An ophthalmic disease diagnosis system based on remote slit lamp

Abstract

The present invention belongs to the field of intelligent medical care, and specifically relates to an ophthalmic disease diagnosis system based on a remote slit lamp. The system includes a remote control end and a proximal execution end, and the remote control end and the proximal execution end are connected to each other through a network. The system performs the following method: S1: the remote control end acquisition unit acquires the control instruction of the electrical signal and inputs it into the remote signal conversion unit, which encodes the control instruction into a corresponding audio signal and sends it to the proximal execution end through the audio channel of the remote interface unit; S2: the proximal execution end receives the audio signal through the audio channel of the proximal interface unit, and the proximal signal conversion unit decodes the audio signal into the control instruction of the electrical signal. Based on the control instruction execution unit, the slit lamp is driven to perform the action corresponding to the control instruction, and the camera unit captures the ophthalmic diagnosis image. The present application will realize plug-and-play remote medical slit lamp diagnosis, eliminating the inconvenience of installing a driver, and making the system more compatible.

Description

Technical Field

[0001] The present invention relates to the field of intelligent medical treatment, and more specifically, to an ophthalmic disease diagnosis system based on a remote slit lamp. Background Art

[0002] Medical and health resources are unevenly distributed, with scattered and numerous primary care resources and a relatively insufficient supply of high-quality medical and health resources. Leveraging advanced technologies, the establishment of telemedicine systems and telemedicine management can effectively address this uneven distribution of medical resources. Current telemedicine systems primarily include remote surgery, remote consultation, and remote outpatient services. The clinical application of telesurgery has been reported both domestically and internationally. These procedures utilize dedicated networks, remote consultations, and remote outpatient services. Consulting experts provide remote guidance, and diagnoses are made based on uploaded data. The primary purpose of dedicated networks is to ensure communication quality during remote diagnosis and treatment. However, the high maintenance costs of dedicated networks make them prohibitive for systems designed to promote widespread remote diagnosis and treatment. Consequently, signal transmission speeds are a significant challenge for remote diagnosis and treatment over the internet.

[0003] The hospital's ophthalmology remote virtual diagnosis and treatment platform adopts the aforementioned remote consultation and remote outpatient model. Both consultations and outpatient clinics are conducted through remote film reading and remote review of test reports. Ophthalmology clinics utilize numerous instruments, and the slit lamp microscope, as the most basic medical device, requires ophthalmologist operation, with examination results closely tied to the doctor's operation. To further improve access to medical resources for people in special occupations and enhance the accuracy of eye disease diagnosis, a slit lamp diagnosis system for remote virtual diagnosis and treatment platforms using dedicated networks for these populations is urgently needed. Que Tianxing et al. designed a remote slit lamp diagnosis platform based on the Internet of Things (IoT). This platform utilizes a doctor-side / server-side / device-side distributed model, but it can only operate under dedicated remote consultation software, an operator-side driver, and an execution-side driver. Chen Junfa et al. designed a web-based remote slit lamp microscope diagnosis system. This system utilizes a B / S architecture, but requires the installation of a dedicated driver on the handle. G. Lahaie Luna et al. proposed a 3D robot-controlled unmanned slit lamp. The system uses a dedicated network for communication data and command data transmission. N. Tanabe et al. proposed a remote-operated slit lamp microscope system based on a client-server architecture and implemented with supporting remote consultation software. D. Nankivil et al. designed a robotic remote-controlled slit lamp system that uses a remote desktop to remotely control the slit lamp, requiring the installation of corresponding remote desktop control software.

[0004] It can be seen that the remote control system in the existing technology often involves the installation of drivers and desktop operating programs, and the drivers, desktop programs, etc. are written in a specific programming language, and their operation process often depends on the specific language and language dependency packages. Since the system platforms and system environments of the remote computers used are different, it is very easy to encounter installation failure problems in actual application scenarios, which also makes the system deployment, implementation and promotion of remote diagnosis and treatment difficult. Summary of the Invention

[0005] In view of the above problems, the present invention provides an ophthalmic disease diagnosis system based on a remote slit lamp. By converting the control electrical signal of the slit lamp into an audio frequency signal and transmitting and decoding it, driver-free remote diagnosis is achieved. At the same time, encoding is used to compress the space during signal transmission to speed up data transmission, thereby constructing a lightweight remote slit lamp ophthalmic disease diagnosis system.

[0006] The present application (first aspect) discloses an ophthalmic disease diagnosis system based on a remote slit lamp, comprising:

[0007] A remote control end and a near-end execution end, wherein the remote control end and the near-end execution end are connected to each other through a network, wherein the remote control end includes: an acquisition unit, a far-end signal conversion unit, and a far-end interface unit; and the near-end execution end includes: a near-end interface unit, a near-end signal conversion unit, an execution unit, and a camera unit; and the system executes the following method:

[0008] S1: The remote control end acquisition unit acquires the control instruction of the electrical signal and inputs it into the remote signal conversion unit, which encodes the control instruction into a corresponding audio signal and sends it to the local execution end through the audio channel of the remote interface unit;

[0009] S2: The proximal execution end receives the audio signal through the audio channel of the proximal interface unit, and the proximal signal conversion unit decodes the audio signal into a control instruction of an electrical signal. Based on the control instruction execution unit, the slit lamp is driven to perform an action corresponding to the control instruction, and the camera unit then captures an ophthalmic examination image.

[0010] Furthermore, the method performed by the system further includes:

[0011] S3: sending the ophthalmological examination image to the remote control terminal through the video channel of the proximal interface unit,

[0012] S4: The remote control terminal receives the diagnosis image through the video channel of the remote interface unit, and obtains an updated control instruction from the acquisition unit according to the diagnosis image;

[0013] S1-S4 are repeated until the remote control terminal receives the ophthalmic examination image required for ophthalmic disease examination, and the ophthalmic disease diagnosis is performed based on the ophthalmic examination image.

