An individualized laryngeal mask processing method based on an Ai vision model

Abstract

The application belongs to the field of laryngeal mask processing, and particularly relates to an individualized laryngeal mask processing method based on an Ai vision model: input and size calibration: collecting multi-view photos of the head and neck of a patient, calibrating the pixel size and the actual size, and obtaining a scale factor of each photo. External morphological feature vector: quantifying the scale factor of each photo and the head and face contour into an external morphological feature vector x. Laryngeal mask parameter set generation and definition: defining a laryngeal mask structure parameter set P, and converting the external morphological feature vector into a laryngeal mask parameter estimate value and an uncertainty index by using a mapping model of the Ai vision model. Geometric constraint and safety margin correction: applying geometric, manufacturing constraints and safety margins to the parameters to obtain final parameters. Geometric modeling: based on constructing a laryngeal mask integrated function model, deriving a 3D printing file and printing a finished product, and ensuring model adaptation and accurate position.

Description

Technical Field

[0001] This invention belongs to the field of laryngeal mask fabrication, specifically a personalized laryngeal mask fabrication method based on an AI visual model. Background Technology

[0002] Currently, laryngeal masks are widely used in clinical anesthesia because they have a lower incidence of oropharyngeal complications compared to endotracheal intubation.

[0003] A laryngeal mask airway is a supraglottic airway device designed to create a seal around the larynx without directly contacting the glottis.

[0004] However, studies have found that the correct placement rate of laryngeal masks in clinical practice is low, ranging from only 12.8% to 49%, which increases the risk of airway obstruction, hypoxia, and even suffocation during anesthesia management.

[0005] The main reasons include: 1. Fixed laryngeal mask size: The commonly used 3# and 4# laryngeal masks for adults are suitable for patients weighing 30~50kg and 50~70kg respectively. However, the upper respiratory tract anatomy may differ among patients of the same weight, which is the direct reason why the position of the same laryngeal mask may be inconsistent when patients of the same weight or within the applicable weight range use the same model of laryngeal mask.

[0006] 2. Different insertion depths of the laryngeal mask can cause deviations in the relative positions of the inner cuff ring and the glottis.

[0007] 3. Different inflation pressures of the laryngeal mask CUFF will change the anterior-posterior diameter of the ventilation mask section, thus causing the glottis position to shift.

[0008] In addition, differences in laryngeal mask models can directly affect the insertion depth and the anterior-posterior diameter of the mask. Summary of the Invention

[0009] The purpose of this invention is to overcome the above-mentioned shortcomings and provide an individualized laryngeal mask fabrication method based on an AI visual model.

[0010] To address the aforementioned technical problems, this invention provides the following technical solution: a personalized laryngeal mask fabrication method based on an AI visual model, comprising the following steps: Input and size calibration steps: Acquire multi-view photographs of the patient's head and neck; convert the pixel scale in the photographs to the actual size through calibration to obtain the scale factor of each photograph; External morphological feature extraction steps: Perform artificial intelligence reasoning on multi-view photos to extract external anatomical points, and quantize the anatomical points and contours into external morphological feature vectors x; Parameter set generation steps: AI visual model is established by learning the anatomical structure of the supraglottic airway and the position of the laryngeal mask from 3D-CT images of patients with different body types. The external morphological feature vector x is analyzed by the AI ​​visual model to obtain the parameters of each part of the laryngeal mask with the optimal position of the patient's supraglottic airway anatomical structure and laryngeal mask. The parameter set definition steps are as follows: Define the laryngeal mask structure parameters as a set P, including ventilation tube parameters, basic bag geometry parameters, and connector parameters; then utilize the mapping model of the trained AI vision model. The external morphological feature vector x is mapped to parameter estimates: ,in These are the initial estimates of the parameters. For each parameter, there is an uncertainty index or confidence level. Geometric constraint and safety margin correction steps: By applying geometric constraints, manufacturing constraints, and safety margins, the parameters for the final model are obtained. ; Geometric modeling steps: Based on A functional algorithm model for the connection of the ventilation tube, connecting section and bag of the laryngeal mask was constructed, and a file suitable for 3D printing was exported to print the finished laryngeal mask.

[0011] Preferably, in the geometric modeling step, the functions of the function algorithm model include geometric parameter formula algorithms for the ventilation duct, connecting section, diaphragm, and integrated model.

