Portable intelligent constant-temperature eye atomization drug delivery system based on digital twinning and pneumatic self-cleaning

Abstract

The invention discloses a portable intelligent constant-temperature eye atomization drug delivery system based on digital twinning and pneumatic self-cleaning, and relates to the technical field of ophthalmology medical instruments. The system comprises a portable host unit and an end effector worn on the eyes. A micro-channel heating module, a gas-liquid separation buffer cavity and a micropore piezoelectric atomization piece are integrated in the end effector, and a one-way clean gulp valve with a micropore sterilization filter membrane is arranged on the side wall of the buffer cavity. When dosing is finished, the system executes a pneumatic broken bridge self-cleaning program, the precise conveying pump generates reverse negative pressure to open the gulp valve, clean air is introduced to cut off a liquid medicine bridge on the surface of the inner side of the micropore piezoelectric atomization piece, residual liquid medicine is withdrawn to the deep position, and crystallization blockage and back suction pollution are physically prevented. Meanwhile, a tear film digital twinborn model and a reinforcement learning algorithm are built in the central control unit, the tear film rupture time is predicted in real time in combination with multi-dimensional sensor data, and a drug delivery strategy is adaptively optimized according to subjective feedback of a user. The problems of cold stimulation, easy blockage of the nozzle, reverse pollution, control lag and the like in the prior art are effectively solved, and constant-temperature, sanitary, accurate and personalized eye administration is realized.

Description

Technical Field

[0001] This invention relates to the field of medical devices and intelligent drug delivery technology, specifically to a portable intelligent constant temperature eye atomization drug delivery system based on digital twin and pneumatic self-cleaning. Background Technology

[0002] Dry eye syndrome and post-operative eye care require long-term, frequent instillation of medication into the eye. Existing portable devices (such as CN219700291U) primarily use a pump-driven piston to directly deliver medication. This technology has significant drawbacks: Cold stimulation and low absorption rate: Directly instilling room-temperature medication onto the ocular surface triggers a momentary cold shock, leading to reflexive tearing that washes away the medication. Furthermore, the traditional droplet volume far exceeds the conjunctival sac capacity, resulting in significant overflow and waste. Lack of environmental prediction: Relying solely on ocular condition feedback is a delayed response, unable to intervene in advance when entering a dry environment. Lack of personalized adaptation: Different patients have vastly different sensitivities to dryness; existing systems use fixed algorithms and cannot meet individual needs. Hygiene risks: Residual medication at the catheter tip can easily crystallize, clogging the nozzle or fostering bacterial growth.

[0003] Existing patents such as CN108938187A (wet room glasses) use water storage in the temples to generate steam, or CN113230504A uses a long tube to deliver the mist. These solutions have a common problem: the mist is prone to condensing into water droplets in the long tube, resulting in inaccurate dosage; and the end of the tube is always damp, making it very easy for bacteria to grow.

[0004] Existing devices also rely on simple threshold control (e.g., spraying when the number of blinks exceeds X). However, different patients have different sensitivities to the environment, and a single rule cannot meet the individualized comfort needs of each patient, nor can it proactively predict the tear film state. Summary of the Invention

[0005] The present invention aims to solve the above problems and provide a systematic solution by using an intelligent eye drug delivery system with constant temperature nebulization drug delivery, active environmental sensing, negative pressure self-cleaning, and human-computer interaction adaptive learning functions to solve the above problems.

[0006] The specific technical solution is as follows:

[0007] This invention provides a portable intelligent constant-temperature ocular nebulization drug delivery system based on digital twin and pneumatic self-cleaning, comprising a portable main unit that integrates: a drug storage unit, a precision delivery pump, a central control unit, and a rechargeable power module; the input end of the precision delivery pump is connected to the drug storage unit; it also includes an end effector configured to be worn near the patient's eye and connected to the output end of the precision delivery pump via a drug delivery tube; and a sensor module and a human-machine interface terminal connected to the central control unit. The end effector includes a housing, a microfluidic heating module, a gas-liquid separation buffer chamber, and a microporous piezoelectric atomizing plate. The microfluidic heating module is wrapped around the outside of the drug flow channel. The gas-liquid separation buffer chamber is located between the microfluidic heating module and the microporous piezoelectric atomizing plate and is used to temporarily store the heated drug solution. A one-way clean air supply valve is provided on the side wall of the gas-liquid separation buffer chamber. The one-way clean air supply valve connects the external atmosphere with the inside of the buffer chamber, and a microporous sterilization filter membrane is integrated in the valve body. The central control unit is configured to control the precision delivery pump to execute a "pneumatic bridge-breaking self-cleaning program" when the drug administration command ends: drive the precision delivery pump to generate a reverse negative pressure, causing the one-way clean air supply valve to open due to the pressure difference. The outside air enters the gas-liquid separation buffer chamber after being filtered by the microporous sterilization filter membrane. The introduced air flow cuts off the drug solution bridge formed on the inner surface of the microporous piezoelectric atomizing plate and retracts the residual drug solution into the depth of the microfluidic channel. The sensor module includes an environmental sensing unit disposed on the surface of the host unit, and an ocular surface physiological sensor group disposed inside the end effector. The ocular surface physiological sensor group includes a non-contact infrared ocular surface temperature sensor and a capacitive biosensor for detecting blinking and tear overflow. The central control unit, which incorporates a tear film digital twin model and a reinforcement learning algorithm module, is configured to perform the following operations: receive data from the sensor module to calculate the estimated tear evaporation rate, and dynamically adjust the flow rate of the precision delivery pump, the power of the microchannel heating module, and the start / stop of the microporous piezoelectric atomizing sheet based on feedback from the sensor module; receive user subjective feedback commands from the human-computer interaction terminal via the wireless communication module; and adaptively correct key parameters in the tear film digital twin model based on the user subjective feedback commands to construct a personalized drug delivery model.

