charging. The outside of the host part 3 is packaged with medical metal materials, and a digital display screen is provided on the surface, which is connected to the microprocessor (MCU) circuit for real-time display of the mechanical data obtained by measurement; the packaging shell of the host part 3 also has an optical fiber interface 5 connected to the catheter part 2 and a guide block of the sliding guide rail 37.

[0087] like Figure 2 As shown, below the host part 3 is the displacement part 4, the housing of the displacement part 4 is provided with a mechanical transmission device, and the housing is provided with a displacement switch 6 and a guide groove of a sliding guide rail 37 for controlling the mechanical transmission device, and the displacement switch 6 only has a triggering effect on the mechanical transmission device, and the host part 3 is slidably connected with the guide groove of the sliding guide rail 37 on the displacement part 4 through the guide block of the sliding guide rail 37. When the displacement switch 6 is pressed by the physician, it will swing backwards, and the mechanical transmission device inside the displacement part 4 is triggered; when the physician stops pressing the displacement switch 6, it will immediately return to its original position; during the return period of the displacement switch 6, it does not interfere with the movement of the mechanical transmission device.

[0088] like Fig.11 As shown, after the displacement switch 6 is activated, the displacement transmission device will push the main body part 3 together with the catheter part 2 and the measuring head part 1 to move forward a constant displacement L. Fig.10As shown, the interior of the mechanical transmission device includes an active member 28, a driving pawl 29, a ratchet 30, a swing mechanism 31, a slider 32, a driving member 33, and a first spring 34, a second spring 35 and a third spring 36 fixed to the housing with a reset function. The active member 28 is in contact with the displacement switch 6. The displacement switch 6 pushes the active member 28 to rotate clockwise when rotating counterclockwise. The active member 28 is fixedly connected to the first spring 34. One end of the driving pawl 29 is fixedly connected to the active member 28, and the other end is in contact with the ratchet 30 and is adjacent to the active member 28. The third spring 36 is fixedly connected, and the middle of the swing mechanism 31 uses a bracket fixed on the shell as a fulcrum, one end is in contact with the ratchet 30 and is fixedly connected to the second spring 35 at the adjacent position, and the other end is connected to the slider 32 through a slide groove, and the slider is fixedly connected to the driving member 33, and the driving member is fixedly connected to the main unit 3. Since the main unit is slidably connected to the displacement unit 4 through the sliding guide rail 37, the driving member 33 can drive the main unit 3 to make reciprocating linear motion along the sliding guide rail 37, thereby driving the catheter unit 2 connected to the main unit 3 to also make reciprocating linear motion.

[0089] Its specific movement principle is: the active member 28 rotates clockwise under the push of the displacement switch 6, and drives the driving pawl 29 thereon to push the ratchet teeth of the ratchet 30 to rotate around the bracket at a certain angle. The fixed fulcrum of the first spring 34 is located in front of the active member 28. After the active member 28 rotates counterclockwise, the doctor stops pressing the displacement switch 6, and the active member 28 is pulled by the first spring 34 to return to the initial position and complete the reset; in the above process, the end of the swing mechanism 31 close to the ratchet 30 plays the role of a "brake pawl", which makes the ratchet 30 only rotate in one direction clockwise; in the above ratchet 30, the ratchet 30 rotates clockwise in one direction. During the clockwise rotation, the swing mechanism 31 will be pushed by the ratchet teeth of the ratchet wheel 30 to rotate counterclockwise. At this time, a horizontal force component is generated to push the slider forward, so that the slider 32 and the driving member 33 are loaded forward with a constant displacement L; the second spring 35 is arranged in front of the swing mechanism 31, and then due to the pull of the second spring 35, the swing mechanism 31 rotates clockwise, and its end close to the second spring 35 contacts the ratchet teeth of the ratchet wheel 30 again to complete the reset. In this process, the slider 32 and the driving member 33 return to the initial position. The above is the movement principle of the internal mechanical transmission device during a displacement loading.

[0090] like Figure 4As shown in the figure, when there is no contact load, the optical path inside the tactile sensor is as follows: the laser diode generates near-infrared light with a wavelength of 850nm, which is transmitted into the optical fiber through the optical fiber circulator 16. The optical fiber circulator is a multi-port non-reciprocal optical device that can realize bidirectional optical signal transmission on a single optical fiber. The light in the single-mode optical fiber is collimated by the GRIN lens 7 and then transmitted to the waveguide; at this time, the light that meets the waveguide conditions will be stably transmitted inside the waveguide in the form of total internal reflection; the output light of the waveguide is converged into the single-mode optical fiber through the GRIN lens 7; finally, the output light of the single-mode optical fiber is transmitted to the photodetector 18 through the optical fiber circulator 16, completing the signal conversion process from "force-light-electricity".

