Three-dimensional reconstruction system and training model for needle insertion route in vitreous injection training

Abstract

The present application relates to the technical field of teaching training apparatus, and discloses a needle insertion route three-dimensional reconstruction system and a training model in vitreous injection training, which comprises: three-layer grid conductive planes arranged in a simulated eyeball module, each layer being composed of a plurality of independently addressable conductive units; a detection and collection module connected with the conductive units and collecting electrical signals generated when a metal injection needle successively contacts each layer of conductive units during puncture, to generate a three-dimensional coordinate point set; a main controller; and an artificial intelligence model deployed in the main controller, which is used for curve fitting of the three-dimensional coordinate point set to reconstruct a three-dimensional advancing route of the needle insertion route, calculate a needle insertion angle and a needle insertion depth, and compare the needle insertion angle and the needle insertion depth with preset standard parameters to generate an operation score. The present application can realize three-dimensional reconstruction and quantitative evaluation of the needle insertion route without relying on external cameras, VR devices or positioning devices, has low hardware cost and a simple system, and provides objective and accurate feedback basis for vitreous injection skill training.

Description

Technical Field

[0001] This invention relates to the field of teaching and training equipment technology, specifically a three-dimensional reconstruction system and training model for needle insertion path in vitreous injection training. Background Technology

[0002] Intravitreal injection is a primary method of intraocular drug administration. During the procedure, the needle must be inserted at a specific location behind the limbus, with strict control over the angle and depth to avoid critical structures such as the lens and retina. To improve operational safety, various intraocular injection training models have emerged. Some models use conductive rubber and needle puncture sensors to achieve simple judgments of needle insertion correctness, while others rely on virtual reality devices and external sensors for 3D positioning and display. However, traditional training models can only provide binary prompts of correct or incorrect insertion, failing to record quantitative parameters such as the actual trajectory of the needle tip inside the eyeball, insertion angle, and depth. While VR-based systems can achieve 3D visualization, they require high-precision sensors, external rendering machines, and force feedback devices, making them complex, costly, and demanding significant operator experience. Therefore, how to achieve 3D reconstruction and quantitative evaluation of the needle insertion path during intravitreal injection training at low cost and without relying on external vision or positioning equipment is a pressing technical problem that needs to be solved. Summary of the Invention

[0003] Therefore, it is necessary to provide a three-dimensional reconstruction system and training model for the needle insertion path in vitreous injection training that can realize three-dimensional reconstruction and quantitative evaluation of the needle insertion path without the need for external positioning equipment, in order to address the above-mentioned technical problems.

[0004] In a first aspect, the present invention discloses a three-dimensional reconstruction system and training model for needle insertion path in intravitreal injection training, the system comprising: Three layers of mesh conductive planes are set inside the simulated eyeball module. Each layer of mesh conductive plane includes multiple conductive units spliced ​​together. Adjacent conductive units are separated by insulating gaps, and each conductive unit is led out through an independent wire. The detection and acquisition module is connected to each of the conductive units of the three-layer grid conductive plane. It is used to detect the electrical signals generated when the metal injection needle comes into contact with the conductive units on each layer of the conductive plane in sequence during the puncture process, and to generate a three-dimensional coordinate point set based on the layer number and two-dimensional coordinates within the triggered conductive unit. Main controller; An artificial intelligence model is deployed on the main controller. The artificial intelligence model receives the set of three-dimensional coordinate points, performs curve fitting on the set of three-dimensional coordinate points to reconstruct the three-dimensional travel path of the metal injection needle in the simulated eyeball module, calculates the needle insertion angle and needle insertion depth based on the reconstructed three-dimensional travel path, and compares the calculated needle insertion angle and needle insertion depth with a preset standard parameter range to generate an operation score.

[0005] In one embodiment, the three-layer mesh conductive plane includes: a first mesh layer located on the inner surface of the sclera, a second mesh layer located between the first mesh layer and the center of the eyeball, and a third mesh layer passing through the center point of the eyeball.

[0006] In one embodiment, the detection and acquisition module includes a multiplexed acquisition unit, which sequentially detects the on / off state of all conductive units on each conductive plane at a scanning frequency of greater than or equal to 1 kHz, and outputs the layer number, two-dimensional coordinates within the layer, and trigger timestamp of the triggered conductive unit as the three-dimensional coordinate point set.

[0007] In one embodiment, the artificial intelligence model uses a cubic spline interpolation algorithm to perform curve fitting on the three-dimensional coordinate point set in order to reconstruct the three-dimensional travel path of the metal injection needle inside the eyeball.

[0008] In one embodiment, the artificial intelligence model uses a weighted scoring algorithm to output a comprehensive score of 0 to 100 and generates a learning curve, which is used to record the temporal changes of the score over multiple operations.

