Disclosure of Invention
The invention aims to provide an interventional touch sensor.
Another technical problem to be solved by the present invention is to provide an application method of the above-mentioned interventional tactile sensor.
In order to solve the technical problems, the technical scheme of the invention is as follows:
an interventional touch sensor comprises a head measuring part, a conduit part, a host part and a displacement part, wherein the conduit part is connected with the head measuring part and the host part,
The measuring head comprises a GRIN lens (self-focusing lens), a light filtering element, a waveguide substrate, a waveguide core layer and an elastomer element with a hemispherical top, wherein the front end of the GRIN lens is fixedly connected with the light filtering element and the elastomer element, the waveguide substrate and the waveguide core layer are sequentially arranged on the outer side of the light filtering element, the waveguide core layer is embedded into the elastomer element, a plurality of elastic contacts are arranged on the inner surface of the elastomer element, an air gap (about a few microns) is reserved between the tip end of each elastic contact and the waveguide core layer, and the elastic contacts are identical in height or different in bottom surface diameter;
The catheter part comprises a fiber core of an optical fiber, a cladding, a coating (protecting) layer and an optical fiber interface, wherein one end of the fiber core is connected with the GRIN lens, the other end of the fiber core is connected with the optical fiber interface, the cladding is arranged outside the fiber core, more than half of the lateral area of the cladding and the GRIN lens is coated with the coating, and the optical fiber interface is in threaded connection with the host part;
The host part comprises an optical fiber circulator, a laser light source, a photoelectric detector and a Microprocessor (MCU), wherein the Microprocessor (MCU) and the optical fiber circulator are respectively connected with the laser light source and the photoelectric detector in a circuit manner, and an optical fiber interface of the catheter part is connected to a packaging shell of the host part;
The displacement portion is located host computer below, and the casing inside of displacement portion is provided with mechanical transmission, and the casing outside is provided with displacement switch, wherein, mechanical transmission includes driving piece, drive pawl, ratchet, swing mechanism, slider, driving piece and fixes first spring, second spring and the third spring that have reset function on the casing, driving piece and displacement switch contact connection, displacement switch promote driving piece clockwise rotation when anticlockwise rotation, driving piece and first spring fixed connection, drive pawl one end and driving piece fixed connection, the other end is connected and adjacent department and third spring fixed connection with the ratchet contact, the centre of swing mechanism regard as the fulcrum with a support of fixing on the casing, one end and ratchet contact connection and adjacent department and second spring fixed connection, the other end and slider pass through the spout connection, slider and driving piece fixed connection, driving piece and host computer portion fixed connection, host computer portion passes through sliding guide and displacement portion sliding connection.
According to the intervention type touch sensor, the guide block of the sliding guide rail is arranged on the packaging shell of the host part, the guide groove of the sliding guide rail is arranged on the shell of the displacement part, and the host part is in sliding connection 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 piece can drive the host part to do reciprocating linear motion along the sliding guide rail, and further drive the guide pipe part connected with the host part to do reciprocating linear motion.
The waveguide substrate, the waveguide core layer and the air layer (waveguide cover layer) between the waveguide core layer and the elastic contact of the probe head form the basic structure of the waveguide, and the GRIN lens is used as a relay lens, and can expand the output light of the optical fiber (comprising the fiber core and the cladding) into parallel light with 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 to enable the output light to enter the optical fiber entirely. The laser light source generates near infrared light with 850nm wavelength, the near infrared light is transmitted into the measuring head through the fiber circulator, the near infrared light enters the waveguide after being collimated through the GRIN lens, the elastomer is not contacted with the waveguide in the case of no contact load, the light meeting the guided wave condition is stably transmitted in the waveguide in the form of total internal reflection, the output light of the waveguide enters the fiber through the GRIN lens to form a 'ring-shaped optical loop', and finally, the output light of the fiber is transmitted into the photoelectric detector through the fiber circulator, and the light path trend is the light path trend in the case of no contact load.
Preferably, the interventional touch sensor is prepared by manufacturing a die of the elastomer element by a 3D printing process, wherein the die comprises a die A and a die B, printing and manufacturing a micro-nano structure on the inner surface of the die B, injecting a silicone rubber prepolymer solution into the die, performing vacuum degassing, separating the cured silicone rubber prepolymer solution from the die to obtain the elastomer element, wherein the inner surface of the elastomer element is provided with elastic contact arrays with different sizes, and the outer surface of the elastomer element is a micro-nano structure surface with super-hydrophobic characteristics.
Preferably, in the above-mentioned interventional touch sensor, the elastomer element is made of silicone rubber, and its elastic modulus changes with the proportion of the main agent and the curing agent, so as to realize flexible adjustment of the measuring range and resolution of the sensor.