[0012] Preferably, the algorithm for the geometric parameters of the ventilation duct includes: Cross-sectional area of ​​cylinder: S0 = π·r 2 Circumference of the cylinder's cross-section: L0 = 2πr; Volume of the cylinder: V0 = πr 2 ·h, where r is the radius of the cylinder and h is the length of the ventilation duct.

[0013] Preferably, the geometric formula algorithm for the diaphragm includes: Thickness non-uniformity function: ; Water droplet flattening correction function: ; Circular cross-sectional area function: ; Integral function of volume: ; The function for summing the equal parts of a volume: ; in, Indicates the circumferential angle ∈[0,2π], Indicates the principal direction of thickness. Indicates non-uniform amplitude. Represents the original constant parameter. The direction and angle of compression It is a non-negative constant. Represents the cross-sectional area of ​​the reference annular ellipse. Indicates the radius of the center line of the hood sweep. A small increment in the circumferential angle, The circumferential angle corresponding to the i-th segment.

[0014] Preferably, the algorithm for the geometric parameter formula of the connecting segment includes: Connecting segment volume function: ; Transition segment volume function: ; in Indicates the length of the transition section. Indicates the area of ​​the end cross section. This represents the cross-sectional area at the starting end of the transition section. This represents the cross-sectional area at the middle of the transition section. Indicates the radius of the cylinder. This indicates the length of the connecting segment; the transition segment is used to connect the connecting segment to the diaphragm.

[0015] Preferably, the algorithm for the geometric parameter formula of the integrated model includes: Overall volume: .

[0016] Preferably, the multi-view photograph includes at least a frontal view, a side view, and an upward view.

[0017] Preferably, in the external morphological feature extraction step, the external anatomical points include at least the chin point, mandibular angle point, mandibular border contour, thyroid cartilage region, suprasternal notch projection point, and oral fissure boundary point.

[0018] Preferably, in the input and dimension calibration step, the calibration method includes at least one of the following three: Take a picture of a reference ruler of known length placed in the shooting area; Use a mobile terminal with ranging capabilities to obtain the shooting distance and perform scale restoration; Statistical dimensions of the external anatomy were used as weak calibration.

[0019] Preferably, the anatomical structures of the supraglottic airway include the maxilla, mandible, oral cavity, hard palate, soft palate, pharynx, larynx, epiglottis, base of tongue, cricoid cartilage, thyroid cartilage, arytenoid cartilage, and glottis.

[0020] Compared with the prior art, the present invention has at least the following beneficial effects: This invention collects multi-view photographs of the patient's head and neck via a mobile terminal APP, extracts external anatomical features using an artificial intelligence visual model, and maps them to a set of laryngeal mask airway (LMA) structural parameters. After parameter generation, the manufacturability and clinical applicability of the parameters are ensured by combining geometric constraints and safety margin mechanisms based on literature. Finally, based on this parameter set, a continuous integrated function algorithm model of the LMA ventilation tube (Stem), connecting segment (Link), and CUFF is constructed using AI combined with calculus. Closed mesh files such as STL / 3MF that can be used for 3D printing are exported, and then the LMA is printed to ensure its model compatibility and positional accuracy. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the laryngeal mask according to an embodiment of the present invention; Figure 2 This is a structural diagram of the ventilation tube model of the laryngeal mask according to an embodiment of the present invention; Where ① Stem long axis: equivalent to the distance 2r between the patient's maxillary canines; ② Stem short axis: equivalent to 2 / 3 of the patient's mouth opening 2r; ③ Stem length: equivalent to the distance h from the patient's incisors to the root of the tongue; Figure 3 This is a structural diagram of the laryngeal mask airway model according to an embodiment of the present invention. Figure 1 ; The green line represents the long axis d of the Cuff outer ring: the distance from the base of the tongue to the esophageal sphincter. The red line represents the long axis e of the inner ring of the Cuff: the distance from the epiglottis to the interarytenoid notch; The brown line represents the short axis f of the outer ring of the Cuff: the distance between the lateral edges of the throat. The light blue line represents the short axis g of the inner ring of the Cuff: the distance between the walls of the arytenoid epiglottis; The light orange line represents the short axis g of the inner ring of the Cuff: the distance between the walls of the arytenoid epiglottis; Light gray represents the distance i from the epiglottis to the root of the tongue; ←→ indicates the distance j from the interarytenoid notch to the posterior laryngeal wall at the C5 level; Figure 4 This is a structural diagram of the laryngeal mask airway model according to an embodiment of the present invention. Figure 2 ; Figure 5 This is a structural diagram of the connecting section of the laryngeal mask according to an embodiment of the present invention; Where e: major axis of the inner ring of the CUFF; m: major axis of the connection between the Link and the CUFF; α: angle between the CUFF and the Stem; h': inner height of the Link; r: radius of the Stem cylinder; H: outer height of the Link; Figure 6 This is a 3D-CTR image showing the optimal position of the laryngeal mask and upper respiratory tract in an embodiment of the present invention. Level 0 represents the optimal position, and the parameter set P is shown. * It is derived from level 0; in Figure 6 A is a 3D-CTR horizontal bit image. Figure 6 B is a 3D-CTR coronal image; Figure 7 This is a flowchart of the laryngeal mask manufacturing method according to an embodiment of the present invention. Detailed Implementation