[0009] Secondly, this application discloses a vitreous injection training model, the model comprising: A simulated head model, wherein the simulated head model is provided with an eye socket structure; A simulated eyeball module is installed within the orbital structure. The simulated eyeball module includes an outer shell layer and an internal structure. The outer shell layer includes a conjunctival conductive layer and a scleral conductive layer from the outside to the inside. The scleral conductive layer is divided into multiple electrically isolated conductive regions along the limbus towards the posterior pole of the eyeball. The conductive regions include a near injection point area, a correct injection area, and a far injection point area. The internal structure includes simulated components of multiple key anatomical structures. The surface of each simulated component of a key anatomical structure is provided with an independent conductive film. And the three-dimensional reconstruction system for the needle insertion path as described above; The conjunctival conductive layer is connected to the ground or power supply terminal of the main controller. Each conductive region of the conjunctival conductive layer and the scleral conductive layer is connected to the first set of general-purpose input / output ports of the main controller of the needle insertion path three-dimensional reconstruction system via wires. The conductive films of the key anatomical structures are connected to the second set of general-purpose input / output ports of the main controller via independent wires. The main controller determines the needle insertion point position and identifies the key anatomical structures that have been accidentally touched by detecting the level changes of each general-purpose input / output port.

[0010] In one embodiment, the device further includes an eyeball rotation mechanism, which includes a universal joint mechanism and a drive assembly. The simulated eyeball module is fixed to the universal joint mechanism, and the drive assembly drives the universal joint mechanism to rotate the simulated eyeball module to the injection position.

[0011] In one embodiment, the conjunctival conductive layer is made of a conductive silicone rubber film with a Shore hardness of 10HA to 15HA and a thickness of 0.3mm to 0.5mm; the scleral conductive layer is made of conductive silicone rubber with a Shore hardness of 30HA to 45HA and a thickness of 0.8mm to 1.2mm; and a lubricating gel layer with a thickness of 0.08mm to 0.12mm is disposed between the conjunctival conductive layer and the scleral conductive layer.

[0012] In one embodiment, both the conjunctival conductive layer and the scleral conductive layer are made of a self-healing conductive silicone rubber material based on a hydrogen bond crosslinking network.

[0013] In one embodiment, the plurality of key anatomical structures include the lens, suspensory ligaments, iridociliary body, inner corneal layer, and posterior retina.

[0014] The aforementioned 3D reconstruction system and training model for the needle insertion path in vitreous injection training utilizes a three-layer conductive mesh plane embedded within a simulated eyeball module. In conjunction with a detection and acquisition module, it sequentially collects the 3D coordinate points of the needle tip contacting each conductive unit during the insertion of the metal injection needle. An artificial intelligence model deployed on the main controller then performs curve fitting on this point set to reconstruct the 3D path of the needle tip within the eyeball. Based on this, it automatically calculates the insertion angle and depth, compares them with preset standard parameters, and generates an operation score. Simultaneously, by connecting the three zones of the conjunctival conductive layer and the scleral conductive layer, as well as the conductive films of key anatomical structures, to the general-purpose input / output ports of the main controller, the main controller detects changes in voltage levels to determine the needle insertion point location in real time and identify any mis-touched key anatomical structures. This achieves data acquisition, processing, and real-time feedback without relying on cameras, VR headsets, external positioning devices, or force feedback devices. It only requires a conductive mesh layer and conductive zone structure embedded within the eyeball, along with a conventional microcontroller, reducing hardware costs and system complexity. The visualization of the three-dimensional movement path, the quantitative output of angle and depth, the real-time judgment of the needle insertion point, and the warning of accidental touch on key structures provide trainees with objective feedback throughout the entire process from needle insertion positioning and intraocular movement to post-operation evaluation, which helps them to correct errors independently and improve their skills. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, 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.

[0016] Figure 1 This is a structural block diagram of a three-dimensional reconstruction system for the needle insertion path during intravitreal injection training in one embodiment. Figure 2 This is a structural block diagram of a vitreous injection training model in one embodiment. Detailed Implementation

[0017] To facilitate understanding of the technical solutions provided in the embodiments of this application, the background technology involved in the embodiments of this application will be described below.

[0018] Intravitreal injection is a core route of drug delivery for treating fundus diseases such as macular edema, retinal vein occlusion, and neovascular ophthalmopathy. This procedure requires the operator to perform the puncture within a narrow area of ​​3.5 mm to 4.0 mm behind the limbus. The angle and depth of needle insertion, as well as its relative position to important structures such as the lens, retina, and ciliary body, must be precisely controlled. Any deviation can lead to serious complications such as intraocular hemorrhage, retinal detachment, cataracts, or endophthalmitis. Therefore, systematic, precise, and quantifiable skills training for operators is crucial in the preclinical stage.