[0008] According to one embodiment of the present invention, the microporous piezoelectric atomizing plate is a metal microporous plate driven by a centrally located annular piezoelectric ceramic oscillator, wherein the micropore diameter of the microporous plate is 3μm to 8μm; the inner surface of the microporous plate is coated with a nano-superhydrophobic coating to reduce droplet adhesion in conjunction with the pneumatic self-cleaning process; the airflow introduced by the one-way clean air supply valve is laterally blown towards the inner surface of the microporous piezoelectric atomizing plate.

[0009] According to one embodiment of the present invention, the central control unit has a built-in tear film digital twin model and reinforcement learning algorithm module; the tear film digital twin model is configured to receive ambient temperature collected by the sensor module. Ambient relative humidity and ocular surface temperature The predicted tear evaporation rate was calculated based on the following tear evaporation kinetics formula. ":

[0010] in, The preset mass transfer coefficient, To achieve effective ocular surface evaporation area, For temperature The saturated water vapor pressure is approximately determined by the Antoine equation; the tear film digital twin model also calculates the predicted tear film breakup time based on the following tear film stability formula. ":

[0011] in, This is the critical tear film rupture thickness constant. The time interval between the previous blinks. The tear secretion replenishment coefficient, This serves as a lipid layer stability correction factor; the reinforcement learning algorithm module employs Q-learning or SARSA strategies, based on the current environmental state. The input is the drug administration action taken, and the environmental state is... Include and The drug administration action Includes flow rate, duration, and temperature, and is based on reward signals fed back by the human-computer interaction terminal (700). The action value function of the drug administration strategy is iteratively updated according to the following Bellman update formula. :

[0012] in, For learning rate, As a discount factor, The next state after the action is performed. This represents the maximum expected value in the next state.

[0013] According to one embodiment of the present invention, the opening pressure threshold of the one-way clean air supply valve is set to 0.5 kPa to 1.5 kPa; the pore size of the microporous sterilization filter membrane... It is used to block bacteria and particulate matter from entering the buffer chamber.

[0014] According to one embodiment of the present invention, the precision delivery pump is a high-precision piezoelectric pump or a peristaltic pump driven by a stepper motor; in the pneumatic bridge self-cleaning process, the precision delivery pump outputs a pulsed negative pressure waveform with a frequency of 5Hz-20Hz to generate a fluid oscillation effect, which helps to peel off the crystal residue at the microporous piezoelectric atomizing sheet.

[0015] According to one embodiment of the present invention, the human-computer interaction terminal provides a multi-level subjective comfort rating interface; the reinforcement learning algorithm module maps the user's rating to a reward signal Rt+1, wherein "comfort" is mapped to a positive reward value, "dryness" or "foreign body sensation" is mapped to a negative reward value, and when the biosensor detects reflexive tearing, it is mapped to a strong negative reward value.

[0016] According to one embodiment of the present invention, the drug delivery tube adopts a double-layer coaxial structure, with the inner tube transmitting the drug solution and the annular gap between the inner and outer tubes filled with a phase change heat storage material or an aerogel insulation layer to maintain the temperature stability of the drug solution during the transmission process.

[0017] According to one embodiment of the present invention, the central control unit loads a general-purpose... The parameter values ​​are dynamically adjusted during operation using the reinforcement learning algorithm module based on the reward signal history of a specific user, and the km and Parameter values ​​are used to achieve personalized calibration of the digital twin model.

[0018] According to one embodiment of the present invention, the microfluidic heating module includes a flexible PTC heating film surrounding the drug liquid channel and a heat insulation layer wrapping the flexible PTC heating film. The microfluidic heating module is configured to maintain a heating state for a certain period of time when the pneumatic bridge-breaking self-cleaning program is started to dry the drug liquid channel.