[0023] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings. These embodiments are implemented based on the technical solution of the present invention and provide detailed implementation methods and specific operation processes. However, the scope of protection of the present invention is not limited to the following embodiments.

[0024] This embodiment discloses a personalized laryngeal mask fabrication method based on an AI visual model, such as... Figure 7 As shown, it includes the following steps: 1. Input and Dimension Calibration: 1.1. Multi-view photo acquisition: The operator is guided to acquire photos of the patient's head and neck from at least three perspectives via a mobile terminal Ai-APP, preferably including: frontal view, side view, and extension view (or one each of pitch and flexion).

[0025] During the data collection process, the app prompts the patient to maintain a standard posture and open their mouth to a preset level to reduce errors caused by postural differences.

[0026] Optionally, the APP can also collect basic patient information (such as gender, height, weight, neck circumference, mouth opening degree, etc.) to improve the stability of model estimation.

[0027] 1.2. Scale calibration (pixel-millimeter mapping): In order to convert the pixel scale in the photo into the actual size, this embodiment adopts at least one scale calibration method: (1) Place a reference ruler / calibration card of known length in the shooting area (e.g., disposable ruler sticker, standard mouth pad or positioning sticker with scale); (2) Use a mobile terminal with distance / depth measurement capability to obtain the shooting distance and perform scale recovery; (3) In the absence of a ruler, use the statistical scale of external anatomy as weak calibration.

[0028] The above calibration yields the scale factor for each image, which is used to convert the subsequently detected key distances and angles into physical quantities.

[0029] 2. External morphological feature extraction (observable part in Ai-APP): 2.1. External Anatomical Key Point Detection and Segmentation: Perform artificial intelligence reasoning on multi-view photos to execute at least one or more tasks: face / head and neck key point detection, neck contour segmentation, oral fissure region detection, mandibular contour fitting, etc.

[0030] The extracted external anatomical points include, but are not limited to: chin point, mandibular angle point, mandibular border contour, thyroid cartilage region (or its projection location), suprasternal notch projection point, oral fissure boundary point, etc.

[0031] 2.2 Constructing external morphological feature vectors: The extracted external anatomical points and contours are quantified into morphological feature vectors, preferably including: (1) Scales related to the oropharyngeal pathway: mouth opening width, mandibular width, mandibular-chin length, mandibular protrusion, etc.; (2) Angles related to head and neck position: head and neck flexion angle, mandibular-neck angle, etc.; (3) Proportions related to soft tissue shape: neck circumference to mandibular width ratio, neck contour curvature statistics, etc.

[0032] All of the above features are observable in photographs and can be calibrated to millimeters or angles, thus achieving consistency across devices and shooting distances.

[0033] 3. Parameter set generation (Photo → Parameters): 3.1. Based on historical big data (different body types), 3D-CTR images (sagittal, coronal, horizontal, and 3D stereoscopic images) of the anatomical structure of the supraglottic airway and laryngeal mask airway in patients are obtained. The 3D-CTR images include sagittal, coronal, and horizontal cricoid CT images of the maxilla, mandible, oral cavity, pharynx, hard palate, soft palate, larynx, epiglottis, trachea, base of tongue, cricoid cartilage, thyroid cartilage, arytenoid cartilage, laryngeal mask airway, and cervical spine. AI is used to learn the anatomical structure of the supraglottic airway in patients of different body types to establish an AI visual model and construct upper respiratory tract anatomical images. 3.2. According to literature reports, the optimal laryngeal mask position is based on Cormack-Lehane grading labels, and the image features are scored using a logistic regression model to determine the laryngeal mask position (grade 0 is the optimal position, see...). Figure 6 ): 1) Figure 6 A: 3D-CTR horizontal image after laryngeal mask airway insertion under direct oral visualization.