[0014] Furthermore, the method executed by the system further includes: the proximal execution end further includes an adaptive control unit, the adaptive control unit automatically executing a set of control instructions based on an adaptive control strategy according to the received control instructions to adjust the slit lamp to a next examination state, and the method for obtaining the adaptive control strategy includes:

[0015] Step 1: Obtain an initial robotic arm motion and an initial state corresponding to the initial motion, wherein the initial state is a current examination setting of the slit lamp;

[0016] Step 2: Reinforcement learning takes the next action in the action space, which includes adjusting the slit lamp intensity, slit width, slit color, slit lamp diffusion, lamp illumination angle, imaging system angle, microscope left and right movement, microscope front and back movement, microscope lifting movement, and imaging magnification;

[0017] Step 3: After the action is executed, the slit lamp switches to the next examination setting and an updated examination image is obtained based on the current examination setting.

[0018] Step 4: Obtain the reward function of the action based on the updated inspection image, and obtain the adaptive control strategy of the robotic arm through training iterations of the reward function.

[0019] Furthermore, the acquisition unit uses the controller handle to acquire control instructions for the electrical signal of the handle operation;

[0020] Optionally, the remote signal conversion unit is built into the controller handle;

[0021] Optionally, the remote signal conversion unit is connected to a controller handle and is connected to a computer via an audio interface of the computer at the remote control end.

[0022] Furthermore, the control instructions or the actions corresponding to the control instructions include adjusting any one or more of the following: slit intensity, slit width, slit color, slit angle, slit light diffusion, slit lamp irradiation angle, imaging system angle, left and right movement of the microscope, front and back movement of the microscope, lifting and lowering movement of the microscope, and imaging system magnification.

[0023] Furthermore, encoding the control instruction into an audio signal is to use different audio frequencies to represent different control instructions, and different control instructions correspond to corresponding actions of the instrument end;

[0024] Optionally, the remote interface unit and the proximal interface unit are two computers connected via a network.

[0025] Furthermore, the camera unit of the proximal execution end includes a slit lamp microscope inspection camera and a monitoring camera, the slit lamp microscope inspection camera is used to capture the inspection image of the microscope, and the monitoring camera is used to capture the instrument operation status image when the slit lamp microscope performs inspection; the proximal execution end transmits the inspection image and the instrument operation status image to the remote control end through the proximal interface unit for display.

[0026] Furthermore, the slit lamp microscope inspection camera is an electronic eyepiece;

[0027] Optionally, the proximal execution end also includes a video unit, which merges the inspection image of the microscope and the instrument operation status image as the main screen and secondary screen of the picture-in-picture respectively into one image and transmits it to the instrument end interface unit, and transmits the inspection image to the remote control end through the video interface; the remote control end acquisition unit also obtains the switching instruction for the main screen and secondary screen of the picture-in-picture and transmits it to the remote interface unit, the remote interface unit transmits the switching instruction and audio signal to the proximal interface unit and then separates the audio signal and the switching instruction, the video unit sets the main screen of the original picture-in-picture as the secondary screen of the new picture-in-picture according to the received switching instruction, sets the main and secondary screens of the original picture-in-picture to the main screen of the new picture-in-picture, and then transmits the new picture-in-picture to the proximal interface unit and then sends it to the remote control end.

[0028] Furthermore, the acquisition unit also includes a microphone module for acquiring the doctor's voice communication signal to the patient, and transmitting the voice communication signal to the remote interface unit. The control end interface unit transmits the reverberation signal obtained by reverberating the audio signal and the voice communication signal to the near-end interface unit. The near-end interface unit separates the reverberation signal to obtain the audio signal and the voice communication signal, and the voice communication signal is played through the voice playback device.

[0029] Furthermore, the remote control end also includes a camera unit for capturing the communication screen between the doctor and the patient, transmitting the communication screen to the control end interface unit, transmitting the communication screen to the instrument end interface unit through the data interface, and the instrument end interface unit transmitting the communication screen to the instrument end display for display.

[0030] Furthermore, the remote control terminal includes multiple terminals; one of the multiple remote control terminals is a master control terminal used to control the actions of the near-end execution terminal, and the near-end execution terminal transmits the picture-in-picture to multiple remote control terminals for display, and the multiple remote control terminals are interconnected for video conferencing.

[0031] This application has the following beneficial effects:

[0032] (1) This application converts the instructions of the doctor's handle control into audio signals and parses them on the instrument side, thereby realizing plug-and-play remote medical slit lamp examination, eliminating the inconvenience of installing a driver and making the system more compatible;

[0033] (2) The instrument control command signal and the voice communication signal are mixed to send the command without hindering the communication between the two parties. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 work.

[0035] Figure 1 is a schematic diagram of a system provided by the first aspect of an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of a remote slit lamp medical diagnosis system provided by an embodiment of the present invention;

[0037] Figure 3 This is a schematic diagram of a remote-controlled slit lamp functional unit provided by an embodiment of the present invention;

[0038] Figure 4 1 is a physical diagram of a remote-controlled slit lamp provided by an embodiment of the present invention, wherein (a) shows the front of the remote-controlled slit lamp; (b) shows the side of the remote-controlled slit lamp;

[0039] Figure 5 A schematic diagram of the components of a control handle provided by an embodiment of the present invention;

[0040] Figure 6 A physical diagram of a control handle provided in an embodiment of the present invention;

[0041] Figure 7 Schematic diagram of system test results provided by an embodiment of the present invention: (a) shows a remote examination room at the hospital; (b) shows a slit lamp examination room at the instrument; (c) shows a front view of the system test, with picture-in-picture technology displaying the examination image and the status of the slit lamp; (d) shows a side view of the system test, with picture-in-picture technology displaying the examination image and the side status of the slit lamp;

[0042] Figure 8 A schematic diagram of remote control command audio encoding and near-end instrument decoding provided by an embodiment of the present invention;

[0043] Figure 9A schematic diagram of a near-end instrument-side shooting data transmission provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0045] In some of the processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The serial numbers of the operations, such as S101, S102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., and do not represent the order of precedence, nor do they limit "first" and "second" to be different types.