[0019] According to one embodiment of the present invention, the outer shell of the portable host unit is provided with a wearing connection structure, and the system is configured in any of the following ways. The present invention also provides a wearing form: ear-hook structure: the host unit is arc-shaped and fits behind the patient's ear, and the end effector extends to the front of the ocular surface through a semi-rigid gooseneck tube; split suspension structure: the host unit is provided with a back clip or lanyard hole, configured to be fixed to the patient's collar or belt, and the medication delivery tube is an adjustable-length anti-tangle silicone tube; integrated frame structure: the host unit is embedded in the temple of the eyeglass frame, and the end effector extends to the nose pad or the edge of the lens rim.

[0020] The beneficial effects of this invention are as follows: 1. This invention completely solves the problems of "reverse contamination" and "crystallization blockage" in micro-drug delivery systems (due to its physical structural advantages), avoiding the back-suction action that draws unfiltered air (even containing eye secretions, bacteria, or dust) from the outside of the nozzle into the tubing, thus preventing a serious risk of reverse contamination. This invention innovatively constructs a "gas-liquid separation buffer chamber" and integrates a "one-way clean air supply valve" in the end effector. During the self-cleaning phase, the system uses the air supply valve to introduce sterile air filtered through a 0.22μm filter membrane to fill the negative pressure space. This design brings a dual breakthrough: Pneumatic bridge breaking and anti-clogging: A clean airflow instantly cuts off the liquid bridge inside the microporous atomizing plate, forcibly "pushing" residual medication back into the deep flow channel, keeping the precision microporous plate dry when not in operation. This effectively prevents device failure caused by crystallization of medication at the micropores due to natural drying, significantly extending the service life of core components. Simultaneously, physical contamination prevention: Ensures that the fluid medium (air) entering the pipeline during the backflow process is clean and sterile, physically blocking the path of external pathogens to enter the drug delivery system, meeting the high hygiene standards of ophthalmic medical devices.

[0021] 2. This invention represents a leap from "passive response" to "active prediction and personalized adaptation" in control. Existing technologies (such as CN108938187A) largely rely on passive responses after detecting dry eye symptoms (such as increased blinking frequency), which is a reactive, "locking the stable door after the horse has bolted" approach. This invention, through a built-in "tear film digital twin model," combined with Fick's diffusion law and real-time environmental / vital data, can accurately calculate the "predicted tear film breakup time (pBUT)" before the user experiences discomfort, thereby enabling preventative micro-dose administration and maintaining ocular surface homeostasis. Furthermore, this invention introduces a "reinforcement learning (RL) algorithm," converting each subjective feedback from the user (such as "comfort" or "dryness") into a reward signal, dynamically adjusting model parameters (such as the mass transfer coefficient). This allows the system to overcome the limitations of general medical models, automatically adapting to individual patient differences (such as differences in evaporation rate due to varying lipid layer thickness), truly achieving "personalized" precision medicine.

[0022] 3. Significantly improves drug bioavailability and patient compliance (clinical experience advantage). Existing technologies mostly use room temperature drug droplets or pipeline delivery nebulization (such as CN113230504A), which can easily cause cold stimulation or drug loss due to pipeline condensation. This invention adopts a technical solution combining "terminal microchannel instantaneous heating" and "microporous piezoelectric nebulization": Eliminating cold shock: The medication is heated to 32-38°C (near body temperature) within milliseconds before being sprayed, eliminating the instantaneous cold stimulation caused by traditional eye drops and avoiding reflex tearing. This prevents the medication from being washed away by tears and significantly increases the residence time of the medication on the ocular surface.

[0023] Highly efficient absorption: the resulting Micron-sized droplets have extremely low momentum, allowing them to gently and evenly cover the corneal and conjunctival surfaces. They are not only rapidly absorbed, but also eliminate the "water jet impact" and "blurred vision period" associated with traditional droplet administration, significantly improving patient comfort and long-term adherence.