[0019] Currently available intraocular injection training models can be mainly divided into three categories. The first category is purely mechanical structural models, such as simple devices that adjust the eyeball position using springs or sliders. These models can only train basic hand-eye coordination and cannot provide any feedback on the needle insertion angle, depth, or whether critical structures have been accidentally touched. The training effect relies entirely on the teacher's on-site observation and guidance, resulting in low efficiency and high subjectivity. The second category is electronic models based on conductive rubber and needle puncture sensors. These models can determine whether the needle insertion position is correct or whether secondary damage has occurred when the needle tip contacts a specific conductive layer, and provide binary prompts of correct or incorrect via indicator lights. However, this type of model can only indicate whether the operation result is correct; it cannot display the actual trajectory of the needle tip after entering the eyeball, nor can it quantify the specific deviations in the needle insertion angle and depth. Trainees cannot self-check the cause of errors through the system and still need to rely on the teacher to correct them one by one. The third category is highly realistic systems combining virtual reality and external positioning devices. These systems can track the position of the injection needle through sensors and display the needle tip position in real time in a virtual three-dimensional eyeball model. They can even simulate the resistance of different tissues to the needle tip. However, such systems require high-precision electromagnetic positioners, multiple rendering computers, force feedback motors, or head-mounted display devices, resulting in extremely high hardware costs, complex system architecture, and significant challenges in calibration and maintenance. Furthermore, they demand a high level of experience from operators, making it difficult to promote and popularize them in grassroots teaching units and large-scale training scenarios.

[0020] In summary, the existing technology has significant shortcomings in the following two core issues: First, it cannot automatically acquire quantitative parameters such as the three-dimensional trajectory of the injection needle inside the eyeball, the needle insertion angle, and the needle insertion depth without relying on external positioning or vision devices; Second, it lacks a low-cost, integrated feedback mechanism that can simultaneously achieve real-time judgment of the needle insertion point, early warning of accidental contact with key structures, and reconstruction and scoring of the three-dimensional route after the operation.

[0021] Therefore, this application provides a three-dimensional reconstruction system and training model for needle insertion path in vitreous injection training.

[0022] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present application.

[0023] Firstly, in vitreous injection training, the operator cannot obtain quantitative parameters such as the three-dimensional trajectory, insertion angle, and insertion depth of the injection needle after entering the eyeball, resulting in a lack of objective evaluation criteria for the training process. Traditional methods rely on external cameras or electromagnetic positioning devices for three-dimensional tracking, but such systems are complex, costly, and difficult to integrate into physical training models. Therefore, in one embodiment, such as... Figure 1 As shown, a three-dimensional reconstruction system for the needle insertion path in intravitreal injection training is provided. The system includes: Three layers of grid conductive planes are set inside the simulated eyeball module. Each layer of grid conductive plane includes multiple conductive units spliced ​​together. Adjacent conductive units are separated by insulating gaps, and each conductive unit is led out through an independent wire. The detection and acquisition module is connected to each conductive unit of the three-layer grid conductive plane. It is used to detect the electrical signals generated when the metal injection needle comes into contact with the conductive units on each layer of conductive plane in sequence during the puncture process, and to generate a three-dimensional coordinate point set based on the layer number and two-dimensional coordinates within the triggered conductive unit. Main controller; The artificial intelligence model is deployed on the main controller. The artificial intelligence model receives a set of three-dimensional coordinate points, performs curve fitting on the set of three-dimensional coordinate points to reconstruct the three-dimensional travel path of the metal injection needle in the simulated eyeball module, calculates the needle insertion angle and depth based on the reconstructed three-dimensional travel path, and compares the calculated needle insertion angle and depth with the preset standard parameter range to generate an operation score.

[0024] The simulated eyeball module is a training model component that mimics a real human eyeball. It is approximately spherical in shape and hollow inside. Inside the simulated eyeball module, three layers of conductive mesh planes are arranged along the anterior-posterior axis of the eyeball. Each conductive mesh plane is manufactured using flexible printed circuit board technology, with a polyimide film substrate. The surface is etched to form an array of conductive units. Each conductive unit is square, with a side length of 0.5 mm, and the insulating gap between adjacent conductive units is 0.05 mm wide. The back of each conductive unit is led to a connector at the edge of the substrate via an independent printed wire.

[0025] The detection and acquisition module is electrically connected to all conductive units via independent wires. The module includes a multiplexed acquisition unit, such as a 74HC4067 analog multiplexer chip. The main controller uses address lines to control the multiplexer to sequentially select each conductive unit, connecting the currently selected unit to the detection loop. In the detection loop, the common reference terminal is connected to the main controller's ground terminal. The currently selected conductive unit is connected to the main controller's general-purpose input / output port via the multiplexer's common output terminal. This port is configured for input mode and has its internal pull-up resistor enabled. When the metal injection needle contacts a conductive unit, the needle short-circuits the conductive unit to the common reference terminal, causing the general-purpose input / output port level to change from high to low. The main controller detects this level change and determines that the conductive unit has been triggered. The main controller records the layer number (1, 2, or 3) of the conductive unit, as well as its row and column numbers within that layer. An internal timer records the trigger time as a timestamp. Multiple triggered points are arranged in a three-dimensional coordinate point set according to their timestamps.