[0046] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0047] Figure 1 An embodiment of the present invention provides an ophthalmic disease diagnosis system based on a remote slit lamp. The system includes a remote control terminal and a proximal execution terminal, which are connected to each other via a network. The remote control terminal includes: an acquisition unit, a remote signal conversion unit, and a remote interface unit; the proximal execution terminal includes: a proximal interface unit, a proximal signal conversion unit, an execution unit, and a camera unit. The system performs the following method:

[0048] S1: The remote control end acquisition unit acquires the control instruction of the electrical signal and inputs it into the remote signal conversion unit, which encodes the control instruction into a corresponding audio signal and sends it to the local execution end through the audio channel of the remote interface unit;

[0049] S2: The proximal execution end receives the audio signal through the audio channel of the proximal interface unit, and the proximal signal conversion unit decodes the audio signal into a control instruction of an electrical signal. Based on the control instruction execution unit, the slit lamp is driven to perform an action corresponding to the control instruction, and the camera unit then captures an ophthalmic examination image.

[0050] In order to realize the control of slit lamp microscope for limited subject consultation based on a dedicated network remote virtual diagnosis and treatment platform, a remote slit lamp microscope diagnosis system based on audio and video signal transmission instructions was designed.

[0051] First, a conventional digital slit lamp microscope was retrofitted with electronic control. Then, audio signals were used to transmit instrument control and training commands. These mixed command signals and voice communication signals were transmitted via the audio channel of a dedicated network diagnosis and treatment platform for specialized personnel. Using video mixing technology, instrument monitoring images and diagnostic data were transmitted via the dedicated network diagnosis and treatment platform's video channel.

[0052] The remote slit lamp diagnosis system's handle and instrument are connected to a computer via headset and camera interfaces. Experimental results demonstrate that the system can remotely control common slit lamp functions, and to date, 50 patients have been remotely examined. This marks the first slit lamp microscope that can be remotely controlled via a dedicated network remote diagnostic and treatment platform.

[0053] The aforementioned remote slit lamp research is difficult to implement using existing video consultation platforms and cannot be directly applied to current dedicated network virtual diagnosis and treatment platforms. Considering the confidentiality and information security requirements of dedicated networks, the system should avoid installing software and device drivers. In response to the imbalanced, dispersed, confidential, and insecure nature of dedicated network virtual diagnosis and treatment platforms and the medical resources required for specialized populations, this paper designs a remote slit lamp diagnostic system based on a dedicated network remote virtual diagnosis and treatment platform.

[0054] First, a conventional digital slit lamp microscope was retrofitted with electronic control. Then, audio signals were used to transmit instrument control commands, mixed command signals, and voice communication codes via the audio channel of a dedicated network diagnosis and treatment platform. Using video mixing technology, instrument monitoring and examination images were transmitted via the platform's video channel. The remote slit lamp medical examination system's handle and instrument interface were connected to a computer via headset and camera ports, eliminating the need for software or device driver installation. This resulted in the first slit lamp microscope capable of remote control via a dedicated network remote diagnosis and treatment platform.

[0055] 1. System Overview

[0056] 1.1 System Architecture

[0057] The overall block diagram of the remote slit lamp medical diagnosis system is as follows: Figure 2The system consists of three parts: a doctor's operating terminal, an instrument execution terminal, and a video conferencing platform. The doctor's operating terminal retains the doctor's operating habits and is designed with a HID (Human Interface Device) handle controller. On the instrument execution terminal, an existing slit lamp microscope is modified to enable remote control of 11 common slit lamp functions. Regarding the video conferencing platform, the hospital's existing video consultation system is utilized to meet the confidentiality requirements and information security characteristics of the dedicated network. Instructions and data are transmitted through the audio and video channels of the virtual diagnosis and treatment platform, eliminating the need to install proprietary software and drivers.

[0058] Figure 2 This is a schematic diagram of a remote slit lamp medical diagnosis system provided by an embodiment of the present invention;

[0059] 1.2 Remote control principle based on dedicated network

[0060] On the doctor's end, the control handle modulates the control commands into transmittable audio signals, which are then connected to the audio interface of a computer connected to a dedicated network. This audio signal is then transmitted via the established video consultation using the audio channel of the existing diagnosis and treatment platform. The system uses audio mixing and signal modulation circuits to combine the instrument control command signals with the voice communication signals, effectively sending commands and enabling communication between the two parties.

[0061] On the device side, the remote-controlled slit lamp connects to the audio interface of the "East Network" computer to receive commands. The instrument uses a frequency detection circuit, signal conditioning circuit, and motor driver circuit to decode the level signal, extract the control instructions, and drive the slit lamp for observation. The physician operating and instrument execution functions of the remote slit lamp diagnosis system are connected to the computer via a video interface, eliminating the need to install device drivers or video consultation software. The remote virtual diagnosis and treatment platform can only transmit one video channel. This paper utilizes video mixing technology to transmit the instrument's office monitoring video and the slit lamp diagnosis video to the physician through the diagnosis and treatment platform's video channel.