[0024] 4. Achieves precision and temperature stability in micro-drug administration. By employing a combination of a double-layered coaxial insulated conduit and a precision delivery pump, this invention not only isolates the thermal effects of ambient temperature on extremely small amounts of drug solution (microliter level), but also achieves nanoliter (nL) level flow control precision through a pulsed pumping waveform combined with the end-effector structure. It is particularly suitable for the precise delivery of highly active biological agents or expensive ophthalmic drugs. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the overall wearing and structure of the system provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the connection structure of each module of the system provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the internal longitudinal cross-sectional structure of the end effector provided in an embodiment of the present invention; Figure 4 A diagram illustrating the pneumatic self-cleaning mechanism for bridge breakage provided in an embodiment of the present invention; Figure 5 A schematic diagram of an intelligent control logic structure based on digital twins and reinforcement learning provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the different wearing configurations of the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] like Figure 1-6 As shown in this embodiment, a portable intelligent constant-temperature ocular nebulization drug delivery system based on digital twin and pneumatic self-cleaning includes a portable host unit 10, which integrates: a drug storage unit 100, a precision delivery pump 200, a central control unit 500, and a rechargeable power module 800; the input end of the precision delivery pump 200 is connected to the drug storage unit 100; it also includes an end effector 300, configured to be worn near the patient's eye, and connected to the output end of the precision delivery pump 200 via a drug delivery conduit 400; and a sensor module 600 and a human-machine interface terminal 700 connected to the central control unit 500; The end effector 300 includes a housing, a microfluidic heating module 310, a gas-liquid separation buffer chamber 320, and a microporous piezoelectric atomizing plate 330. The microfluidic heating module 310 is wrapped around the outside of the drug flow channel. The gas-liquid separation buffer chamber 320 is located between the microfluidic heating module and the microporous piezoelectric atomizing plate 330 and is used to temporarily store the heated drug solution. A one-way clean air supply valve 340 is provided on the side wall of the gas-liquid separation buffer chamber 320. The one-way clean air supply valve connects the external atmosphere with the inside of the buffer chamber, and a microporous sterilization filter membrane 342 is integrated in the valve body. The central control unit 500 is configured to control the precision delivery pump 200 to execute a "pneumatic bridge-breaking self-cleaning program" when the drug administration command ends: drive the precision delivery pump 200 to generate a reverse negative pressure, causing the one-way clean air supply valve 340 to open due to the pressure difference. After being filtered by the microporous sterilization filter membrane, the outside air enters the gas-liquid separation buffer chamber 320. The introduced air flow is used to cut off the drug liquid bridge formed on the inner surface of the microporous piezoelectric atomizing plate 330, and the residual drug liquid is withdrawn to the depth of the microchannel. The sensor module 600 includes an environmental sensing unit 610 disposed on the surface of the host unit 10, and an ocular surface physiological sensor group 620 disposed inside the end effector 300. The ocular surface physiological sensor group 620 includes a non-contact infrared ocular surface temperature sensor 621 and a capacitive biosensor 622 for detecting blinking and tear overflow. The central control unit 500 has a built-in tear film digital twin model and reinforcement learning algorithm module, and is configured to perform the following operations: (a) receive data from the sensor module 600 to calculate the estimated tear evaporation rate, and dynamically adjust the flow rate of the precision delivery pump 200, the power of the microchannel heating module 310, and the start / stop of the microporous piezoelectric atomizing sheet 330 based on the feedback from the sensor module 600; (b) receive user subjective feedback commands from the human-computer interaction terminal 700 through the wireless communication module; (c) based on the user subjective feedback commands, adaptively correct the key parameters in the tear film digital twin model, and construct a personalized drug delivery model.

[0028] Example 1: Overall Hardware Architecture and Wearing Form of Portable System like Figure 1 As shown in Figures 2 and 6, this invention provides a portable intelligent constant-temperature ocular nebulization drug delivery system based on digital twin and pneumatic self-cleaning. The system mainly consists of two parts: a portable host unit 10 and an end effector 300, which are connected by a flexible drug delivery tube 400.

[0029] Portable main unit 10: As the control and power core of the system, the main unit 10 integrates a drug storage unit 100, a precision delivery pump 200, a central control unit 500, and a rechargeable power module 800 within its housing.

[0030] Drug storage unit 100: It adopts a replaceable cartridge design and is preferably made of medical-grade cyclic olefin copolymer (COC), which has good drug stability and is used to store the drug solution to be delivered.

[0031] Precision delivery pump 200: Preferably a piezoelectric micropump, which is characterized by its small size, low noise, and high precision. This pump has bidirectional drive capability, capable of forward delivery of pharmaceutical solutions with nanoliter (nL) precision, and can also generate instantaneous negative pressure of -5kPa to -10kPa in self-cleaning mode.

[0032] The Central Control Unit 500 is the "brain" of the system, featuring a built-in high-performance, low-power MCU that runs the tear film digital twin model and reinforcement learning algorithms. To reduce power consumption, edge computing and a hibernation mechanism are employed: during non-drug administration intervals, the system enters deep hibernation, retaining only low-power Bluetooth broadcasting; AI model training (weight updates) can be performed while charging, or data can be uploaded to a mobile device (cloud) via wireless connection to complete calculations, after which only the updated strategy parameters are sent to the portable host, thereby reducing local power consumption.

[0033] Rechargeable power module 800: supplies power to all components of the system and supports long-term standby.

[0034] Wearing Connection Structure: To adapt to different patients' living scenarios, the shell of the portable main unit 10 is designed with a wearing connection structure, supporting multiple wearing modes: (a) Ear-hook structure: The main unit 10 is ergonomically curved and fits behind the patient's ear (similar to a hearing aid). The end effector 300 extends to the front of the ocular surface through a semi-rigid gooseneck tube, providing good concealment and a secure fit.