[0026] The main controller uses a 32-bit ARM architecture microcontroller chip (such as the STM32F103 series), with built-in random access memory and flash memory to store the three-dimensional coordinate point set and preset standard parameter range.

[0027] The artificial intelligence model is deployed as a software program in the flash memory of the main controller. The model first preprocesses the set of three-dimensional coordinate points: mapping depth coordinates according to layer numbers (pre-calibrated Z1, Z2, Z3), converting row and column numbers into spatial lateral coordinates (X = (column number - C0) × D, Y = (row number - R0) × D, where D is the center-to-center spacing of conductive units, and C0 and R0 are the row and column numbers of the reference center point), resulting in a sequence of three-dimensional spatial coordinate points. Then, using a cubic spline interpolation algorithm, with depth Z as the independent variable, piecewise cubic polynomials are constructed for X and Y respectively, fitting a continuous and smooth three-dimensional spatial curve, which represents the needle tip's path. The needle insertion angle is defined as the angle between the tangent vector at the curve's starting point and the normal vector of the scleral surface, calculated using the vector angle formula; the needle insertion depth is defined as the arc length from the curve's starting point to its ending point, obtained by numerically integrating the derivative of the arc length. Finally, the calculated needle insertion angle and depth are compared with the preset standard range: if both the angle and depth are within the standard range, the score is 100 points; otherwise, points are deducted according to the linear rule (4 points are deducted for every 1 degree deviation in angle, and 20 points are deducted for every 1 millimeter deviation in depth, up to 0 points). The scoring results are output to the display screen.

[0028] Through the collaborative work of the three-layer mesh conductive plane, detection and acquisition module, and artificial intelligence model, this system automatically acquires the three-dimensional discrete trajectory points of the injection needle and reconstructs the continuous travel route without relying on external cameras or positioning devices. It realizes the quantitative measurement of needle insertion angle and depth and the automatic generation of operation scores, providing trainees with objective and accurate training feedback.

[0029] To ensure that the collected discrete points can cover the critical path of the needle tip from entering the vitreous cavity to passing through the center of the eyeball, in one embodiment, the three-layer mesh conductive plane includes: a first mesh layer located on the inner surface of the sclera, a second mesh layer located between the first mesh layer and the center of the eyeball, and a third mesh layer passing through the very center of the eyeball.

[0030] The first mesh layer is positioned on the inner surface of the sclera in the simulated eyeball module. Since the injection needle enters the vitreous cavity immediately after piercing the sclera, the first mesh layer is installed flush against the inner wall of the sclera, its curvature matching the inner surface of the sclera, to record the position coordinates of the needle tip immediately upon entering the vitreous cavity. The second mesh layer is positioned at an intermediate depth between the first mesh layer and the center of the eyeball, for example, approximately half the radius of the eyeball. The third mesh layer is positioned on a plane passing through the exact center of the eyeball. The three mesh layers are arranged sequentially along the anterior-posterior axis of the eyeball. By setting three mesh layers at different depths, the system can acquire the two-dimensional coordinates of the needle tip at three different depth positions, providing at least three discrete sampling points for subsequent curve fitting. Three sampling points are the minimum necessary number, uniquely determining the control points for a quadratic or cubic spline curve, thus controlling the total number of conductive units and system complexity while ensuring reconstruction accuracy.

[0031] If each conductive unit is configured with a separate detection channel, the hardware cost will increase dramatically. Therefore, in one embodiment, the detection and acquisition module includes a multiplexed acquisition unit. The multiplexed acquisition unit sequentially detects the on / off state of all conductive units on each conductive plane at a scanning frequency of greater than or equal to 1 kHz, and outputs the layer number of the triggered conductive unit, the two-dimensional coordinates within the layer, and the trigger timestamp as a three-dimensional coordinate point set.

[0032] Specifically, each layer of the grid's conductive plane contains approximately 1800 conductive units. The multiplexing acquisition unit uses multiple cascaded 74HC4067 analog multiplexer chips. Each 74HC4067 has 16 input channels and 1 common output channel, with the currently active channel selected via 4 address lines. The common output of all chips is connected to the same detection port of the main controller, and the address lines and chip select lines are controlled by the main controller's general purpose input / output port. During scanning, the main controller sequentially selects each conductive unit and detects its on / off state. To meet real-time requirements, a hierarchical parallel architecture can be adopted: the three grid layers are independently configured with multiplexing acquisition units, operating in parallel, increasing the scanning frequency to over 10kHz; or a hybrid architecture of interrupt triggering and parallel scanning can be used, where when the probe tip triggers a conductive unit, the address encoder immediately outputs the unit coordinates and triggers an external interrupt, with a response time of less than 10 microseconds. The multiplexing acquisition unit outputs the layer number, row number, column number, and timestamp of the triggered unit as a set of three-dimensional coordinate points. Employing multiplexing technology, only a few general-purpose input / output ports are needed to detect a large number of conductive units, reducing hardware costs and wiring complexity. A scanning frequency of 1 kHz or higher ensures that the needle tip can be captured in time during rapid puncture, avoiding missed detections.