Preferably, in the above-mentioned interventional touch sensor, the elastic contacts are conical, each elastic contact is disposed inside the elastic element in a ring-shaped manner, and the diameter of the bottom of the elastic contact dispersed in the outer ring is larger than the diameter of the bottom of the elastic contact dispersed in the inner ring.
When a contact load is applied, the elastic contact on the inner surface of the elastic element is contacted with the waveguide core layer, so that the guided wave condition of the elastic contact is destroyed to generate radiation light, and the larger the contact area of the elastic contact and the waveguide core layer is, the more leakage light is generated, so that a force-light conversion relation is established. The hemispherical shell-shaped waveguide has dense light near the center and sparse light at the edges, which results in the light intensity in the waveguide exhibiting a decreasing distribution from the center to the edges according to a certain rule. Therefore, the invention optimizes the arrangement mode of the cone-shaped elastic contact arrays on the inner surface of the elastic body element, and sets the elastic contact arrays to be distributed in a mode of gradually increasing the diameters of the bottom surfaces from the center to the edge, so as to compensate the uneven problem of the light intensity distribution in the waveguide, thereby realizing isotropic measurement resolution for the contact loads with the same size in the sensing range of 180 degrees of the measuring head part.
The measuring head of the interventional touch sensor is preferably prepared by adopting a filter element with high absorption characteristic for near infrared light with the wavelength of 850nm as a substrate, sequentially depositing a waveguide substrate and a waveguide core layer on the upper surface of the filter element by adopting a vacuum evaporation process, and sequentially bonding the filter element and the elastomer element with the GRIN lens by adopting medical UV glue to finish the fixation of each element in the measuring head.
Preferably, in the above-mentioned intervention type touch sensor, the outside of the host portion is encapsulated by using a medical metal material, and a digital display screen is further disposed on the surface and connected with the microprocessor through a circuit.
Preferably, in the above-mentioned intervention type touch sensor, the host portion further includes a bluetooth transmission module and a wireless charging module, and the bluetooth transmission module and the wireless charging module are respectively connected with the microprocessor circuit.
Preferably, in the above-mentioned interventional touch sensor, the laser light source is near infrared light with a wavelength of 850nm, and the near infrared light is transmitted through an optical fiber, collimated by a GRIN lens, enters a waveguide core layer of the near infrared light from one side of the waveguide, then exits from the other side of the waveguide, is converged by the GRIN lens, enters an optical fiber core for transmission, and finally is received by a photodetector.
The application method of the interventional touch sensor comprises the steps of uploading palpation data obtained by a measuring head to a computer through a Bluetooth module of a host computer, drawing a force-displacement curve through an interpolation method, normalizing force values at a palpation initial contact state to zero values through preprocessing of each group of palpation data, performing disease diagnosis through a deep learning algorithm, firstly extracting primary characteristics of the force-displacement curve through a Fully Connected Neural Network (FCNN), then extracting space-time characteristics of diagnostic data in medical records of the same patient at different times through a long-term and short-term memory network (LSTM), and finally predicting the output vectors of the characteristic extraction network in two forms through a FCNN characteristic classification network, wherein the type and the stage of the disease are predicted, and the prediction result is stored in a medical database.
Preferably, in the application method of the interventional touch sensor, a NoSOL database technology is used for carrying out data butt joint with an existing medical database, palpation data obtained by a measuring head and a disease result of a medical expert through visual diagnosis of an endoscope are stored in the medical database, a attention mechanism is utilized for dynamically adjusting classification weights of palpation data distribution attribute information and medical expert knowledge experience information, abnormal samples are reclassified into different disease period categories, and the palpation data set is established and used as a training set of the deep learning algorithm.
In the use process, the gauge head of the touch sensor enters the human body through the clamp channel of the endoscope, the rotation of the front end is controlled by the bent part of the endoscope, and a measuring object is searched through the visual information of the objective lens, so that the problem that the endoscope can only observe the corroded area of the surface of pathological tissues and can not acquire the mechanical information of the surface of the tissues is solved, and the combination of observation and measurement is realized. Based on the medical database, the measurement deviation of the tactile sensor and the mechanical difference of the soft tissues in different patients are corrected by a model algorithm, so that the disease stage is accurately and rapidly quantitatively diagnosed.