[0062] 1.3 Remote Control Slit Lamp Execution Terminal Structure

[0063] Remote control of slit lamp is realized by modifying the conventional digital slit lamp (S390L, Shanghai Meiwo) Figure 9 The remote-controlled slit lamp receives the command signal and drives the slit lamp body to complete the medical examination.

[0064] as follows Figure 3 It is a functional unit for remote control of slit lamp. The audio signal from the doctor side is connected to the audio input port of the execution side interface unit;

[0065] The interface unit separates the input audio signal into two paths through a filter: one path is sent to the MCU core unit, which controls the motor transmission by converting the audio signal into operation instructions;

[0066] The other path is transmitted to the external speaker through the audio output port to play the doctor's voice guidance to the subject.

[0067] The executive end sensor unit is used to detect the status of the slit lamp body.

[0068] The MCU core unit decodes the audio signal from the interface unit, controls the video unit to switch screens according to instructions, and controls the motor drive unit to drive the slit lamp body to complete the instruction action.

[0069] The operator remotely controls the brightness of the slit light band, the angle of the slit illumination source, and the imaging magnification to examine the patient's eyes. The operator observes the patient's examination room and the operating status of the instrument, completing remote interaction with the patient. The transmission system is designed to adjust its physical structure to complete the transformation of the mechanical part. All motors are driven by the core control unit. Appropriate sensors such as Hall sensors and angle sensors are set to monitor the operating status of the instrument to achieve closed-loop control of the stepper motor. The actual remote-controlled slit lamp is shown in the figure, and the remote slit lamp function description is listed in Table 1:

[0070] Table 1 Description of remote-controlled slit lamp functions

[0071]

[0072]

[0073] The sampling rate of a general computer sound card is 44KHz, which means that the selected frequency cannot exceed 22KHz in theory.

[0074] Among them, imaging magnification, slit spot width, slit spot color, slit spot angle, and diffuse light are discrete audio frequencies;

[0075] Others are based on continuous audio frequency control.

[0076] In some embodiments, 8 frequencies are selected ( Figure 8 (as shown) or more, f1 to f8, each frequency represents a binary bit, with the presence or absence of that frequency corresponding to 1 and 0. Codes f1 to f4 indicate function. For example, if the slit (slit light spot) intensity is 0001 and the slit width is 0010, codes f5 to f8 indicate the amount of motion. For example, if the slit (slit light spot) width is 5mm, the code is 00100101, resulting in the presence of frequencies f3, f6, and f8 in the generated signal. The specific values ​​of f1 to f8 can be set according to actual conditions, as long as they can be transmitted through the video conferencing system's audio channel.

[0077] In some embodiments, the frequencies used are 697, 770, 852, 941, 1209, 1336, 1477, and 1633 Hz;

[0078] In some embodiments, the frequencies used are integral increases of the above frequencies, such as 8000 Hz, then the frequencies are 8697, 8770, 8852, 8941, 9209, 9336, 9477 and 9633 Hz.

[0079] 1.4 Composition of the Doctor's Remote Control Terminal

[0080] The function of the doctor's side control handle is to convert the doctor's operation into control instructions and modulate them into audio signals that can be transmitted. Figure 3 The figure shows a schematic diagram of the components of the control handle. The control handle consists of a sensor unit, an MCU core unit, and an interface unit. The sensor unit contains a rotary potentiometer, a gear switch, a lever, and a push button switch, which convert the doctor's operation into an electrical signal; the MCU core unit collects the electrical signal output by the sensor and generates a command signal (audio) to control the slit lamp according to the coding rules: the interface unit completes the microphone drive on the doctor's side, and reverberates the microphone signal with the control command audio signal. The control handle is ergonomically designed to continue the doctor's operating habits to the greatest extent, making the new system easy to use. The actual doctor's handle is shown in the figure below. Figure 6 shown.

[0081] The remote slit lamp controller is deployed in the hospital's ophthalmology clinic, while the physician control terminal is located in the remote clinic of the ophthalmology department of a hospital in Beijing. Both terminals are connected via the hospital's dedicated network remote virtual diagnosis and treatment platform. The physician uses a handheld controller to remotely control the slit lamp, performing the 11 common functions listed in Table 1.

[0082] The system runs on the hospital's dedicated network remote virtual diagnosis and treatment platform, which can obtain clear images that meet the diagnosis requirements and achieve good real-time interaction in remote consultations. Figure 7 As shown in Figure 2. So far, the system has provided remote diagnosis and treatment for 50 patients.

[0083] Telemedicine consultations via video conferencing have become a means of increasing access to medical resources, and remote control of medical examination instruments has become a means of improving the quality of remote consultations. Existing literature on remote slit lamp microscopy systems is complex to deploy and cannot be directly applied to existing video consultation platforms. These systems require the installation of device drivers, resulting in low convenience and information security. This paper proposes a slit lamp examination system based on a dedicated network-based remote virtual diagnosis and treatment platform. This system leverages the existing dedicated network-based remote virtual diagnosis and treatment platform, uses audio and video transmission for data transmission, and does not require the installation of any software or device drivers, thus improving the ease of system deployment.

[0084] In actual testing, the remote-controlled slit lamp was able to remotely control 11 common basic functions, obtaining images that met diagnostic requirements, with video clarity and smoothness meeting diagnostic requirements. The latency of slit lamp microscope command transmission was less than 1 second. The proposed slit lamp diagnosis system, based on a dedicated network-based remote virtual diagnosis and treatment platform, improves the convenience and network security of the remote diagnosis system, achieving plug-and-play operation.