[0035] (b) Split-type suspension structure: The main unit 10 is equipped with a back clip or lanyard hole, which can be fixed to the patient's collar, pocket, or belt. The medication catheter 400 is an adjustable-length, anti-tangle silicone tube to reduce the burden on the head.

[0036] (c) Integrated frame structure: The main unit 10 is miniaturized and embedded inside the temple of a special eyeglass frame. The end effector 300 extends invisibly to the nose pad or edge of the frame, and its appearance is no different from ordinary eyeglasses.

[0037] Drug delivery conduit 400: The conduit connecting the main unit and the end unit adopts a double-layer coaxial structure. The inner tube is responsible for conveying the drug solution, and the space between the outer and inner tubes is filled with a phase change heat storage material or aerogel insulation layer. This design can effectively maintain the temperature stability of the drug solution during the transmission process, prevent the drug solution from cooling down in the pipeline due to low ambient temperature, and ensure that the drug solution is within a controllable temperature range when it reaches the heating module.

[0038] Example 2: End effector structure and pneumatic self-cleaning mechanism like Figure 3 and 4 As shown, the end effector 300 is the core execution component of this invention. Its internal structure is specially designed with fluid dynamics to achieve "constant temperature atomization" and "pneumatic self-cleaning".

[0039] 1) Constant temperature atomization component: Microfluidic heating module 310: Wrapped around the outside of the liquid flow channel, it includes a flexible PTC heating film 311 and a heat insulation layer 312. When the liquid flows through this area, the central control unit 500 controls the power of the PTC heating film through a PID algorithm, so that the liquid is instantly heated to 32℃-38℃ close to the human body temperature before being sprayed out, thus eliminating cold stimulation.

[0040] Microporous piezoelectric atomizing plate 330: Located at the very end of the flow channel, it is a metal microporous plate driven by a centrally located annular piezoelectric ceramic oscillator. The micropore diameter is 3μm-8μm. When a high-frequency driving voltage is applied, the microporous plate oscillates, dispersing the heated medication into micron-sized droplets that gently cover the ocular surface. The inner surface of the microporous plate is coated with a nano-superhydrophobic coating with a contact angle greater than 150° to reduce medication adhesion and prepare for subsequent self-cleaning.

[0041] 2) Pneumatic bridge self-cleaning program: In response to the problem that microporous atomizers are prone to clogging due to crystallization of residual medicine, the present invention sets up a gas-liquid separation buffer chamber 320 between the microchannel heating module 310 and the microporous piezoelectric atomizing plate 330, and sets up a one-way clean air supply valve 340 on its side wall.

[0042] One-way clean air supply valve 340: connects the external atmosphere to the inside of the buffer chamber, and the opening pressure threshold is set to 0.5kPa to 1.5kPa. The microporous antibacterial filter membrane 342 ensures that the incoming air is sterile and dust-free.

[0043] Self-cleaning Workflow: 1. Dosing Completion: Upon completion of a single dosing command, the central control unit 500 immediately initiates the "pneumatic bridge-breaking self-cleaning program." 2. Negative Pressure Generation: The precision delivery pump 200 is driven to reverse, generating negative pressure within the flow channel. 3. Valve Opening: Due to the extremely small pores (3-8 μm) of the microporous piezoelectric atomizing plate 330, the capillary resistance of the liquid is extremely high, making it difficult for external air to flow back from the nozzle. At this time, the one-way clean air supply valve 340 on the side wall opens first due to the pressure difference. 4. Airflow Bridge Breaking: After being filtered by the filter membrane 342, the outside air forms a clean airflow that enters the buffer chamber 320 laterally and blows directly onto the inner surface of the microporous piezoelectric atomizing plate 330. This airflow quickly cuts off / destroys the liquid bridge formed by surface tension at the micropores. 5. Retraction and Drying: The residual drug solution, having lost its adhesion points, is cleanly and efficiently retracted into the depths of the microchannel (i.e., behind the microchannel heating module 310) under negative pressure. At this time, the inside of the microporous sheet becomes a dry air environment, physically preventing crystallization blockage. 6. Drying: Simultaneously, the microchannel heating module 310 maintains heating for a certain period (e.g., 5-10 seconds), utilizing residual heat to dry the trace amounts of moisture remaining on the channel walls, further inhibiting bacterial growth.

[0044] Example 3: Intelligent Control Based on Digital Twin and Reinforcement Learning To support the high-precision prediction of the digital twin model and the reward and punishment mechanism of the reinforcement learning algorithm, this system is equipped with a multi-dimensional sensor module 600. For example... Figure 1 and Figure 2 As shown, the module is physically distributed in two locations: the portable host unit 10 and the end effector 300, specifically including: Environmental sensing unit (610): Location: Located on the outer surface of the portable host unit 10, avoiding the area of ​​human skin contact, in order to obtain accurate external environmental data.

[0045] Composition: Employs a high-precision digital temperature and humidity sensor (such as the Sensirion SHT4x series).