The beneficial effects are that:
The interventional touch sensor is an interventional touch sensor with large sensing angle isotropy resolution, is based on the total internal reflection principle of a waveguide, solves the core problems of electric passivity, miniaturization, large sensing angle, adjustable resolution and the like required in interventional touch diagnosis by utilizing the rule that the internal transmission optical power of the waveguide changes along with the external contact load in the 'living-mechanical' contact process, can keep measurement consistency for different azimuth angles, and can be used for eliminating cross infection hidden trouble among different patients by developing a disposable palpation instrument, and prevent the occurrence of unexpected medical accidents such as viral hepatitis or AIDS and the like of the patients caused by incomplete disinfection. Specifically:
1. elastic contacts with the same height and different bottom diameters are arranged on the inner surface of the elastic element, so that the problem of uneven light intensity distribution of the center and the edge of the hemispherical shell-shaped waveguide is solved, and the measuring head can realize an isotropic resolution measuring effect within a 180-degree sensing range.
2. The medical UV glue is used for respectively bonding the elastomer and the filter element on the GRIN lens, so that the complex process of assembling and debugging an optical machine structure is avoided, and the probe head can be miniaturized so as to be applied to interventional palpation.
3. The die of the elastomer element is manufactured by using a 3D printing process, and the micro-nano structure is printed and manufactured on the contact part of the die and the outer surface of the elastomer element, so that the surface of the elastomer micro-nano structure with super-hydrophobic property can be obtained in batches, and body fluid on the surface of the soft tissue is prevented from adhering to the surface of the measuring head, and the repeatability of measurement is affected.
4. The catheter part is in threaded connection with the main machine part through the optical fiber interface, and is used as medical consumable together with the gauge head part, and after a patient uses the catheter part, a doctor can replace the catheter part by utilizing the optical fiber interface, so that the problem of cross infection among patients caused by incomplete disinfection is prevented.
5. By utilizing the influence of the proportion of the main silicone rubber agent and the curing agent on the elastic modulus of the material, the elastic bodies with different hardness are manufactured, and the measuring range and the resolution of the touch sensor can be adjusted on the premise that the basic structure of the measuring head part is unchanged, so that the touch sensor is suitable for detection of different types of soft tissues.
6. The tactile sensor enters the body from the forceps channel of the endoscope to perform in-situ palpation, solves the problem that the endoscope can only observe the pathological changes of tissues on the premise of not changing the structure of the existing endoscope, provides important biomechanical diagnosis standard for doctors, and enables the endoscope to combine 'observation' and 'measurement'.
7. The palpation data set is established by utilizing the existing medical database, a rich training set is provided for the deep learning model algorithm, and the classification weights of palpation data distribution attribute information and medical expert knowledge experience information are reasonably distributed through the attention mechanism, so that the reliability of the training set is improved.
8. The deep learning model algorithm is trained on the palpation data set based on the medical database, and the rich data set enables the algorithm to realize accurate and efficient disease diagnosis and stage prediction so as to cope with complex and various internal soft tissue environments, reduce missed diagnosis and misdiagnosis of an endoscope, further realize early discovery and early treatment of diseases and improve the cure rate of the diseases, in addition, the palpation data is subjected to standardized processing by the algorithm, so that the error caused by uncertainty of the contact initial state of palpation is solved, and the palpation result is stored in the medical database each time, so that not only the materials of pathological research are increased, but also the training data set of the deep learning model is enriched, and the positive feedback effect is generated on the accuracy rate of disease diagnosis and stage prediction.
Detailed Description
The present invention is described below with reference to examples and drawings, in which the present invention is applicable to an interventional tactile sensor and an application method.
Example 1
1. Design and manufacturing method of head measuring part
As shown in fig. 1, an interventional tactile sensor includes a probe section 1, a catheter section 2, a main body section 3, and a displacement section 4. Since the sizes of the jaws of different types of flexible endoscopes are different, as shown in fig. 9, the present embodiment takes an example of a digestive tract endoscope having a jaw diameter of 3.0mm, and designs a catheter portion of a tactile sensor having a diameter of 2.5mm to fit the jaw size.
As shown in FIG. 3, the appearance of the probe part 1 is a sphere top cylinder, and can be divided into two parts according to purposes, one part is a mechanical sensitive element, namely an elastomer element 11 with a hemispherical top, and the other part is a force-light conversion element, comprising a GRIN lens 7, a light filtering element 8, a waveguide substrate 9 and a waveguide core layer 10, wherein the light filtering element 8 has good light absorption performance on light with the wavelength of 850 nm. The core function of the probe part 1 is to perform a "force-light" conversion, wherein the principle of the "force-light" conversion is as follows:
The optical waveguide consists of an outer two cladding layers of low refractive index medium and a substrate and an inner core layer of relatively high refractive index medium, the refractive index distribution of which is n1＞n2≥n3, wherein n1、n2 and n3 are the refractive indices of the core layer, the substrate and air, respectively.