[0033] A suitable interpolation algorithm is needed to ensure the smoothness and accuracy of the reconstructed trajectory. Linear interpolation can produce sharp angles, while high-order polynomial interpolation is prone to edge oscillations. Therefore, in one embodiment, the artificial intelligence model uses a cubic spline interpolation algorithm to perform curve fitting on the three-dimensional coordinate point set to reconstruct the three-dimensional travel path of the metal injection needle inside the eyeball.

[0034] Specifically, let the sequence of three-dimensional spatial coordinate points obtained after preprocessing be... ,in Cubic spline interpolation using depth coordinates As the independent variable, respectively for coordinates and Construct a piecewise cubic polynomial using coordinates. For Coordinates, in each sub-interval Construct a cubic polynomial on (k=1,2,…,m−1): ; The following conditions must be met: , (Through known points); (The first derivative is continuous); (The second derivative is continuous); Boundary conditions are taken as natural spline conditions: , .

[0035] The above conditions constitute a relationship with the coefficients. , , The system of linear equations, which has a tridiagonal matrix form, can be solved efficiently using the Thomas algorithm (the chasing method). The solution process is common knowledge in numerical analysis; the specific algorithm can be found in textbooks on numerical analysis. For... The coordinates are constructed using the exact same steps. The final parametric representation of the three-dimensional travel path is as follows: , , ; The curve exhibits continuous first and second derivatives, accurately reflecting the natural and smooth trajectory of needle puncture, avoiding the sharp angles of linear interpolation and the edge oscillations of higher-order polynomial interpolation. The reconstructed 3D path can be output to a display screen for visualization using a transparent eyeball model and a colored trajectory curve.

[0036] When scoring, the quality of the operation needs to be quantified into a specific score, reflecting the trainee's progress trend. Simple binary judgment cannot meet the training requirements. Therefore, in one embodiment, the artificial intelligence model uses a weighted scoring algorithm to output a comprehensive score from 0 to 100, and generates a learning curve. The learning curve is used to record the temporal changes of the scores of multiple operations.

[0037] Specifically, the weighted scoring algorithm quantifies and scores four evaluation indicators: needle insertion point position deviation (Euclidean distance between the actual needle insertion point and the standard point, in millimeters), needle insertion angle deviation (absolute value of the deviation of the actual angle from the standard range [25°, 35°], in degrees), needle insertion depth deviation (absolute value of the deviation of the actual depth from the standard range [5mm, 7mm], in millimeters), and a critical structure mis-touch flag (0 or 1). The original deviation value of each indicator is mapped to a sub-score of 0 to 100 points through a piecewise linear function: 20 points are deducted for every 1 mm increase in position deviation, 4 points for every 1 degree increase in angle deviation, and 20 points for every 1 mm increase in depth deviation. A mis-touch results in a sub-score of 0 points, otherwise 100 points. The sub-scores are weighted and summed, with typical weights of 0.3 for position, 0.3 for angle, 0.2 for depth, and 0.2 for mis-touch. The weighted sum is the comprehensive score, rounded to the nearest integer. After each operation, the model stores the score and operation sequence number in the non-volatile memory of the main controller. When a user views the learning curve, the main controller plots a line graph with the operation number on the x-axis and the score on the y-axis, displaying statistical information such as the highest score, lowest score, and average score. The aforementioned penalty coefficients and weights can be pre-obtained through supervised learning: professional physicians are invited to perform standard operations, feature indicators and physician scores are recorded, and the optimal parameters are fitted using the least squares method. After being solidified, the model only performs forward computation during runtime. This lightweight model occupies less than 1 kilobyte of storage space, has minimal inference computation, and can run in real-time on a low-cost microcontroller.

[0038] Secondly, in intravitreal injection training models, the operator not only needs to understand the three-dimensional path after needle insertion, but also needs to know in real time whether the needle insertion point is accurate and whether critical structures have been accidentally touched during the puncture. Existing models either lack real-time feedback or rely on complex external devices. Therefore, in one embodiment, such as... Figure 2 As shown, a vitreous injection training model is provided, the model including: A simulated head model, which includes eye socket structures; The simulated eyeball module is installed within the orbital structure. The simulated eyeball module includes an outer shell layer and an internal structure. The outer shell layer includes a conjunctival conductive layer and a scleral conductive layer from the outside to the inside. The scleral conductive layer is divided into multiple electrically isolated conductive regions along the limbus towards the posterior pole of the eyeball. The conductive regions include the near injection point area, the correct injection area, and the far injection point area. The internal structure includes simulated parts of multiple key anatomical structures. The surface of each simulated part of a key anatomical structure is provided with an independent conductive film. And as mentioned above, a three-dimensional reconstruction system for the needle insertion path; The conjunctival conductive layer is connected to the ground or power supply terminal of the main controller. Each conductive area of ​​the conjunctival conductive layer and the scleral conductive layer is connected to the first set of general-purpose input / output ports of the main controller of the needle insertion path three-dimensional reconstruction system via wires. The conductive films of key anatomical structures are connected to the second set of general-purpose input / output ports of the main controller via independent wires. The main controller determines the needle insertion point position and identifies the key anatomical structures that have been accidentally touched by detecting the level changes of each general-purpose input / output port.