[0034] Using the dashed line of the sagittal section of the glottis midline as the standard, the degree of elevation of the arytenoid arch (interary notch) is graded from 0 to 3: (Grade 0: horizontal to the cricoid arch; Grade 1: elevation of the lower 1 / 3 of the cricoid arch to the glottis midline; Grade 2: elevation of the middle 1 / 3 of the cricoid arch to the glottis midline; Grade 3: elevation of the upper 1 / 3 of the cricoid arch to the glottis midline).

[0035] 2) Figure 6 B: 3D-CTR coronal image after laryngeal mask airway insertion under direct oral visualization.

[0036] A laryngeal mask bag tip located between the upper and lower edges of the sixth cervical vertebra is grade 0; located between the upper and lower edges of the fifth or seventh cervical vertebra is grade 1; and located between the upper and lower edges of the fourth cervical vertebra is grade 2.

[0037] 3.3. Using the above upper respiratory tract anatomical images and based on the Cormack-Lehane classification of the optimal laryngeal mask position, the parameters of each part of the optimal laryngeal mask position were measured: Parameter set definition (corresponding one-to-one with the laryngeal mask structure) In this embodiment, the laryngeal mask structural parameters are defined as a set. This is used for subsequent 3D modeling.

[0038] The parameter set includes at least the following three categories: Stem (ventilation tube) parameters: distance between the corners of the patient's mouth, mouth opening, and distance from the incisors to the base of the tongue. Stem: Cylinder radius; :Stem length (can be estimated from a photo or a standard length can be used and subsequent cutting / adaptation is allowed).

[0039] 2) Basic geometric parameters of the CUFF: The inner and outer rings of the CUFF are concentric teardrop-shaped ellipses. : CUFF outer ring long axis (distance between the base of the tongue and the upper esophageal sphincter); : CUFF inner ring long axis (distance between the epiglottis and the interarytenoid notch); : CUFF outer ring short axis (distance between the lateral edges of the throat); : CUFF inner ring short axis (the distance between the walls of the aryepiglottic ring in a patient); CUFF front and rear diameters (estimated in photo-only mode, with a safety margin to be introduced later).

[0040] 3) Link (connector segment) parameter : The long axis at the connection between the Link and the CUFF (aligned with the long axis direction of the inner ring of the CUFF); The angle between CUFF and Stem (reflecting the difference between head and neck position and insertion direction); Link inner height; :Link outer height.

[0041] Among them, parameters and The orientation alignment is used to ensure that the position of the Link connected to the CUFF is consistent with the longitudinal axis of the patient's pharynx; parameters This describes the tilting transition characteristics of a Link, where the inner and outer sides of the Link are at different heights, thus better reflecting the asymmetry of real anatomical space.

[0042] 4) (Added shape parameter description) To describe the thickness non-uniformity and teardrop / leaf-shaped cross-section of the CUFF in actual bonding, the parameter set also includes a small number of shape parameters: : Thickness non-uniformity range (intensity of thickness difference). Main direction of thickness (thickest / thinnest direction); : Flattening strength (the degree of tearing caused by adhesion / compression); : Pressing the center direction; : Compression coverage area (compression of the angular width of the circumferential coverage).

[0043] 5) To achieve continuous fusion of Link and CUFF, the parameter set also includes: : Transition length (the length of the loft from the end of the Link to the beginning of the CUFF).

[0044] The newly added parameters can be output by the mapping model; when the model confidence is insufficient, a preset template can also be used, and restrictions and corrections can be made in combination with a safety margin.

[0045] 6) AI Mapping and Uncertainty Output: Using the trained mapping model external features Mapped to parameter estimates: in These are the initial estimates of the parameters. These are the uncertainty indicators or confidence levels corresponding to each parameter, used for subsequent safety margin adjustments.

[0046] 7) Geometric constraints and safety margin adjustments: right By applying geometric constraints, manufacturing constraints, and safety margins, the final parameters used for modeling are obtained. .