[0085] 2. Intelligent system optimization

[0086] The slit lamp diagnosis system can be further optimized. On the basis of closed-loop control, a control algorithm can be used to establish a remote control mathematical model to improve the system control performance and dynamic characteristics; artificial intelligence technology can be integrated into the system to obtain images for preliminary diagnosis, thereby improving the efficiency and accuracy of diagnosis.

[0087] Some proximal execution ends also include an adaptive control unit, which automatically executes a set of control instructions based on an adaptive control strategy according to the received control instructions to adjust the slit lamp to the next inspection state, wherein the adaptive control strategy is obtained based on reinforcement learning.

[0088] Reinforcement learning consists of two themes and four parts. The two main parts are:

[0089] Agent: An agent is an individual who takes an action;

[0090] Environment: The environment can provide feedback on behavioral actions;

[0091] The four parts include: A, S, R, and P; A represents the action space, that is, the space of all actions in which the agent takes action; S represents the state space, which is the feedback state after the agent takes action; R represents the reward, which is a real number. Positive values ​​represent rewards, and negative values ​​represent penalties. P represents the strategy, that is, what action a∈A the agent will take in state s∈S (sometimes A is used to represent the strategy).

[0092] The algorithm uses the agent to interact with the environment, and uses the rewards and states of the environment to guide the agent to update its actions, with the goal of maximizing long-term rewards, and then obtaining an optimal strategy. The typical Q-learning in reinforcement learning is summarized as follows: First, the agent collects data through sampling, {(s i ,a i ,s′ i ,r i )}, where s i Execute action a in the current state i After the state changes to s ′ i , and get reward ri ,

[0093] Then, the Q function (action value function) is: Among them, r(s,a) in the Q function represents the current instantaneous reward, represents the action that can get the maximum reward by taking different actions at the next state s′; γ represents the discount factor;

[0094] Our target is:

[0095] Then the objective function is: Q-learning updates the Q-value table by continuously sampling and obtaining different rewards, so that the optimal action under different states can be determined based on the Q-table.

[0096] When the state is simple, the Q-table can be obtained through exhaustive updates. However, when the state space and action space become increasingly complex, Q-learning gradually becomes inadequate. Therefore, by introducing neural networks in deep learning to maintain a neural network representing the Q-table, the DQN algorithm was generated, and reinforcement learning evolved into deep reinforcement learning (DRL).

[0097] Deep reinforcement learning is the process in which an agent takes action A based on state S in an environment E and continuously trains to generate a strategy P in order to obtain the maximum reward R.

[0098] Symbolic Representation in Deep Reinforcement Learning:

[0099] π: Function symbol, representing the policy function. The parameter is often the state returned by the environment, and the output is a specific action. It can be seen that a=π(s)∈A.

[0100] r: The immediate reward obtained after taking action a.

[0101] G: cumulative reward, G t represents the cumulative reward from time t to the end of the action;

[0102] Q π (s,a): state-action value function, sometimes also referred to as action value function, which is the expected value of the cumulative reward expected to be obtained by taking action a and using strategy π at time t in state s.

[0103] V π(s): State value function. This is the expected value of the cumulative reward obtained by taking action a and using strategy π at time t in state s.

[0104] For policy P, there are roughly two methods for DRL to generate policy P, namely policy-based and value-based. These two strategies are not completely mutually exclusive and can be used together.

[0105] Among them, the value-based strategy is based on the action value function Q π (s,a) and state value function V π (s), find the strategy π that maximizes the reward, that is, traverse all states and actions and find the maximum value function Q π (s,a) and V π (s) strategy. The main methods include: Dynamic Programming, Monte-Carlo Methods, Temporal-Difference Learning. Among them, Q-learning and DQN mentioned above are both temporal-difference learning, that is, they use value-based strategies.

[0106] All value-based methods ultimately learn through value functions, which may be related to actions or states. Policy-based methods, on the other hand, use gradient descent to directly learn policies.

[0107] In the value-based method, the policy is π(s), which is only related to the state. We modify the policy through the feedback of the value function and then calculate the value function based on the policy.

[0108] In the policy-based approach, the policy is π(s|a,θ), with a set of parameters θ added. Our goal is no longer to indirectly optimize π(s) based on the Q and V functions, but to directly optimize the policy π by optimizing the parameters θ. θ Optimize.

[0109] To optimize the learning objective, we must have a Loss function. Below we introduce the loss functions in two spaces:

[0110] The Loss function in discrete state is:

[0111]

[0112] Here, S1 represents the initial state, and V1 represents the cumulative reward from the initial state to the end. In other words, since the loss function is fixed at θ throughout each game, simply looking at the initial state S1 tells us how much reward a strategy with this set of parameters will achieve. However, since some games don't make actions deterministic but instead introduce some randomness (for example, for the same state and action, the game itself generates noise to prevent identical outcomes), the loss function here is the expected value.

[0113] What we have to do is to maximize J(θ) by learning this set of parameters θ. This is what the policy-based approach does.

[0114] For the Loss function of continuous space, we have:

[0115]

[0116] in, It is a stationary distribution of π(θ) in the Markov chain, which can be simply understood as J(θ) is the expected value of V under various states.

[0117] The value-based method scores the behavior in each state, like training a critic. Based on the critic's score of the state behavior, the highest score is selected to achieve the optimization result.

[0118] Policy-based methods, on the other hand, ignore the state-action scores and focus on optimizing themselves. They are like an actor: as long as they learn their parameters, they will naturally know which action to take in each state.

[0119] In actual training, policy-based methods also use V(s) or Q(s,a), but when actually selecting actions, they do not rely on these two functions. This is the difference between policy-based methods and value-based methods.

[0120] The Actor-Critic method combines the two, adding a Value-based method on the basis of Policy-based, using the QV function to estimate the cumulative reward, reducing the number of samples and improving training effect and efficiency.