Stable light transmission in a waveguide requires that two basic conditions be met, namely
n2k0＜β＜n1k0 (I)
κd=mπ+δ12+δ13 (II)
Wherein formula (I) is a guided wave condition, wherein beta is a propagation constant, k0 is a wave number of a light wave in vacuum, only the light wave with the propagation constant beta satisfying formula (I) can propagate in the waveguide, formula (II) is a dispersion equation, the optical path difference of the light wave is required to satisfy a coherence strengthening condition when the light wave is stably transmitted in the waveguide, wherein the formula is an x component of a wave vector of a core layer, d is a thickness of the core layer of the waveguide, delta12 and delta13 are total reflection phase shifts of the light at the interfaces of the core layer and the substrate and the core layer and the cover layer respectively, m is called a mode number, and the value of the m is a non-negative integer of 0,1,2.
As shown in FIG. 6 (a), when the elastomer is not in contact with the waveguide, i.e., no pressure event occurs, the light rays satisfying formulas (I) and (II) propagate in the waveguide and form guided modes by coherent enhancement, while those light waves having propagation constants of 0.ltoreq.β.ltoreq.n2k0 are refracted outwards to form radiated waves.
As shown in FIG. 6 (b), when the elastomer is in contact with the waveguide, i.e., a pressure event occurs, since the elastomer has a refractive index greater than that of the waveguide substrate, i.e., n4＞n2, the guided wave condition of the light at the contact location of the two will no longer be formula (II) but will be
n4k0＜β＜n1k0 (III)
Thus, those rays having a propagation constant of n2k0≤β≤n4k0 will radiate outward out of the waveguide's confinement. If the contact area of the elastic body with the waveguide is continuously enlarged under the action of external load, the waveguide is caused to radiate more energy outwards, so that the optical power transmitted in the waveguide is continuously reduced, and a functional relation between the optical power variation delta P transmitted in the waveguide and the contact load delta F is established, and the conversion from force to light is completed.
From the "force-light" conversion principle, it is known that the functional relationship between the amount of change Δp in the transmitted optical power in the waveguide and the contact load Δf is established by intermediating the contact area of the elastomer contact with the waveguide. However, as shown in (d) of fig. 6, the light intensity distribution inside the hemispherical shell-shaped waveguide is schematically shown, since the light inside the hemispherical shell-shaped waveguide is input and output through the edges, the light intensity inside the waveguide is gradually increased from the edges to the center, so that even if the contact areas are the same, if the contact positions of the elastomer contacts and the hemispherical shell-shaped waveguide are different, the waveguides may generate different amounts of radiant energy, namely, when the radiant light energy of the elastomer contacts and the hemispherical shell-shaped waveguide are the same on the same cross section of the hemispherical shell-shaped waveguide, otherwise, the radiant light energy radiated by the contact is different, namely, the measuring head shows anisotropic measurement results for different sensing angles.
To solve this problem, the present embodiment contemplates an array of elastomeric contacts 12 of equal height with the bottom surface diameter decreasing from edge to center. As shown in fig. 6 (b) and (c), the elastomer contacts in fig. (c) have larger diameters than the bottom surface in fig. (b) when the same pressure event occurs, so that the waveguide generates more optical energy loss under the same guided wave condition (formula (III)). Therefore, the invention compensates the problem caused by uneven light intensity distribution in the hemispherical shell-shaped waveguide by changing the diameter of the bottom surface of the contact, so that the measuring head shows isotropic mechanical resolution within a large sensing angle range of 180 degrees, as shown in figure 7.
As shown in fig. 6 (a) and (b), since the maximum contact area Smax of each of the elastomer contacts with the hemispherical-shaped waveguide is a certain value, it is known from the "force-light" conversion principle that each of the elastomer contacts causes the maximum light loss in the waveguide to be a constant value. Thus, the range of the tactile sensor of the present invention can be defined as the amount of pressure at which the contact area of the elastomeric contact with the waveguide is Smax, denoted Fmax.
Because of the difference in mechanical properties of the soft tissue surfaces at different locations, when Smax is unchanged, for those biological tissues with harder surfaces, an elastomer with a higher elastic modulus needs to be selected, namely Fmax is increased, and for those biological tissues with softer surfaces, when disease diagnosis is performed on the biological tissues, an elastomer with a lower elastic modulus needs to be selected, so that the resolution of the sensor delta Fmin is increased, and the mechanical information of the biological tissues is captured more accurately.
In this embodiment, polydimethylsiloxane (PDMS) is selected as the material of the elastomer member 11, and PDMS is a silicone rubber material frequently used in the biomedical field, and has good biocompatibility. When the proportions of the main agent and the curing agent are different, PDMS can show elastic moduli with different magnitudes, and when the PDMS contacts with loads with the same magnitude, the PDMS can cause different deformation amounts, so that the contact areas of the elastic contacts 12 and the waveguide core layer 10 are different, and different degrees of optical loss are generated, thereby obtaining the measuring head part 1 with different mechanical measuring ranges and resolutions, and the elastic moduli of the measuring head part 1 are sequentially decreased according to the common proportions of 5:1, 10:1, 15:1 and 20:1. Therefore, according to the different properties of the measured object, the elastomer element 11 can be made of PDMS materials with different proportions to obtain different measuring ranges and resolutions.