[0091] like Figure 5 As shown, when there is a contact load, that is, the probe head 1 contacts the soft tissue, the elastic contact 12 in the probe head 1 will contact the waveguide. According to the "force-light" conversion principle, the waveguide will produce light loss, that is, the output light power of the waveguide will decrease, which will eventually manifest as a weakening of the output electrical signal of the photodetector, thereby resolving the measured force value.

[0092] 3. How to perform palpation

[0093] During use, if Fig. 9 As shown, doctor A holds the endoscope 19 and inserts the insertion tube 22 into the patient's body, and uses the objective lens 25 to find the soft tissue to be examined under the illumination of the light guide mirror 26; after locking the target soft tissue, doctor B holds the displacement part 4 with one hand and holds the catheter part 2 with the other hand to introduce the measuring head 1 from the clamp channel entrance 21 of the endoscope and lead it out from the clamp channel exit 24, and determines the "initial contact state" between the measuring head 1 and the target soft tissue through the reading on the digital display screen of the sensor, that is, the position of the measuring head 1 when the digital display screen just displays the reading; at this time, doctor B presses the displacement switch 6 to start the mechanical transmission device inside the displacement part 4, and after the displacement switch 6 is started, the mechanical transmission device will push the main unit together with the catheter part 2 and the measuring head 1 to move forward a constant displacement L, and then immediately return to the "initial contact state", that is, complete a palpation process, as shown in FIG. Fig.11 After a palpation is completed, the doctor B holds the catheter part 2 and guides the measuring head 1 to the side of the forceps channel entrance 21, and stops when the measuring head 1 cannot be observed in the objective lens 25. Then the doctor A continues to control the insertion tube 22 to move forward / backward and find the next soft tissue to be examined through the objective lens 25. Alternatively, the doctor A can also control the bending part 23 of the endoscope to drive the tip part 27 to rotate "up", "down", "left" or "right" by manipulating the angle control knob 20 of the bending part of the endoscope to find the soft tissue to be examined. Fig.12 When doctor A obtains the location information of the target soft tissue through the objective lens, doctor B can repeat the above palpation process to perform palpation. The overall process of intelligent palpation is as follows: Figure 5 shown.

[0094] 4. Intelligent palpation algorithm

[0095] The intelligent palpation algorithm described in this embodiment is implemented in the following four steps:

[0096] Step 1: Use NoSQL database technology to connect to the current medical database

[0097] During clinical trials of palpation, medical experts can use the palpation method to perform in situ palpation in patients to assist endoscopic vision in the diagnosis and staging of diseases. Figure 5 As shown, the data results of the palpation will be uploaded to the computer through the Bluetooth module of the host unit 3 for data processing. The palpation data and the staging results and treatment recommendations of the doctor's diagnosis are stored in a medical database. The medical database is an existing database, such as a hospital information system (HIS), a clinical data warehouse (CDR), an electronic medical record system (EMR), and a laboratory information management system (LIS), etc., which facilitates the storage and access of palpation data and its connection with existing medical data.

[0098] This embodiment uses NoSQL database technology to connect with the existing database. It is a non-relational database technology with good flexibility and scalability. By using NoSQL database technology, this embodiment can efficiently store and manage the unstructured data in the medical database, providing rich medical samples for in vivo palpation and pathology research.

[0099] Step 2: Establishment of palpation data set and reclassification of abnormal samples

[0100] For patients diagnosed as being in the middle or late stages of a disease by endoscopic vision, their corresponding palpation data are directly stored in the medical database; for patients diagnosed as being disease-free, having a benign disease, or having an early stage of a disease by endoscopic vision, since the possibility of misdiagnosis of the diagnosis is relatively high, after storing their corresponding palpation data in the medical database, medical experts also need to use the medical database to continuously track the subsequent development of the patient's disease. Medical experts need to observe whether patients diagnosed as being disease-free or having a benign disease subsequently develop disease or worsen their condition, and whether patients diagnosed as being in the early stages of a disease subsequently show no disease, benign disease, or worsen their condition. If the patient subsequently develops the above two situations, then there is a misdiagnosis problem in the diagnosis result previously obtained by the medical expert using endoscopic vision, and the corresponding palpation data of the patient is defined as an "abnormal sample".

[0101] This embodiment adopts a feature fusion network based on the attribute distribution information of palpation data and the knowledge and experience information of medical experts based on the attention mechanism. First, the K-means clustering method is used to re-cluster the "abnormal samples" in the palpation data set according to the disease type and stage using the attribute distribution of the diagnostic data. Then, the feature fusion network of the above-mentioned attention mechanism is used to dynamically adjust the weights of the K-means classification results and the medical expert's diagnosis results, so as to perform a more accurate secondary classification of the "abnormal samples" to reduce the impact of the subjectivity of the medical experts' medical experience and the limitations of the K-means clustering method on the diagnosis results. Finally, NoSQL database technology is used to update and expand the data with the existing medical database to further improve the accuracy and reliability of the palpation data set, such as Fig.13 shown.