[0039] The simulated head model is made of silicone and has the shape of a real human head, with eye sockets in the eye area. The simulated eyeball module is fixed in the eye socket structure by a ball-and-socket joint mount, allowing for omnidirectional rotation.

[0040] The outer shell of the simulated eyeball module has a multi-layered composite structure. The outermost layer is the conjunctival conductive layer, composed of a conductive silicone rubber film with a Shore hardness of 10HA~15HA and a thickness of 0.3~0.5mm. The inner layer is the scleral conductive layer, made of conductive silicone rubber with a Shore hardness of 30HA~45HA and a thickness of 0.8~1.2mm. The scleral conductive layer is divided into three electrically independent annular regions: the near zone (from the limbus to 3.5mm posterior to the limbus), the correct zone (3.5mm to 4.0mm posterior to the limbus), and the far zone (4.0mm posterior to the limbus and beyond to the equator of the eyeball). These three regions are separated by a 0.2mm wide non-conductive silicone rubber insulating strip. A lubricating gel layer with a thickness of 0.08~0.12mm is also placed between the conjunctival and scleral conductive layers, simulating Tenon's fascia, to create a layered sensation of "soft first, then hard, two breakthroughs" during puncture. A 0.1mm thick elastic film is further bonded to the inner side of the conductive layer of the sclera to simulate the feeling of the needle falling into the vitreous cavity after passing through the sclera.

[0041] The internal structure includes simulated components of the lens, suspensory ligament, iris-ciliary body, inner corneal layer, and posterior retina. Each simulated component is coated with an independent conductive film, which is electrically isolated from each other.

[0042] The outer shell of the simulated eye module is detachably connected to its internal structure using a snap-fit ​​mechanism. Specifically, the outer shell has multiple latches on the side closest to the internal structure, and the corresponding positions on the internal structure have multiple slots. The latches and slots are locked together by pressing, and can be separated by pulling in the opposite direction. As an independently replaceable consumable component, the outer shell can be manually removed and a new shell installed after its design lifespan has expired, with the replacement time not exceeding 30 seconds and without affecting the internal structure and circuitry.

[0043] The aforementioned conductive structure is connected to the main controller via wires. The conjunctival conductive layer serves as a common electrode connected to the ground terminal of the main controller. The three regions of the scleral conductive layer are respectively connected to the first set of general-purpose input / output ports (e.g., GPIO0, GPIO1, GPIO2) of the main controller, and the conductive film of each key anatomical structure is respectively connected to the second set of general-purpose input / output ports (e.g., GPIO3~GPIO7). Each port is configured for input mode and internal pull-up resistors are enabled. When the metal injection needle is inserted, the needle body successively contacts the conjunctival conductive layer and a certain region of the sclera (or the conductive film of a key structure), forming a current loop, and the corresponding port level changes from high to low. The main controller determines whether the needle insertion point is in the near zone, the correct zone, or the remote zone based on the port number of the level change, or identifies the specific key structure that has been mistakenly touched, and outputs a prompt or alarm in real time.

[0044] By directly connecting the conjunctival conductive layer, the scleral three-zone conductive layer, and the key structural conductive film to the general-purpose input / output ports of the main controller, this model achieves real-time three-level judgment of the needle insertion point and accurate identification of accidental contact with key structures without adding complex sensors. Together with the needle insertion path 3D reconstruction system, it constitutes a complete training solution from needle insertion positioning and intraocular navigation to operational evaluation. The model's detection and acquisition module and main controller are powered by 5V DC low voltage, supplied by an external power adapter or USB interface, ensuring safe use. The main controller integrates a low-dropout linear regulator to convert 5V to 3.3V for the chip core.

[0045] In real clinical injections, patients need to turn their eyes upwards towards the nose to expose the injection area. Existing models have fixed eye positions and cannot simulate this physiological posture. Therefore, in one embodiment, the model also includes an eye rotation mechanism, which includes a gimbal mechanism and a drive component. The simulated eye module is fixed to the gimbal mechanism, and the drive component drives the gimbal mechanism to rotate the simulated eye module to the injection position.