Because the body fluid of a person has certain viscosity, a small amount of liquid may adhere to the outer surface of the probe during the contact process, so that a larger error exists in the next measurement result. This embodiment overcomes this problem by improving the manufacturing process flow of the elastomeric element 11. As shown in fig. 8, first, a mold of the elastomer element 11 is manufactured using a 3D printing process, the mold including two parts of a mold a and a mold B, a micro-nano structure is printed on an inner surface of the mold B, and then a silicone rubber prepolymer solution is injected into the mold and vacuum deaerated, and the resulting elastomer element 11 is separated from the mold after curing. The elastic contact 12 arrays with different sizes are arranged on the inner surface of the elastic element 11, and the outer surface is a micro-nano structure surface with super-hydrophobic characteristic, so that the problem that the measuring head is easy to be adhered by body fluid is solved, and the repeatability of measurement is ensured. In addition, in order to prevent interference of radiation light of the waveguide to internal transmission light, 1-2% of dark red silicon paste material is doped into the PDMS material, so that the PDMS material has good light absorption performance at a wavelength of 850 nm.
This embodiment refers to a material arrangement of an SOI waveguide with a core layer, cladding layer and substrate of silicon (Si), air and silicon dioxide (SiO2) in that order. At a wavelength of 850nm, the refractive index of Si is 3.64 and the refractive index of SiO2 is 1.45.
Since the elastomer element (11) is made of PDMS material with the ratio of the main agent and the curing agent being 15:1, the refractive index of the elastomer element is 1.40, which is smaller than that of the SOI waveguide substrate SiO2, and n4＞n3 (as shown in fig. 6 (b)) is required to be replaced in order to satisfy the guided wave conditions of the formulas (I) and (II), the guided wave medium MgF2 with a lower refractive index is selected, which is 1.38, and the design requirement can be satisfied. Thus, the material of the waveguide in this embodiment is a core layer (Si, n1 =3.64), a substrate (MgF2,n2 =1.38), and a cladding layer of air (n3 =1).
When total reflection occurs, the light rays in the waveguide are not completely reflected back into the core layer at the interface, but penetrate into the substrate or the cover layer to a certain depth, and propagate along the interface for a certain distance, and then are emitted along the direction of the reflected light rays, wherein the depth is called penetration depth, and the light waves propagating along the interface are called evanescent waves. The evanescent wave takes part in the propagation of the guided mode field and carries a certain energy, which increases the effective thickness de of the guided mode field over the geometric thickness d of the core layer, i.e
Wherein 1/p and 1/q are the penetration depths of light into the substrate and the cover layer, respectively, and p and q are the attenuation constants in the x-direction in the substrate and the cover layer, respectively, i.e
Where k0 =2pi/λ. The thickness of the substrate of the final design waveguide is 20 mu m, and the thickness of the covering layer is 200 mu m by comprehensively considering the application requirements of the waveguide, the processing technology and other factors. Thus, the present embodiment completes the design of the characteristic parameters d, n1、n2、n3 and the transmission wavelength λ of the waveguide.
As shown in fig. 4, the GRIN lens 7 is used as a relay lens, which is a cylindrical optical lens different from a normal lens in that the refractive index distribution of the self-focusing lens material is gradually reduced in the radial direction, and can continuously refract light transmitted in the axial direction, thereby realizing smooth and continuous convergence of outgoing light. The embodiment adopts a GRIN lens with the pitch of 0.25P, which can collimate and expand the output light of the optical fiber to a larger area so as to lead part of the light to be waveguide input light, and can also converge the output light of the waveguide so as to lead the output light to enter the optical fiber completely, thereby forming a ring-shaped optical circuit. The use of GRIN lenses, in combination with the unique design of the probe head, solves the complex coupling problem between the hemispherical shell shaped waveguide and the optical fiber.
As shown in fig. 3, the radial dimension of the head of this embodiment is 2mm, i.e., the outer surface radius of the elastomeric element and the bottom surface diameter of the GRIN lens are both 2mm. The manufacturing method of the probe part 2 comprises the steps of firstly adopting a light filtering element 8 as a substrate, wherein the light filtering element 8 has high absorption characteristic on near infrared light with the wavelength of 850nm, so that interference of stray light such as radiated light of a waveguide is avoided, then adopting a vacuum evaporation process method to sequentially deposit a waveguide substrate 9 and a waveguide core layer 10 on the upper surface of the element 8, and sequentially adopting medical UV glue for bonding between the light filtering element 8 and the elastic body 11 and the GRIN lens, so that fixation of all elements in the probe can be completed without a mechanical sleeve, and complex processes of assembling and debugging an optical-mechanical structure are avoided, as shown in figure 8.