[0047] The constraints include, but are not limited to: The outer ring is larger than the inner ring: , ; Link skew consistency: ; Connection direction consistency: Long axis at Link connection With CUFF inner ring long axis Orientation alignment; The parameter value is within the printable range: for example , wait; According to uncertainty Set a safety margin: for example, conservatively increase the size of dimensions that might be too small and cause leakage, and impose an upper limit constraint on dimensions that might be too large and cause compression, thereby outputting dimensions that meet the safety boundary. .

[0048] The above steps ensure that individualized parameters that meet manufacturing and usage requirements can still be generated even in the absence of internal images.

[0049] 8) Bridging section: Parameters → Geometric modeling (use the output parameters of the photo as subsequent input) Complete the parameter set After generation and constraint correction, the system enters the geometric modeling stage: with As input, construct a parametric model of "non-uniform thickness + water droplet cross-section" for CUFF, and construct an integrated transition lofting model for Link–CUFF.

[0050] The geometric modeling process ensures that the generated entities satisfy both watertightness and continuity (at least [number missing] connections). The requirement of tangential continuity allows for direct export of 3D printing file formats (STL / 3MF).

[0051] 4. First, treat the laryngeal mask as a "regular body" and use calculus algorithms to construct a preliminary function.

[0052] Specifically, such as Figure 1As shown, the individualized laryngeal mask is divided into four parts: the ventilation tube part Stem (purple), the ventilation mask part CUFF (red), the connecting section part Link (green), and the ventilator connector (blue). The length, radius, diameter, and other parameters of the first three parts are determined according to the anatomical structure of the supraglottic respiratory tract constructed by the AI ​​visual model.

[0053] 4.1. Stem (e.g.) Figure 2 (as shown) Cross-sectional area of ​​the Stem cylinder: S0 = π·r 2 Circumference of the cross-section of the Stem cylinder: L0 = 2π·r Volume of Stem cylinder: V0 = π·r 2 ·h 4.2. CUFF (e.g.) Figure 3 , Figure 4 (as shown) CUFF volume: Using Pappus's center of mass theorem: ① Establish three coordinate axes: x, y, and z; ② The red circle is obtained by translating the center of the CUFF along the X-axis by a distance b, and its radius is 1 / 2j, b = 1 / 2(de); ③ The CUFF ring is formed by rotating the red circle around the Z-axis by 360°n in equal parts, and the distance of each rotation is set as dy, dy = (2·π·b) / n.

[0054] 4.3. Link (e.g.) Figure 5 (as shown) α angle: according to Figure 6 A, α is the angle between Link and CUFF when the interarytenoid notch is raised to grade 0 at the level of the cricoid cartilage arch, where the lengths of m and h' are related to the angle α.

[0055] Link volume: V3 = π·r 2 ·H.

[0056] 5. For CUFF: Transform the CUFF into an irregular body (non-uniform thickness + teardrop-shaped cross-section, P×→CUFF geometry), and derive the formula for calculating the irregular body CUFF using the principles of AI combined with calculus: A. For CUFF: Uneven thickness + teardrop cross-section modeling ( →CUFF Geometry) A1: Definition of Circumferential Angle and Degeneracy Consistency Define the cuff circumferential angle This is used to describe the thickness and flattening variations of a CUFF along different circumferential directions.

[0057] make in This indicates the radius of the CUFF sweep centerline.

[0058] A2: Circumferential non-uniformity function of thickness / front-to-back diameter The original constant parameter Extended to a function that varies with the circumferential direction. .

[0059] Introduction: (Non-uniform amplitude) (Main direction), definition: ; when It degenerates to the original constant thickness parameter. .

[0060] in Control the degree of unevenness in thickness. The "direction of compression of the patient's anatomical structure" is encoded into the model.

[0061] A3: Water Droplet Flattening Template and Correction Factor To represent the teardrop / blade-like cross-section caused by localized compression, a flattening strength is introduced. , compression center direction With coverage And it adopts a segmented definition: Introducing a water droplet correction factor (area attenuation factor): satisfy ,and When it increases Reduced (the cross-sectional area becomes smaller, making it closer to a teardrop shape).

[0062] A4: Circumferential cross-sectional area and circumferential variation Considering the existence of inner and outer rings in the CUFF, while maintaining the original meaning of the measurement, the cross-sectional area of ​​the reference annular ellipse is defined as follows: (corresponding to the area of ​​the outer ellipse) Subtract the area of ​​the inner ellipse ).