[0102] Step 3: Preprocessing of palpation data

[0103] First, a palpation data set is extracted from a medical database. The palpation data set is a two-dimensional data set. The data types are all fixed-length serialized data types. The independent variable is the contact displacement between the measuring head 1 and the soft tissue. The function value is the force value output by the tactile sensor. The "force-displacement" curve of palpation can be obtained by interpolation. Then, due to the uncertainty of the "initial contact state" of the palpation between the measuring head 1 and the soft tissue at each measurement, the force at the origin of the displacement in the palpation data is not zero and is slightly larger than zero. In order to solve this problem, before processing the data using the deep learning model algorithm, this embodiment needs to pre-process the palpation data: subtract the force value at the zero displacement value from all the mechanical values ​​in each group of palpation data, so that the force values ​​of all groups of palpation data in the data set start from zero.

[0104] Step 4: Feature extraction and disease prediction based on palpation data

[0105] like Fig.13 As shown, this embodiment uses a fully connected neural network (FCNN) as a feature extraction and disease staging network. First, the palpation data set is input into the neural network FCNN to extract the change trend characteristics of the "force-displacement curve".

[0106] This embodiment uses a long short-term memory network (LSTM) to extract the spatiotemporal features of diagnostic data in the medical records of the same patient at different times in a medical database. First, the LSTM network has memory. It encodes the time series of the patient's previous diagnostic data and maps the diagnostic information of each time step into a vector representation of a fixed size. Then, these vector representations are arranged in sequence in the time series to form a spatiotemporal feature representation of the diagnostic data, which can capture the changing process of the patient's disease condition. The use of the LSTM network is conducive to medical experts to grasp the changing trend and evolution process of patient diagnostic data in the time dimension, so as to give more scientific diagnosis results and treatment opinion reports, and reduce the generation of "abnormal samples".

[0107] This embodiment uses the FCNN neural network to classify the output vector of the feature extraction network. The FCNN feature classification neural network predicts the disease in two forms: the type of disease and the stage.

[0108] In the actual application process, each set of data obtained from clinical palpation is stored in the medical database together with the diagnosis results after disease diagnosis and staging prediction by the above-mentioned deep learning algorithm, which will further enrich the size of the palpation data set; and the size of the palpation data set in the medical database will determine the prediction accuracy of the deep learning algorithm in disease diagnosis and staging. The two complement each other and can continuously improve the accuracy of the prediction results of the deep learning algorithm to cope with complex and diverse clinical cases, thereby better assisting the endoscope to perform accurate and efficient disease diagnosis and staging prediction, achieve "early detection and early treatment" of the disease, and improve the clinical cure rate.

[0109] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An intrusive tactile sensor, characterized in that: It comprises a measuring head (1), a catheter part (2), a main body part (3) and a displacement part (4), wherein the catheter part (2) connects the measuring head (1) and the main body part (3), wherein: The measuring head (1) comprises a GRIN lens (7), a filter element (8), a waveguide substrate (9), a waveguide core layer (10) and an elastomeric element (11) at the top, the front end of the GRIN lens (7) is fixedly connected to the filter element (8) and the elastomeric element (11), the outer side of the filter element (8) is provided with a waveguide substrate (9) and a waveguide core layer (10) in sequence, the waveguide core layer (10) is embedded in the elastomeric element (11), the inner surface of the elastomeric element (11) is provided with a plurality of elastic contacts (12), an air gap is left between the tip of the elastic contact (12) and the waveguide core layer (10), the bottom diameters of the plurality of elastic contacts (12) are different, and the arrangement mode of the elastic contacts (12) in the elastomeric element (11) is that the heights are the same and the bottom diameters increase step by step from the center to the edge; The catheter part (2) has one end connected to the GRIN lens (7) and the other end connected to the host part (3); The host part (3) internally comprises an optical fiber circulator (16), a laser light source (17), a photodetector (18) and a microprocessor, wherein the microprocessor and the optical fiber circulator (16) are respectively connected to the laser light source (17) and the photodetector (18) by lines, and externally has an optical fiber interface (5) connected to the catheter part (2); The displacement part (4) is provided with a mechanical transmission device inside and a displacement switch (6) outside, wherein the mechanical transmission device comprises an active member (28), a driving pawl (29), a ratchet (30), a swing mechanism (31), a slider (32), a driving member (33), and a first spring (34), a second spring (35) and a third spring (36) fixed on the housing and having a reset function, wherein the active member (28) is in contact with and connected to the displacement switch (6), the active member (28) is fixedly connected to the first spring (34), and the driving pawl (29) is in contact with and connected to the first spring (34). ) one end is fixedly connected to the active member (28), the other end is in contact with the ratchet (30) and is fixedly connected to the third spring (36) at an adjacent position, the middle of the swing mechanism (31) uses a bracket fixed to the shell as a fulcrum, one end is in contact with the ratchet (30) and is fixedly connected to the second spring (35) at an adjacent position, and the other end is connected to the slider (32) via a sliding groove, the slider is fixedly connected to the driving member (33), and the driving member is fixedly connected to the main unit (3); the main unit (3) is slidably connected to the displacement unit (4) via a sliding guide rail (37).