[0121] The Soft Actor-Critic (SAC) algorithm introduces maximum entropy, which requires the entropy of each output action to be maximized when maximizing the expected cumulative reward. It uses maximum entropy to increase exploration. At the same time, the "actor-critic" (AC) framework used in the algorithm uses an online strategy when updating the Q function reward value. That is, the behavior strategy is the same as the target strategy, which has a stable property and solves the instability of the offline strategy in DQN.

[0122] Maximizing entropy ensures strategy randomization, meaning the probability of each action output is as dispersed as possible, rather than concentrated on a specific action. The algorithm also supports automatic adjustment of the entropy temperature coefficient. When the initial temperature coefficient is large, the agent is encouraged to explore. As the agent slowly converges, the temperature coefficient can adaptively decay.

[0123] The SAC algorithm can be divided into two parts: strategy evaluation and strategy improvement:

[0124] In the strategy evaluation part, the state value function Vsoft and the action value function Qsoft are defined as:

[0125] V soft (s t )=E[Q(s t ,a t )]-lgπ(a t |s t )

[0126] Q soft (s t ,a t )=r(s t ,a t )+γE[V_soft(s t+1 )

[0127] In the strategy improvement part, the relative entropy optimization strategy is used. Sampling strategy and reward function: The sampling strategy refers to the selection of input DNN samples in the deep reinforcement learning process, usually the current environment state s t In order to improve the convergence and fitting ability of deep reinforcement learning,

[0128] SAC (Soft Actor-Critic) is a model-based deep reinforcement learning algorithm that combines the advantages of actor-critic algorithms and model-based policy optimization algorithms. SAC can efficiently solve reinforcement learning problems in continuous action spaces and is particularly suitable for problems with high-dimensional state and action spaces.

[0129] The action space in this study includes: adjusting the slit lamp intensity, slit width, slit color, slit lamp diffusion, lamp illumination angle, imaging system angle, microscope left and right movement, microscope front and back movement, microscope lifting and lowering movement, and imaging magnification. Among them, continuous variables such as slit lamp intensity, slit width, and lamp illumination angle are all discretized at fixed intervals.

[0130] Because slit lamp adjustments can involve multiple actions, an action sequence is defined as a temporal combination of actions spanning multiple time steps. The model generates a series of action combinations (i.e., action sequences) through multiple decision-making processes (multiple time steps), gradually approaching the target state. For example: Step 1: Adjust slit width and intensity; Step 2: Correct imaging angle and mechanical position; Step 3: Optimize diffusion and color. Implementation: A policy network dynamically selects actions for each step based on the current state and historical information, ultimately forming a sequence. Without explicitly designing combinations, the reinforcement learning algorithm automatically learns the dependencies between parameters and the temporal logic.

[0131] The status is represented as: the current examination settings of the slit lamp after an action is performed on the slit lamp;

[0132] Feedback is represented by: the diagnostic image obtained based on the current examination settings of the slit lamp;

[0133] Reward function design: During remote slit lamp examinations, doctors need to obtain eye images under a variety of test conditions. Therefore, the examination images must be clear and meet the doctor's requirements for ophthalmological examination conditions. Furthermore, the doctor needs to infer the image under the next examination condition based on the currently examined eye images. Therefore, the reward function design integrates the image clarity requirements, the test meeting the test condition requirements, and the expected adjustment of the slit lamp.

[0134] R t =α1·ImageClarity+α2·ImageRequire+α3·NextRequire

[0135] Among them, ImageClarity represents the clarity score of the current image, ImageRequire represents the degree to which the current examination image meets the examination requirements, and NextRequire represents the degree to which the motion direction of the current examination image meets the expectations compared with the previous time step;

[0136] In some embodiments, the clarity score of an image is obtained by training a clarity scoring network using a contrastive learning network based on images with labeled clarity. ImageRequire is a network for scoring the degree of satisfaction of diagnostic requirements, trained using deep learning based on a set of labeled images. NextRequire is a score for the action taken compared to the state of the previous time step. The action score is derived based on history:

[0137] During its interaction with the environment, the agent receives many observations. For each observation, it takes an action and receives a reward. So history is a sequence of observations, actions, and rewards:

[0138] H t =O1,R1,A1,…,A t-1 ,O t ,R t

[0139] The agent depends on the history it has obtained before when taking the current action, so the entire current state can be regarded as a function of this history:

[0140] S t =f(H t )

[0141] When, after multiple time steps of action, the diagnostic image obtained by the slit lamp is in the next state that meets the diagnostic requirements, the reward NextRequire for the action in this history is higher, and its weight is higher than the weight of clarity and the weight of satisfying the diagnostic requirements;

[0142] In some embodiments, α1 = 0.15, α2 = 0.3, and α3 = 0.55.

[0143] In some embodiments, NextRequire is obtained based on the consistency level between the current action and the action path of the next state that meets the diagnosis requirement.

[0144] In some embodiments, we use the DQN network to complete the training of the adaptive strategy:

[0145] First, the continuous changes in the action space, such as angles, are discretized into 1-degree angles to adapt to the discrete action space of DQN.

[0146] Now our choice of action is decided by DQN, so we need to train DQN to achieve satisfactory performance.

[0147] First, the output of our DQN is the Q value that can be obtained after executing each action in a given state. However, in many cases, we do not know what the optimal Q value is. We use rewards to obtain the optimal Q value:

[0148] That is, for state s, if action a is performed, the reward obtained is certain. Therefore, starting from the reward, the problem of comparing the predicted Q value and the true Q value is transformed into the problem of making the model actually fit the reward.