[0063] Superimposing "uneven thickness" and "droplet flattening" onto the circumferential cross-sectional area function: in It is a non-negative constant, preferably taken as... or It is used to characterize the sensitivity of cross-sectional area to thickness.

[0064] A5: CUFF size The volume of a cuff is essentially an integral of a sweep along a circle: The engineering implementation can be achieved by summing equal parts: when and hour, The model degenerates into a "uniform elliptical ring", thus maintaining consistency with the original scheme.

[0065] 6.B. CUFF–Link Integrated Modeling ( →Continuous entity) To ensure geometric continuity and allow for solid enclosure, the connection between Link and CUFF uses a transition lofting fusion.

[0066] Introducing the length of the transition section (Can be output by the model or given by a preset rule), used to control the loft length from the end of the Link to the start of the CUFF.

[0067] B1: Link end section (connected to Stem) Link can be viewed as a radius Based on the channel, the cross-sectional area at the end is taken as: B2: CUFF starting section (annular teardrop section consistent with section A) To incorporate the actual fit of the CUFF into the connection calculation, the connection direction between the Link and the CUFF is defined as follows: .

[0068] Preferably, it can be taken (i.e., the connection direction is consistent with the direction of the compression center), or determined by the main direction of compression inferred from the photograph.

[0069] The initial cross-sectional area of ​​the CUFF is then taken as the value of the circumferential cross-sectional area defined in Section A in that direction: B3: Transition section volume (used for estimation and constraints) The volume of the transition section is approximated using a prism: The area of ​​the middle cross section It can be obtained by linear interpolation of the Link end section and the CUFF end section; preferably, it can be taken as follows: Alternatively, a more accurate approximation can be obtained by interpolating geometric parameters and then calculating the area.

[0070] B4: Link–CUFF Integrated Geometry Generation To generate a single, 3D-printable solid object, define: Link upper section : Circular cross-section, radius is The normal direction is aligned with the Stem axis; Link lower end connection section Elliptical or quasi-elliptical cross section, with major axis as... And the direction of the major axis is the same as the major axis of the inner ring of the CUFF. Alignment; minor axis is preferred. Alternatively, the thickness can be taken as the local thickness at the connection direction. The function is used to ensure the continuity of the internal cavity and the structural strength at the connection point; Link tilt attitude determined by the included angle Controlled, and by and The inner and outer sides are constrained to have different heights, thus forming a sloping transition shape.

[0071] exist and Several intermediate sections are arranged between them. The intermediate section is obtained by interpolating the dimensional and shape parameters, and the outer and inner surfaces of the Link are generated using lofting. Tangential continuity constraints are applied during the lofting process to ensure that the Link achieves at least tangential continuity at its connection points with the Stem and CUFF. This avoids steps and abrupt changes in curvature.

[0072] Then, the sectional constraints of the Link termination end and the CUFF at the corresponding connection position are aligned, and continuous fusion (including but not limited to Boolean union, shared boundary loft fusion, etc.) is performed to finally obtain a single closed entity of Stem–Link–CUFF.

[0073] Optionally, a rounded transition can be provided at the connection to further improve sealing performance and durability.

[0074] 7. Overall volume and exported print file Under the integrated model, the total volume can be written as: in: We can first retain the original cylindrical approximation (e.g.) (or obtained directly from the lofting entity in the implementation); As given in Section A (thickness non-uniformity + water droplet correction); The volume of the transition section is given in Section B.

[0075] The aforementioned continuous, integrated closed solid can be exported as an STL / 3MF mesh file for 3D printing manufacturing; and a parameter report (including...) can be output. (and shape template parameters) to support traceability and reproduction.

[0076] The above embodiments are merely preferred embodiments of the present invention. Without departing from the spirit of the present invention, those skilled in the art can make equivalent substitutions or modifications to the order of steps, module combinations, or specific implementation methods, all of which should fall within the protection scope of the present invention.