[0046] Function: Real-time monitoring of ambient temperature and relative humidity This data is directly input into the digital twin model to calculate the driving force of moisture mass transfer under different environments (such as dry cabins, humid rainy days, and air-conditioned rooms).

[0047] Ocular surface physiological sensor array (620): Location: Highly integrated into the inner edge of the housing of the end effector 300, with its detection window facing the surface of the patient's eyeball and arranged side by side with the nozzle.

[0048] Component 1: Non-contact infrared temperature sensor (621) Selection: Uses a medical-grade miniature infrared thermopile sensor (such as Melexis MLX90632) with a field of view (FOV) of 35° to accurately cover the corneal area.

[0049] Function: Real-time non-contact measurement of ocular surface temperature This data is used to calibrate the saturated vapor pressure of tears. This is because for every 1°C decrease in ocular surface temperature, the tear evaporation rate may change by 5%-10%.

[0050] Component 2: Biofeedback sensor (622) Selection and principle: Miniature capacitive proximity sensors or electric field sensors are preferred.

[0051] Function A (Blink Detection): By monitoring the subtle changes in the capacitive field caused by eyelid opening and closing, it accurately records the user's blinking frequency and blink interval. This data is a key variable for predicting tear film breakup time (pBUT).

[0052] Function B (Reflexive Tear Detection): When the system administers medication excessively, causing the patient to experience "reflexive tearing," excessive tear fluid accumulates or even overflows at the eyelid margin. Since tears are highly conductive electrolytes, this causes a step-like change in the equivalent capacitance value detected by the sensor.

[0053] Logical determination: After the central control unit 500 detects the mutation signal, it immediately determines it as a "strong negative reward" event. The reinforcement learning algorithm will then significantly reduce the flow rate or temperature of the next drug administration to avoid re-stimulation.

[0054] like Figure 5 As shown, the core software of this system lies in the intelligent algorithm running within the central control unit 500, which realizes the leap from "passive response" to "active prediction and personalized adaptation".

[0055] Tear film digital twin model (prediction layer): The system utilizes real-time data collected by sensor module 600: ambient temperature Ambient relative humidity and ocular surface temperature The model is not a simple table lookup, but rather performs real-time calculations based on physicochemical equations. First, the "predicted tear evaporation rate" is calculated based on the tear evaporation kinetics formula. ":

[0056] in, The Antoine equation is used to calculate saturated vapor pressure. This formula is highly sensitive to environmental changes, such as those occurring inside a dry engine compartment. Lowering will lead to A sharp increase.

[0057] Secondly, combining the user's blinking characteristics Using the tear film stability formula to predict tear film breakup time ":

[0058] When predicted When the blink interval is shorter than the user's natural blinking interval, the system determines that there is a risk of dry eye and triggers medication in advance, rather than waiting until the user feels discomfort.

[0059] mass transfer factor The initial universal value was determined through laboratory in vitro simulation experiments. The specific method was as follows: in a standard constant temperature and humidity chamber (e.g., temperature 25℃, relative humidity 40%), using a simulated ocular surface with a known surface area (a concave container containing physiological saline), the water mass loss per unit time was measured, and the baseline mass transfer coefficient was derived by reverse calculation. (Typical value range is) to ).

[0060] During the user's wearing and use, the system utilizes reinforcement learning algorithms to... Dynamic adjustments will be made. If the user's prediction is incorrect... If a user reports "dryness" before the designated time has elapsed, it indicates that their actual tear evaporation rate is higher than the calculated value by the general model (possibly due to a larger palpebral fissure or a lack of tear lipid layer). The system will automatically increase the sensitivity. This value allows the model's predictions to more closely reflect the user's actual physiological state.

[0061] Lipid layer stability modifier This is an empirical parameter reflecting the tear film lipid layer's ability to inhibit tear film breakage. Its initial value is set based on clinical statistical data from different dry eye subtypes. Upon first use, the system allows users to input simple dry eye questionnaire results (SPEED or OSDI) via the accompanying app. If the user is diagnosed with "lipid-deficient dry eye," the system will load a smaller... Initial value (indicating a thin lipid layer and poor stability); if it is "aqueous deficiency type", then a larger load is applied. Initial value.

[0062] This parameter also participates in the update of the Bellman equation. When the biosensor (622) detects frequent reflexive tearing in the user (indicating sudden tear film instability), the algorithm reduces... Value, thus in calculation This reduces the predicted rupture time, prompting the system to intervene with medication earlier.

[0063] Reinforcement Learning Algorithm Module (Decision and Evolution Layer): In order to adapt to individual differences among different users (such as different lipid layer thickness and tear secretion capacity), the system introduces a reinforcement learning (RL) mechanism for self-evolution.

[0064] Strategy Model: Q-learning or SARSA strategy is adopted.

[0065] state : current , and environmental scenarios.