[0149] The loss function of the model is expressed as:

[0150]

[0151] DQN consists of two deep neural networks with the same structure but different parameters, i.e. the weights and biases of each layer in the neural network. And the other is θ i In each iteration, we update θ i Without updating And it is stipulated that after each C step, And its The network where it is located is called the target network, θ i The network is the Q network.

[0152] The problem now becomes that we need a set of training sets, which can provide a batch of four-tuples (s, a, r, s'), where s' is the next state after s executes a.

[0153] Cache Pool: Since DQN uses a cache pool training network, training is performed by sampling a batch of four-tuples (s, a, r, s') in each batch. Therefore, we need to cache a batch of these four-tuples in the experience pool for training. Since each action can transition to the next state and obtain a reward, we can obtain a four-tuple after each action, and we can also directly store this four-tuple in the experience pool.

[0154] Step 1: Obtain an initial robotic arm motion and an initial state corresponding to the initial motion, wherein the initial state is a current examination setting of the slit lamp;

[0155] Step 2: Reinforcement learning takes the next action in the action space, which includes adjusting the slit lamp intensity, slit width, slit color, slit lamp diffusion, lamp illumination angle, imaging system angle, microscope left and right movement, microscope front and back movement, microscope lifting movement, and imaging magnification;

[0156] Step 3: After the action is executed, the slit lamp switches to the next examination setting and an updated examination image is obtained based on the current examination setting.

[0157] Step 4: Obtain the reward function of the action based on the updated inspection image, and obtain the adaptive control strategy of the robotic arm through training iterations of the reward function.

[0158] Among them, the inspection setting is a state vector composed of the current slit lamp intensity, slit width, slit color, slit lamp diffusion, lamp irradiation angle, imaging system angle, microscope coordinates (x, y, z coordinates) and imaging magnification.

[0159] In some embodiments, DQN uses a buffer pool to obtain a batch of quads, and obtains the adaptive control strategy after the iteration is completed.

[0160] When the remote doctor operates the slit lamp, the control unit at the proximal instrument side acts according to the instructions received from the doctor and adjusts to the next examination position based on the adaptive strategy. After the adaptive adjustment, the doctor can also make fine adjustments based on the current status.

[0161] In some embodiments, the positions of adaptive strategy adjustments and the doctor's fine-tuning operations are recorded, and the model autonomously learns and optimizes.

[0162] In some embodiments, a remote slit lamp diagnosis system: the control instrument on the doctor's side encodes the instrument operation into an audio signal and transmits it to the interactive platform, which transmits it to the instrument side. The driving device on the instrument side decodes the audio signal and drives the slit lamp for inspection. The camera captures the image and transmits it to the interactive platform, which sends it to the display on the doctor's side to present the inspection image.

[0163] In some embodiments, a remote slit lamp diagnosis system: the doctor at the doctor's end controls the movement of the slit lamp at the instrument end through the controller unit, and the control electrical signal is converted into an audio signal and then transmitted to the interface unit, which is transmitted via a dedicated network to the instrument end. The interface unit at the instrument end transmits the received audio signal to the core control unit. The core control unit also includes decoding the received doctor's control signal into a control instruction, driving the motor to move, so that the mechanically bound slit lamp examination instrument performs a corresponding action to perform the examination on the patient, and the electronic eyepiece on the microscope transmits the captured video image to the instrument end interface unit, inputs the input into the interactive platform, and then displays it on the doctor end interface unit.

[0164] In some embodiments, the proximal execution end includes an image encoder, and the diagnostic image is sent to the remote control end through the video channel of the proximal interface unit after encoding; the remote control end includes an image decoder, and the video channel of the distal interface unit receives the encoded image and decodes it to obtain the diagnostic image; the image encoder and decoder are trained based on the self-encoding task of slit lamp diagnostic images.

[0165] In some embodiments, the image encoder is obtained by training a convolutional neural network, and the doctor reconstructs the image through a decoder for the doctor to view and make a diagnosis. At the same time, the encoder's features are input into the auxiliary diagnosis system to output auxiliary diagnosis results, such as conjunctivitis, dry eye, glaucoma, keratitis, cataracts, eye trauma, and eye tumors. Among them, the incidence of diseases such as iritis, eye trauma, and eye tumors is relatively low; keratitis, iritis, and dry eye are relatively common eye diseases, and conventional cross-entropy loss is difficult to effectively distinguish eye diseases; the present invention adopts a weighted focus loss that integrates image features. By combining the scoring standard of slit lamp microscopy and weighted focus loss, the loss function not only takes into account the category imbalance problem of various ophthalmic diseases, but also enhances the consistency of the model with the slit lamp prior knowledge through KL divergence, which helps to more accurately predict ophthalmic diseases and provide higher classification accuracy. The calculation method of the loss function is expressed as:

[0166]

[0167] Where L is the loss function of the convolutional neural network, which is used to guide the training of the convolutional neural network;

[0168] λ i =1+Score i / max(Score) is the weight of the i-th sample, Score i is the slit lamp microscopy score standard of the i-th sample, and max(Score) represents the maximum value of the slit lamp microscopy score standard;

[0169] p i is the highest probability value of ophthalmic diseases predicted based on the features of the diagnostic image,

[0170] γ is the focus factor, preferably set to 2 to reduce the loss contribution of easy-to-classify samples;

[0171] q i is the probability distribution of eye diseases predicted by the convolutional neural network for the i-th sample;

[0172] r i is the prior distribution of ophthalmic diseases based on image features of the i-th sample;

[0173] α is the weight coefficient of the KL divergence term;

[0174] D KL (·||·) is the KL divergence term, which is used to quantify the difference between the predicted distribution and the prior distribution, forcing the network prediction probability to be aligned with the distribution of radiomics scores, thereby enhancing the consistency of the model with domain knowledge.