2. Design of conduit portion, main machine portion and displacement portion
As shown in figure 2, the catheter part 2 is connected with the probe part 1 and the host part 3 and is used for mediating optical signal interaction, the optical fiber consists of a fiber core 13 and a cladding 14, the cladding 14 is arranged outside the fiber core 13, a coating layer 15 of the catheter part 2 is made of medical rubber material, and the optical fiber and more than half of the side area of the GRIN lens 7 of the probe part 1 are coated and fixed with the probe part 1 through medical UV glue. The coating layer 15 not only plays a role of protecting the internal optical fiber but also plays a role of reinforcing the connection of the probe portion 1 and the catheter portion 2. In this embodiment, a single mode fiber is used as the optical fiber in the catheter section 2. The other end of the conduit part 2 is provided with an optical fiber interface 5 which is in threaded connection with the packaging shell of the host part 3, so that the probe part 1 and the conduit part 2 can be conveniently and flexibly replaced, and cross infection among different patients is avoided. After the die of the head measuring part 1 and the catheter part 2 is manufactured, batch production can be realized, the mass production cost is low, the disposable medical consumable can be used as disposable medical consumable, and the disposable medical consumable is replaced with the main machine part 3 through the optical fiber interface 5, so that the problem of cross infection among patients caused by incomplete disinfection is prevented.
As shown in fig. 4, the host portion 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), where the Microprocessor (MCU) is respectively connected to the laser light source 17, the photodetector 18, the bluetooth transmission module, the wireless charging module, and the other circuit modules (such as the displacement module) in a line. Wherein, the laser light source adopts a laser diode with the wavelength of 850nm, and the photoelectric detector adopts a photoelectric diode. The Bluetooth transmission module transmits mechanical data to the computer in real time, and the wireless charging module can eliminate potential electric shock hidden danger caused by the fact that liquid exists on hands when doctors are charged in a wired mode. The outside of the host part 3 is encapsulated by using medical metal materials, a digital display screen is arranged on the surface of the host part and is connected with a micro processor (MCU) circuit for displaying the mechanical data obtained by measurement in real time, and a guide block for connecting the optical fiber interface 5 of the catheter part 2 and the sliding guide rail 37 is arranged on the encapsulation shell of the host part 3.
As shown in fig. 2, the displacement part 4 is arranged below the main body part 3, a mechanical transmission device is arranged inside a shell of the displacement part 4, a displacement switch 6 for controlling the mechanical transmission device and a guide groove of a sliding guide rail 37 are arranged outside the shell, the displacement switch 6 only has a triggering effect on the mechanical transmission device, and the main body part 3 is in sliding connection with the guide groove of the sliding guide rail 37 on the displacement part 4 through a guide block of the sliding guide rail 37. The displacement switch 6 will swing backwards when pressed by the physician, where the mechanical transmission inside the displacement part 4 is triggered, it will return to its home position immediately when the physician stops pressing the displacement switch 6, and it will not interfere with the movement of the mechanical transmission during the return of the displacement switch 6.
As shown in fig. 11, the displacement transmission device pushes the main body part 3 together with the catheter part 2 and the probe part 1 to move forward by a constant displacement L after the displacement switch 6 is started. As shown in fig. 10, the inside of the mechanical transmission device includes a driving member 28, a driving pawl 29, a ratchet 30, a swinging mechanism 31, a sliding block 32, a driving member 33, and a first spring 34, a second spring 35 and a third spring 36 which are fixed on the housing and have a reset function, wherein the driving member 28 is in contact connection with the displacement switch 6, the displacement switch 6 pushes the driving member 28 to rotate clockwise when rotating anticlockwise, the driving member 28 is fixedly connected with the first spring 34, one end of the driving pawl 29 is fixedly connected with the driving member 28, the other end is in contact connection with the ratchet 30 and the adjacent part is fixedly connected with the third spring 36, the middle of the swinging mechanism 31 takes a bracket fixed on the housing as a fulcrum, one end is in contact connection with the ratchet 30 and the adjacent part is fixedly connected with the second spring 35, the other end is connected with the sliding block 32 through a sliding groove, the sliding block is fixedly connected with the driving member 33, and the driving member is fixedly connected with the host portion 3, and since the host portion is in sliding connection with the displacement portion 4 through a sliding guide rail 37, the driving member 33 can drive the host portion 3 to reciprocate linearly along the sliding guide rail 37, thereby driving the host portion 3 to reciprocate linearly and reciprocate with the catheter portion 2.