Claims

1. A method for individualized laryngeal mask fabrication based on an AI visual model, characterized in that, Includes the following steps: Input and size calibration steps: Acquire multi-view photographs of the patient's head and neck; convert the pixel scale in the photographs to the actual size through calibration to obtain the scale factor of each photograph; External morphological feature extraction steps: Perform artificial intelligence reasoning on multi-view photos to extract external anatomical points, and quantize the anatomical points and contours into external morphological feature vectors x; Parameter set generation steps: AI visual model is established by learning the anatomical structure of the supraglottic airway and the position of the laryngeal mask from 3D-CT images of patients with different body types. The external morphological feature vector x is analyzed by the AI ​​visual model to obtain the parameters of each part of the laryngeal mask with the optimal position of the patient's supraglottic airway anatomical structure and laryngeal mask. The parameter set definition steps are as follows: Define the laryngeal mask structure parameters as a set P, including ventilation tube parameters, basic bag geometry parameters, and connector parameters; then utilize the mapping model of the trained AI vision model. The external morphological feature vector x is mapped to parameter estimates: ,in These are the initial estimates of the parameters. For each parameter, there is an uncertainty index or confidence level. Geometric constraint and safety margin correction steps: By applying geometric constraints, manufacturing constraints, and safety margins, the parameters for the final model are obtained. ; Geometric modeling steps: Based on A functional algorithm model for the connection of the ventilation tube, connecting section and bag of the laryngeal mask was constructed, and a file suitable for 3D printing was exported to print the finished laryngeal mask.

2. The personalized laryngeal mask fabrication method based on an AI visual model according to claim 1, characterized in that, In the geometric modeling step, the functions of the function algorithm model include the geometric parameter formula algorithms for the ventilation duct, connecting section, diaphragm, and integrated model.

3. The personalized laryngeal mask fabrication method based on an AI visual model according to claim 2, characterized in that, The algorithm for the geometric parameters of the ventilation duct includes: Cross-sectional area of ​​cylinder: S0 = π·r 2 Circumference of the cylinder's cross-section: L0 = 2πr; Volume of the cylinder: V0 = πr 2 ·h, where r is the radius of the cylinder and h is the length of the ventilation duct.

4. The personalized laryngeal mask fabrication method based on an AI visual model according to claim 2, characterized in that, The geometric formula algorithm for the diaphragm includes: Thickness non-uniformity function: ; Water droplet flattening correction function: ; Circular cross-sectional area function: ; Integral function of volume: ; The function for summing the equal parts of a volume: ; in, Indicates the circumferential angle ∈[0,2π], Indicates the principal direction of thickness. Indicates non-uniform amplitude. Represents the original constant parameter. The direction and angle of compression It is a non-negative constant. Represents the cross-sectional area of ​​the reference annular ellipse. Indicates the radius of the center line of the hood sweep. A small increment in the circumferential angle, The circumferential angle corresponding to the i-th segment.

5. The personalized laryngeal mask fabrication method based on an AI visual model according to claim 2, characterized in that, The algorithm for the geometric parameter formula of the connecting segment includes: Connecting segment volume function: ; Transition segment volume function: ; in Indicates the length of the transition section. Indicates the cross-sectional area at the end. This represents the cross-sectional area at the starting end of the transition section. This represents the cross-sectional area at the middle of the transition section. This represents the radius of the cylinder. This indicates the length of the connecting segment; the transition segment is used to connect the connecting segment to the diaphragm.

6. The personalized laryngeal mask fabrication method based on an AI visual model according to claim 2, characterized in that, The algorithm for the geometric parameter formula of the integrated model includes: Overall volume: .

7. The personalized laryngeal mask fabrication method based on an AI visual model according to claim 1, characterized in that, The multi-view photographs include at least a frontal view, a side view, and an upward-facing view.

8. The personalized laryngeal mask fabrication method based on an AI visual model according to claim 1, characterized in that, In the external morphological feature extraction step, the external anatomical points include at least the chin point, mandibular angle point, mandibular border contour, thyroid cartilage region, suprasternal notch projection point, and oral fissure boundary point.

9. The personalized laryngeal mask fabrication method based on an AI visual model according to claim 1, characterized in that, In the input and dimension calibration steps, the calibration method includes at least one of the following three: Take a picture of a reference ruler of known length placed in the shooting area; Use a mobile terminal with ranging capabilities to obtain the shooting distance and perform scale restoration; Statistical dimensions of the external anatomy were used as weak calibration.

10. The personalized laryngeal mask fabrication method based on an AI visual model according to claim 1, characterized in that, The anatomical structures of the supraglottic airway include the maxilla, mandible, oral cavity, hard palate, soft palate, pharynx, larynx, epiglottis, base of tongue, cricoid cartilage, thyroid cartilage, arytenoid cartilage, and glottis.

Images