[0066] action Dosing strategies (e.g., flow rate in micro-lubrication mode, duration in strong lubrication mode, heating temperature, etc.).

[0067] Human-computer interaction feedback (Reward): Users provide subjective feedback through human-computer interaction terminals (such as mobile apps or buttons on the host).

[0068] If the user clicks "Comfort", the system receives a positive reward. .

[0069] If the user clicks on "dryness / foreign body sensation," or if the biosensor detects reflexive tearing (overdose stimulation), the system receives a negative reward. .

[0070] Model Update: The system updates the action-value function based on the Bellman equation. :

[0071] Parameter fine-tuning: In addition to updating the strategy, the algorithm also dynamically fine-tunes key parameters in the digital twin model (such as the quality transfer coefficient) based on historical reward signals. and secretion coefficient For example, for a patient whose tears evaporate extremely quickly, the system will automatically adjust their personalized [treatment / condition]. The value is adjusted to make the model's predictions more consistent with the patient's actual physiological condition.

[0072] In the formula, : This means the current action has been performed. Then, the system enters the "next state" (e.g., the predicted evaporation rate and burst time updated at the next moment).

[0073] It is in this new state Below, every possible action that the system can theoretically perform.

[0074] This item means "best-case scenario for the future." The algorithm "looks ahead" and assumes all possible actions in the next state. Evaluate all actions to find the one that brings the greatest value (highest Q value), and use this maximum value as the basis for calculating the current value.

[0075] Based on the physical scenario of this invention: Assuming the current ( The system performed an action at that moment. (e.g., "medium flow rate spray").

[0076] The environment then became (For example, "humidity slightly increases, and eye surface temperature decreases").

[0077] at this time, This means that the system will be in the next moment. A set of selectable actions, such as: = Stop spraying; = Maintain the current flow rate; = Increase the flow rate and raise the temperature.

[0078] The algorithm passes To find these Which one is most advantageous in the future, and thus we can deduce the current action? Is it wise?

[0079] In short: It means "all possible moves for the next step".

[0080] In summary, this invention solves the problem of eliminating cold shock and achieving efficient absorption by combining "instantaneous heating of the end microchannel" with "microporous piezoelectric atomization". It solves the reliability problem of micro-drug delivery devices being prone to clogging and contamination by "pneumatic self-cleaning" in hardware, and solves the accuracy problem of traditional devices being unable to cope with environmental changes and individual differences by "digital twin + reinforcement learning" in software. It constitutes a complete intelligent ocular drug delivery system with high innovation and practical value.

[0081] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A portable intelligent constant-temperature ocular nebulization drug delivery system based on digital twin and pneumatic self-cleaning, comprising a portable host unit (10), which integrates: a drug storage unit (100), a precision delivery pump (200), a central control unit (500), and a rechargeable power module (800); the input end of the precision delivery pump (200) is connected to the drug storage unit (100); it also includes an end effector (300), configured to be worn near the patient's eye, and connected to the output end of the precision delivery pump (200) via a drug delivery conduit (400); and a sensor module (600) and a human-machine interface terminal (700) connected to the central control unit (500); Its features are: The end effector (300) includes a housing, a microfluidic heating module (310), a gas-liquid separation buffer chamber (320), and a microporous piezoelectric atomizing plate (330); the microfluidic heating module (310) is wrapped around the outside of the liquid flow channel; the gas-liquid separation buffer chamber (320) is located between the microfluidic heating module and the microporous piezoelectric atomizing plate (330) and is used to temporarily store the heated liquid; a one-way clean air supply valve (340) is provided on the side wall of the gas-liquid separation buffer chamber (320), the one-way clean air supply valve connects the external atmosphere and the inside of the buffer chamber, and a microporous sterilization filter membrane (342) is integrated in the valve body; The central control unit (500) is configured to control the precision delivery pump (200) to perform a "pneumatic bridge-breaking self-cleaning procedure" at the end of the drug administration command: drive the precision delivery pump (200) to generate a reverse negative pressure, causing the one-way clean air supply valve (340) to open due to the pressure difference, and allow outside air to enter the gas-liquid separation buffer chamber (320) after being filtered through the microporous sterilization filter membrane. The introduced air flow is used to cut off the liquid bridge formed on the inner surface of the microporous piezoelectric atomizing plate (330), and the residual liquid is withdrawn to the depth of the microchannel. The sensor module (600) includes an environmental sensing unit (610) disposed on the surface of the host unit (10) and an ocular surface physiological sensor group (620) disposed inside the end effector (300). The ocular surface physiological sensor group (620) includes a non-contact infrared ocular surface temperature sensor (621) and a capacitive biosensor (622) for detecting blinking and tear overflow. The central control unit (500) has a built-in tear film digital twin model and reinforcement learning algorithm module, and is configured to perform the following operations: (a) receive data from the sensor module (600) to calculate the estimated tear evaporation rate, and dynamically adjust the flow rate of the precision delivery pump (200), the power of the microchannel heating module (310), and the start and stop of the microporous piezoelectric atomizing sheet (330) in combination with the feedback from the sensor module (600); (b) receive user subjective feedback instructions from the human-computer interaction terminal (700) through the wireless communication module; (c) based on the user subjective feedback instructions, adaptively correct the key parameters in the tear film digital twin model, and construct a personalized drug delivery model.