2. The intrusive tactile sensor according to claim 1, characterized in that: The elastomeric element (11) is prepared by the following method: a mold of the elastomeric element (11) is made by a 3D printing process, the mold comprising two parts, mold A and mold B, a micro-nano structure is printed on the inner surface of mold B, a silicone rubber prepolymer solution is injected into the mold and vacuum degassed, and the mold is separated after curing to obtain an elastomeric element (11), the inner surface of which has an array of elastic contacts (12) of different sizes, and the outer surface is a micro-nano structure surface with super-hydrophobic properties.

3. The intrusive tactile sensor according to claim 1, characterized in that: The elastic contacts (12) are conical in shape, and each elastic contact (12) is annularly dispersed inside the elastic element (11), and the bottom diameter of the elastic contacts (12) dispersed in the outer ring is larger than the bottom diameter of the elastic contacts (12) dispersed in the inner ring.

4. The intrusive tactile sensor according to claim 1, characterized in that: The measuring head (1) is prepared by the following method: first, a filter element (8) having a high absorption characteristic for near-infrared light with a wavelength of 850 nm is used as a substrate; then, a waveguide substrate (9) and a waveguide core layer (10) are sequentially deposited on the upper surface of the filter element (8) by a vacuum evaporation process; and the filter element (8) and the elastomer element (11) are sequentially bonded to the GRIN lens (7) by medical UV adhesive, thereby completing the fixation of the components inside the measuring head (1).

5. The intrusive tactile sensor according to claim 1, characterized in that: The catheter part (2) comprises an optical fiber core (13), a cladding (14), a coating layer (15) and an optical fiber interface (5); one end of the optical fiber core (13) is connected to the GRIN lens (7), and the other end is connected to the optical fiber interface (5); the outer part of the optical fiber core (13) is provided with a cladding (14); the outer part of the cladding (14) and more than half of the side area of ​​the GRIN lens (7) are covered with the coating layer (15); the optical fiber interface (5) is threadedly connected to the host part (3); the outer part of the host part (3) is encapsulated with medical metal materials, and a digital display screen is also provided on the surface, which is connected to the microprocessor circuit; the inner part of the host part (3) also includes a Bluetooth transmission module and a wireless charging module, and the Bluetooth transmission module and the wireless charging module are respectively connected to the microprocessor circuit.

6. The intrusive tactile sensor according to claim 1, characterized in that: The laser light source (17) is near-infrared light with a wavelength of 850 nm, which is transmitted through an optical fiber, collimated by a GRIN lens (7), enters the waveguide core layer (10) from one side of the waveguide, and then exits from the other side of the waveguide. After being converged by the GRIN lens (7), the light enters the optical fiber core (13) for transmission, and is finally received by a photodetector (18).

7. The intrusive tactile sensor according to claim 1, characterized in that: The elastic element (11) is made of silicone rubber, and its elastic modulus changes with the ratio of the main agent and the curing agent.

8. The method for applying the intrusive tactile sensor according to claim 1, characterized in that: The palpation data obtained by the measuring head (1) is uploaded to a computer via a Bluetooth module of a host unit (3), and a "force-displacement" curve is drawn using an interpolation method; the force value at the "initial contact state" of the palpation is normalized to zero by preprocessing each set of palpation data; and a deep learning algorithm is then used to perform disease diagnosis.

9. The method for applying the invasive tactile sensor according to claim 8, characterized in that: The deep learning algorithm is as follows: first, a fully connected neural network is used to extract primary features of the "force-displacement" curve; then, a long short-term memory network is used to extract the spatiotemporal features of the diagnostic data in the medical records of the same patient at different times in the medical database; finally, an FCNN feature classification network is used to predict the output vector of the above feature extraction network in two forms: the type and stage of the disease, and the prediction results are stored in the medical database.

10. The application method of the intrusive tactile sensor according to claim 8, characterized in that: The NoSOL database technology is used to connect data with the existing medical database, and the palpation data obtained by the measuring head (1) and the disease results diagnosed by the medical expert through endoscope vision are stored in the medical database together. The attention mechanism is used to dynamically adjust the classification weights of the palpation data distribution attribute information and the medical expert knowledge and experience information, and the "abnormal samples" are reclassified into different disease period categories to complete the establishment of the palpation data set, which is used as the training set of the deep learning algorithm.

Images