The specific motion principle is that the driving part 28 rotates clockwise under the pushing of the displacement switch 6 and drives the driving pawl 29 to push the ratchet teeth of the ratchet wheel 30 to rotate around the bracket by a certain angle, the fixed fulcrum of the first spring 34 is positioned in front of the driving part 28, after the driving part 28 rotates anticlockwise, a doctor stops pressing the displacement switch 6, the driving part 28 is pulled by the first spring 34 to return to the initial position to finish the reset, one end of the swinging mechanism 31 close to the ratchet wheel 30 acts as a 'braking claw' in the process, the ratchet wheel 30 can rotate clockwise only, the swinging mechanism 31 can be pushed by the ratchet teeth of the ratchet wheel 30 to rotate anticlockwise in the process of rotating clockwise of the ratchet wheel 30, a horizontal force component is generated to push the sliding block to move forwards at the moment, so that the sliding block 32 and the driving part 33 are loaded forwards by constant displacement L, the second spring 35 is arranged in front of the swinging mechanism 31, and then one end of the swinging mechanism 31 close to the second spring 35 contacts with the ratchet teeth of the ratchet wheel 30 again to finish the reset, namely, the mechanical motion is completed in the process of returning to the initial position in the process of moving along with the ratchet wheel 30.
As shown in FIG. 4, the path inside the tactile sensor runs in the absence of contact loading, with the laser diode generating near infrared light at a wavelength of 850nm, which is transmitted into the fiber via the fiber circulator 16. The optical fiber circulator is a multi-port nonreciprocal optical device and can realize bidirectional optical signal transmission on a single optical fiber. The light in the single-mode fiber is collimated by the GRIN lens 7 and then is transmitted into the waveguide, the light meeting the guided wave condition is stably transmitted in the waveguide in a total internal reflection mode, the output light of the waveguide is converged into the single-mode fiber by the GRIN lens 7, and finally, the output light of the single-mode fiber is transmitted into the photoelectric detector 18 by the optical fiber circulator 16, so that the signal conversion process of force, light and electricity is completed.
As shown in fig. 5, when there is a contact load, that is, the probe portion 1 contacts with soft tissue, the elastic contact 12 in the probe portion 1 will contact with the waveguide, and as known from the "force-light" conversion principle, the waveguide will generate optical loss, that is, the output optical power of the waveguide will be reduced, and finally the attenuation of the output electrical signal of the photodetector will be represented, so as to analyze the magnitude of the measured force value.
3. Operation method of palpation
In use, as shown in fig. 9, a doctor A holds an endoscope 19 and inserts an insertion tube 22 into a patient, under the illumination condition of a light guide lens 26, a lens 25 is used for searching for soft tissues to be detected, after the target soft tissues are locked, a doctor B holds a displacement part 4 with one hand and holds a catheter part 2 to guide a head part 1 from a forceps channel inlet 21 of the endoscope and guide the head part out of a forceps channel outlet 24, the initial contact state of the head part 1 and the target soft tissues is determined through the reading of a digital display screen of a sensor, namely the position of a measuring head part 1 when the digital display screen has just reading display, at the moment, the doctor B presses a displacement switch 6 to start a mechanical transmission device in the displacement part 4, and the mechanical transmission device pushes the host part, the catheter part 2 and the head part 1 to advance by a constant displacement L together, and then returns to the initial contact state immediately, namely the touch process is finished, as shown in fig. 11. After one palpation is completed, the doctor B holds the catheter part 2 to guide out the head part 1 to the side of the forceps channel entrance 21 until the head part 1 is not observed in the objective lens 25, then the doctor a continues to operate the insertion tube 22 to advance/retreat and search for the next soft tissue to be detected through the objective lens 25, or the doctor a can also control the bending part 23 of the endoscope to drive the front end part 27 to rotate up, down, left and right through operating the angle control knob 20 of the bending part of the endoscope to search for the soft tissue to be detected, as shown in fig. 12. When doctor A obtains the position information of the target soft tissue through the objective lens, doctor B can repeat the palpation process to palpation. The overall flow of intelligent palpation is shown in fig. 5.
4. Intelligent palpation algorithm
The intelligent palpation algorithm in this embodiment is implemented by the following four steps:
step one, docking the current medical database by using NoSQL data rate technology
In the clinical trial of palpation, medical professionals can use the palpation method to perform in situ palpation in patients to assist the endoscopic vision to realize diagnosis and staging of diseases. As shown in fig. 5, the data result of palpation is uploaded to the computer through the bluetooth module of the host part 3 for data processing. The palpation data and the staged results of doctor diagnosis and treatment suggestions 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), a laboratory information management system (LIS) and the like, and is convenient for accessing palpation data and interfacing with the existing medical data.