2. The system according to claim 1, characterized in that, The microporous piezoelectric atomizing plate (330) is a metal microporous plate driven by a centrally located annular piezoelectric ceramic oscillator. The micropore diameter of the microporous plate is 3μm to 8μm. The inner surface of the microporous plate is coated with a nano-superhydrophobic coating to reduce droplet adhesion in conjunction with the pneumatic self-cleaning process. The air introduced by the one-way clean air supply valve (340) is blown laterally towards the inner surface of the microporous piezoelectric atomizing plate (330).

3. The system according to claim 1, characterized in that, The central control unit (500) has a built-in tear film digital twin model and reinforcement learning algorithm module; the tear film digital twin model is configured to receive the ambient temperature collected by the sensor module (600). Ambient relative humidity and ocular surface temperature The predicted tear evaporation rate was calculated based on the following tear evaporation kinetics formula. ": in, The preset mass transfer coefficient, To achieve effective ocular surface evaporation area, For temperature The saturated water vapor pressure is approximately determined by the Antoine equation; the tear film digital twin model also calculates the predicted tear film breakup time based on the following tear film stability formula. ": in, This is the critical tear film rupture thickness constant. The time interval between the previous blinks. The tear secretion replenishment coefficient, It is a lipid layer stability modifier; The reinforcement learning algorithm module employs Q-learning or SARSA strategies, based on the current environment state. The input is the drug administration action taken, and the environmental state is... Include and The drug administration action Includes flow rate, duration, and temperature, and is based on reward signals fed back by the human-computer interaction terminal (700). The action value function of the drug administration strategy is iteratively updated according to the following Bellman update formula. : in, For learning rate, As a discount factor, The next state after the action is performed. This represents the maximum expected value in the next state.

4. The system according to claim 3, characterized in that, The opening pressure threshold of the one-way clean air supply valve (340) is set to 0.5 kPa to 1.5 kPa; the microporous sterilization filter membrane (342) It is used to block bacteria and particulate matter from entering the buffer chamber.

5. The system according to claim 1, characterized in that, The precision delivery pump (200) is a high-precision piezoelectric pump or a peristaltic pump driven by a stepper motor; in the pneumatic bridge self-cleaning program, the precision delivery pump outputs a pulsed negative pressure waveform with a frequency of 5Hz-20Hz to generate a fluid oscillation effect, which helps to peel off the crystal residue at the microporous piezoelectric atomizing sheet.

6. The system according to claim 3, characterized in that, The human-computer interaction terminal (700) provides a multi-level subjective comfort rating interface; the reinforcement learning algorithm module maps the user's rating into a reward signal. The value is mapped to "comfort" as a positive reward, "dryness" or "foreign body sensation" as a negative reward, and when the biosensor detects reflexive tearing, it is mapped to a strong negative reward.

7. The system according to claim 1, characterized in that, The drug delivery conduit (400) adopts a double-layer coaxial structure. The inner tube transmits the drug solution, and the annular gap between the inner and outer tubes is filled with a phase change heat storage material or an aerogel insulation layer to maintain the temperature stability of the drug solution during transmission.

8. The system according to claim 3, characterized in that, The central control unit loads a general-purpose system in its initial state. The parameter values ​​are dynamically adjusted during operation using the reinforcement learning algorithm module based on the reward signal history of a specific user. and Parameter values ​​are used to achieve personalized calibration of the digital twin model.

9. The system according to claim 1, characterized in that, The microfluidic heating module (310) includes a flexible PTC heating film (311) surrounding the drug liquid flow channel and a heat insulation layer (312) wrapping the flexible PTC heating film (311). The microfluidic heating module (310) is configured to maintain a heating state for a certain period of time when the pneumatic bridge self-cleaning program starts to be executed, so as to dry the drug liquid flow channel.

10. The system according to claim 1, characterized in that, The portable host unit (10) has a wearing connection structure in its shell, and the system is configured in any of the following wearing forms: (a) ear hook structure: the host unit (10) is arc-shaped and fits behind the patient's ear, and the end effector extends to the front of the ocular surface through a semi-rigid gooseneck tube; (b) split suspension structure: the host unit (10) has a back clip or lanyard hole, configured to be fixed to the patient's collar or belt, and the medication delivery tube is an adjustable length anti-tangle silicone tube; (c) frame integrated structure: the host unit (10) is embedded in the temple of the eyeglass frame, and the end effector extends to the nose pad or the edge of the lens rim.

Images