[0046] Specifically, the universal joint mechanism adopts any existing ball joint structure, with its base fixed inside the simulated head model and its movable end fixedly connected to the rear of the simulated eyeball module. The drive component can use any existing manual knob or micro servo motor. The manual knob drives the universal joint to deflect via a worm gear mechanism, achieving vertical and horizontal angle adjustment of the eyeball. The micro servo motor is automatically controlled by the main controller according to button commands or preset programs, driven by pulse width modulation signals, with an angle control accuracy of more than 1 degree, capable of rotating the eyeball to a standard injection position (e.g., rotating 30 degrees upwards towards the nose). Using this device, the simulated eyeball module can be adjusted to different angles before training to simulate injection scenarios with varying levels of patient cooperation, enhancing the clinical realism of the training.

[0047] To simulate the piercing sensation of a real human eyeball, the hardness and thickness of each conductive layer need to be precisely controlled. Inappropriate hardness or thickness will lead to a distorted piercing sensation. Therefore, in one embodiment, the conjunctival conductive layer is made of a conductive silicone rubber film with a Shore hardness of 10HA–15HA and a thickness of 0.3mm–0.5mm; the scleral conductive layer is made of conductive silicone rubber with a Shore hardness of 30HA–45HA and a thickness of 0.8mm–1.2mm; and a lubricating gel layer with a thickness of 0.08mm–0.12mm is disposed between the conjunctival and scleral conductive layers.

[0048] Specifically, conductive silicone rubber is formed by vulcanizing a silicone rubber matrix with conductive fillers (such as carbon nanotubes, conductive carbon black, or silver nanowires). The lower the Shore hardness value, the softer the material. The conjunctival layer uses low hardness and a thin thickness to simulate the thin and loose characteristics of the real conjunctiva, resulting in low puncture resistance. The scleral layer uses higher hardness and a larger thickness to simulate the density of scleral fibrous tissue, producing a distinct breakthrough sensation during puncture. The lubricating gel layer uses silicone oil or medical-grade lubricating gel, which has a low coefficient of friction, allowing relative sliding between the conjunctival and scleral layers. As the needle tip sequentially passes through the conjunctival layer, lubricating gel layer, and scleral layer, the operator can feel a layered sensation of "soft first, then hard, two breakthroughs," greatly enhancing the realism of the training.

[0049] Repeated punctures can damage the material, affecting conductivity and feel, and shortening the lifespan of the model. Therefore, in one embodiment, both the conjunctival conductive layer and the scleral conductive layer are made of a self-healing conductive silicone rubber material based on a hydrogen-bonded cross-linked network.

[0050] Specifically, self-healing conductive silicone rubber introduces a large number of functional groups (such as urea, amide, or carboxyl groups) capable of forming reversible hydrogen bonds into the polymer matrix. When the puncture needle is removed, the material around the needle hole undergoes hydrogen bond breakage and recombination at room temperature, with dynamic hydrogen bonds re-crosslinking, gradually closing the needle hole and restoring the material's integrity and conductive continuity. The self-healing process typically completes within minutes to tens of minutes, without the need for heating or external reagents. Experiments show that conjunctival and scleral layers made with this material exhibit a resistance change of less than 15% and a peak puncture force change of less than 10% after more than 200 standard injection needle punctures, maintaining good conductivity and puncture feel. The self-healing time is less than or equal to 5 minutes for a single needle hole closure at room temperature. This material can be produced by reacting hydroxyl-terminated polydimethylsiloxane with urea-containing diisocyanates to generate a polysiloxane elastomer containing hydrogen-bonded units, which is then mixed with conductive fillers and vulcanized. The self-healing property allows a single simulated eyeball module to support hundreds of training sessions, extending the model's lifespan and reducing training costs. Specifically, the self-healing conductive silicone rubber material has a self-healing time of ≤5 minutes for a single puncture needle hole closure at room temperature; after 200 punctures, the resistance value shift of the material is ≤15%, and the peak puncture force change is ≤10%.

[0051] To effectively deploy the conductive film, it is necessary to identify which structures are most susceptible to accidental contact during injection. Therefore, in one embodiment, several key anatomical structures include the lens, suspensory ligaments, iris-ciliary body, inner corneal layer, and posterior retina.

[0052] Specifically, the lens, located behind the iris, is the structure most prone to accidental contact when the injection angle is too forward. The suspensory ligament, a ring-shaped fibrous structure connecting the lens and ciliary body, is located close to the needle insertion area. The iridociliary body is located near the limbus and may be touched if the needle is inserted too superficially. The inner corneal layer is located on the inner side of the cornea at the front of the eyeball and may be accidentally touched if the needle is inserted too far forward. The posterior retina is located on the inner side of the posterior wall of the eyeball and may be punctured if the needle is inserted too deeply. Independent conductive films are placed on the surface of the simulation model of these five key structures, each connected to an independent universal input / output port of the main controller. When the needle tip accidentally touches any structure, the main controller immediately identifies it and issues an audible and visual alarm. By covering these most common and dangerous accidental contact scenarios, this training model can effectively cultivate the operator's awareness of structural protection and reduce the risk of complications in real surgery.