The present embodiment interfaces with existing databases using NoSQL database technology. The method is a non-relational database technology and has good flexibility and scalability. By using NoSQL database technology, the present embodiment enables efficient storage and management of unstructured data in the medical database, providing a rich medical sample for in vivo palpation and pathology studies.
Step two, establishing palpation data set and reclassifying abnormal sample
For patients diagnosed by endoscope vision as middle or late stage of disease, the corresponding palpation data is directly stored in a medical database, while for patients diagnosed by endoscope vision as no disease, benign disease or early stage of disease, due to the high possibility of misdiagnosis of the diagnosis result, after the corresponding palpation data is stored in the medical database, medical professionals also need to continuously track the subsequent disease development condition of the patients by using the medical database. Medical professionals need to observe whether a patient diagnosed as having no disease or benign disease subsequently presents with a problem of disease or exacerbation of the disease, and whether a patient diagnosed as early in disease subsequently presents with no disease, benign disease or exacerbation of the disease. If the patient subsequently comes in both cases, the diagnosis result previously obtained by the medical expert using the endoscope vision has a problem of misdiagnosis, and the corresponding palpation data of the patient is defined as an "abnormal sample".
The embodiment adopts a feature fusion network based on the palpation data attribute distribution information and the medical expert knowledge experience information of an attention mechanism. Firstly, the K-means clustering method is adopted to recluster the disease types and stages of the abnormal samples in the palpation data set by using the attribute distribution of the diagnosis data. And then, dynamically adjusting the weights of the K-means classification result and the medical expert diagnosis result by using the characteristic fusion network of the attention mechanism, so that the abnormal sample is more accurately secondarily classified, and the influence of subjectivity of medical expert diagnosis experience and the limitation of the K-means clustering method on the diagnosis result is reduced. And finally, updating and expanding data by using a NoSQL database technology and the existing medical database, thereby further improving the accuracy and reliability of the palpation data set, as shown in fig. 13.
Step three, pretreatment of palpation data
Firstly, 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 of a fixed-length serialization data type, the independent variable is the contact displacement of the measuring head part 1 and soft tissues, the function value is the force value output by a touch sensor, and a palpation force-displacement curve can be obtained by an interpolation method. Then, because of the uncertainty of the palpation "contact initial state" of the gauge head part 1 with the soft tissue at each measurement, the force at the displacement origin of the palpation data is not zero value and slightly larger than zero value. To solve this problem, the present embodiment requires pre-processing of the palpation data prior to processing the data using the deep learning model algorithm by subtracting the force values at the displacement zero values from all the mechanical values in each of the palpation data sets so that the force values for all the palpation data sets start from zero.
Feature extraction and disease prediction based on palpation data
As shown in fig. 13, the present embodiment uses a Fully Connected Neural Network (FCNN) as a feature extraction and disease stage network, and first, a palpation data set is input into the neural network FCNN to extract the trend feature of the "force-displacement curve".
The present embodiment utilizes a long-short term memory network (LSTM) to extract spatiotemporal features of diagnostic data in medical records of different time visits of the same patient in a medical database. First, the LSTM network has memory that maps the diagnostic information for each time step into a fixed-size vector representation by encoding a time series of past diagnostic data for the patient. The vector representations are then sequentially arranged in a time series to form a spatiotemporal feature representation of the diagnostic data, thereby enabling capture of the course of the patient's disease condition. The LSTM network is beneficial for medical professionals to master the change trend and evolution process of the patient diagnosis data in the time dimension, so that diagnosis results and treatment opinion reports are more scientifically given, and the generation of abnormal samples is reduced.
The present embodiment uses FCNN neural networks to classify the disease from the output vector of the feature extraction network. The FCNN-feature classification neural network predicts disease in two forms, the type and stage of disease.
In the practical application process, each group of data obtained by clinical palpation is stored into a medical database together with diagnosis results after disease diagnosis and stage prediction by the deep learning algorithm, the size of palpation data sets is further enriched, the size of palpation data sets in the medical database determines the prediction accuracy of the deep learning algorithm in terms of disease diagnosis and stage, and the two are complemented with each other, so that the accuracy of the prediction result of the deep learning algorithm can be continuously improved to cope with complex and various clinical cases, thereby better assisting an endoscope to accurately and efficiently diagnose and predict the disease stage, realizing early discovery and early treatment of the disease, and improving the clinical cure rate.
The foregoing is merely a preferred embodiment of the present invention and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present invention, which are intended to be comprehended within the scope of the present invention.