[0053] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0054] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A three-dimensional reconstruction system for needle insertion path in intravitreal injection training, characterized in that, The system includes: Three layers of mesh conductive planes are set inside the simulated eyeball module. Each layer of mesh conductive plane includes multiple conductive units spliced ​​together. Adjacent conductive units are separated by insulating gaps, and each conductive unit is led out through an independent wire. The detection and acquisition module is connected to each of the conductive units of the three-layer grid conductive plane. It is used to detect the electrical signals generated when the metal injection needle comes into contact with the conductive units on each layer of the conductive plane in sequence during the puncture process, and to generate a three-dimensional coordinate point set based on the layer number and two-dimensional coordinates within the triggered conductive unit. Main controller; An artificial intelligence model is deployed on the main controller. The artificial intelligence model receives the set of three-dimensional coordinate points, performs curve fitting on the set of three-dimensional coordinate points to reconstruct the three-dimensional travel path of the metal injection needle in the simulated eyeball module, calculates the needle insertion angle and needle insertion depth based on the reconstructed three-dimensional travel path, and compares the calculated needle insertion angle and needle insertion depth with a preset standard parameter range to generate an operation score.

2. The three-dimensional reconstruction system for needle insertion path in intravitreal injection training according to claim 1, characterized in that, The three-layer mesh conductive plane includes: a first mesh layer located on the inner surface of the sclera, a second mesh layer located between the first mesh layer and the center of the eyeball, and a third mesh layer passing through the center point of the eyeball.

3. The three-dimensional reconstruction system for needle insertion path in intravitreal injection training according to claim 1, characterized in that, The detection and acquisition module includes a multiplexed acquisition unit. The multiplexed acquisition unit sequentially detects the on / off state of all conductive units on each conductive plane at a scanning frequency of greater than or equal to 1 kHz, and outputs the layer number, two-dimensional coordinates within the layer, and trigger timestamp of the triggered conductive unit as the three-dimensional coordinate point set.

4. The three-dimensional reconstruction system for needle insertion path in intravitreal injection training according to claim 1, characterized in that, The artificial intelligence model uses a cubic spline interpolation algorithm to fit the three-dimensional coordinate point set to reconstruct the three-dimensional travel path of the metal injection needle inside the eyeball.

5. The three-dimensional reconstruction system for needle insertion path in vitreous injection training according to claim 1 or 4, characterized in that, The artificial intelligence model uses a weighted scoring algorithm to output a comprehensive score of 0 to 100 and generates a learning curve, which is used to record the temporal changes of the score over multiple operations.

6. A vitreous injection training model, characterized in that, The model includes: A simulated head model, wherein the simulated head model is provided with an eye socket structure; A simulated eyeball module is installed within the orbital structure. The simulated eyeball module includes an outer shell layer and an internal structure. The outer shell layer includes a conjunctival conductive layer and a scleral conductive layer from the outside to the inside. The scleral conductive layer is divided into multiple electrically isolated conductive regions along the limbus towards the posterior pole of the eyeball. The conductive regions include a near injection point area, a correct injection area, and a far injection point area. The internal structure includes simulated components of multiple key anatomical structures. The surface of each simulated component of a key anatomical structure is provided with an independent conductive film. And a three-dimensional reconstruction system for needle insertion path as described in any one of claims 1 to 5; The conjunctival conductive layer is connected to the ground or power supply terminal of the main controller. Each conductive region of the conjunctival conductive layer and the scleral conductive layer is connected to the first set of general-purpose input / output ports of the main controller of the needle insertion path three-dimensional reconstruction system via wires. The conductive films of the key anatomical structures are connected to the second set of general-purpose input / output ports of the main controller via independent wires. The main controller determines the needle insertion point position and identifies the key anatomical structures that have been accidentally touched by detecting the level changes of each general-purpose input / output port.

7. The intravitreal injection training model according to claim 6, characterized in that, It also includes an eyeball rotation mechanism, which includes a universal joint mechanism and a drive component. The simulated eyeball module is fixed to the universal joint mechanism, and the drive component drives the universal joint mechanism to rotate the simulated eyeball module to the injection position.

8. The intravitreal injection training model according to claim 6, characterized in that, The conjunctival conductive layer is made of a conductive silicone rubber film with a Shore hardness of 10HA to 15HA and a thickness of 0.3mm to 0.5mm; the scleral conductive layer is made of conductive silicone rubber with a Shore hardness of 30HA to 45HA and a thickness of 0.8mm to 1.2mm; a lubricating gel layer with a thickness of 0.08mm to 0.12mm is disposed between the conjunctival conductive layer and the scleral conductive layer.

9. The intravitreal injection training model according to claim 6, characterized in that, Both the conjunctival conductive layer and the scleral conductive layer are made of self-healing conductive silicone rubber material based on a hydrogen bond cross-linking network.

10. The intravitreal injection training model according to claim 6, characterized in that, The key anatomical structures include the lens, suspensory ligaments, iris-ciliary body, inner corneal layer, and posterior retina.

Images