[0175] In some embodiments, the slit lamp microscopy score is obtained by a doctor's marking based on a slit lamp examination evaluation index grading standard.

[0176] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architectures, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the boxes can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified functions or operations, or can be implemented using a combination of dedicated hardware and computer instructions.

[0177] In general, various example embodiments of the present disclosure may be implemented in hardware or dedicated circuitry, software, firmware, logic, or any combination thereof. Certain aspects may be implemented in hardware, while other aspects may be implemented in firmware or software that may be executed by a controller, microprocessor, or other computing device. When various aspects of the disclosed embodiments are illustrated or described as block diagrams, flow charts, or using some other graphical representation, it will be understood that the blocks, devices, systems, techniques, or methods described herein may be implemented, as non-limiting examples, in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or a controller or other computing device, or some combination thereof.

[0178] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0179] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0180] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0181] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0182] The exemplary embodiments of the present disclosure described in detail above are merely illustrative and not restrictive. Those skilled in the art will appreciate that various modifications and combinations may be made to these embodiments or their features without departing from the principles and spirit of the present disclosure, and such modifications should fall within the scope of the present disclosure.

Claims

1. An ophthalmic disease diagnosis system based on a remote slit lamp, characterized in that: The system includes a remote control end and a near-end execution end, which are connected to each other through a network. The remote control end includes: an acquisition unit, a remote signal conversion unit, and a remote interface unit; the near-end execution end includes: a near-end interface unit, a near-end signal conversion unit, an execution unit, and a camera unit. The system performs the following method: S1: The remote control end acquisition unit acquires the control instruction of the electrical signal and inputs it into the remote signal conversion unit, which encodes the control instruction into a corresponding audio signal and sends it to the local execution end through the audio channel of the remote interface unit; S2: The proximal execution end receives the audio signal through the audio channel of the proximal interface unit, and the proximal signal conversion unit decodes the audio signal into an electrical signal control instruction. Based on the control instruction execution unit, the slit lamp is driven to perform an action corresponding to the control instruction, and the camera unit captures an ophthalmic examination image. The method executed by the system further includes: the proximal execution end further includes an adaptive control unit, the adaptive control unit automatically executing a set of control instructions based on an adaptive control strategy according to the received control instructions to adjust the slit lamp to a next examination state, and the method for obtaining the adaptive control strategy includes: Step 1: Obtain an initial robotic arm motion and an initial state corresponding to the initial motion, wherein the initial state is a current examination setting of the slit lamp; Step 2: Reinforcement learning takes the next action in the action space, which includes adjusting the slit lamp intensity, slit width, slit color, slit lamp diffusion, lamp illumination angle, imaging system angle, microscope left and right movement, microscope front and back movement, microscope lifting movement, and imaging magnification; Step 3: After the action is executed, the slit lamp switches to the next examination setting and obtains an updated examination image based on the current examination setting; Step 4: Based on the updated inspection image, a reward function for the action is obtained. The adaptive control strategy of the robotic arm is obtained through iterative training of the reward function. The reward function is: ;in, Indicates the clarity score of the current state of the image, Indicates the degree to which the current state of the diagnostic image meets the diagnostic requirements. Indicates the degree to which the motion direction of the current inspection image is consistent with the expectation compared with the previous time step; 、 and They are 、 and NextRequire is the score of the action taken compared to the state of the previous time step. The action score is derived based on history, which is a sequence of observations, behaviors, and rewards. The reward of the action in this history is The score is relatively high, and its weight is higher than the weight of the clarity score and the weight of the score of the degree to which the diagnosis meets the requirements.

2. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 1, characterized in that: The method performed by the system further includes: S3: sending the ophthalmological examination image to the remote control terminal through the video channel of the proximal interface unit, S4: The remote control terminal receives the diagnosis image through the video channel of the remote interface unit, and obtains an updated control instruction from the acquisition unit according to the diagnosis image; S1-S4 are repeated until the remote control terminal receives the required diagnosis image for ophthalmic disease examination, and the ophthalmic disease diagnosis is performed based on the diagnosis image.

3. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 2, characterized in that: The proximal execution end includes an image encoder, and the diagnostic image is sent to the remote control end through the video channel of the proximal interface unit after being encoded; the remote control end includes an image decoder, and the video channel of the distal interface unit receives the encoded image and decodes it to obtain the diagnostic image; the image encoder and decoder are trained based on the self-encoding task of slit lamp diagnostic images.

4. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 1, characterized in that: The acquisition unit uses the controller handle to acquire control instructions for the electrical signals of the handle operation.

5. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 4, characterized in that: The remote signal conversion unit is built into the controller handle.

6. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 4, characterized in that: The remote signal conversion unit is connected to the controller handle and is connected to the computer through the audio interface of the computer at the remote control end.

7. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 1, characterized in that: The control instructions or the actions corresponding to the control instructions include adjusting any one or more of the following: slit intensity, slit width, slit color, slit angle, slit light diffusion, slit lamp irradiation angle, imaging system angle, left and right movement of the microscope, front and back movement of the microscope, lifting and lowering movement of the microscope, and imaging system magnification.

8. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 7, characterized in that: Encoding the control instruction into an audio signal means using different audio frequencies to represent different control instructions, and different control instructions correspond to actions of corresponding instrument ends.

9. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 1, characterized in that: The remote interface unit and the near-end interface unit are two computers connected via a network.

10. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 1, characterized in that: The camera unit of the proximal execution end includes a slit lamp microscope inspection camera and a monitoring camera. The slit lamp microscope inspection camera is used to capture the inspection image of the microscope, and the monitoring camera is used to capture the instrument operation status image when the slit lamp microscope is performing inspection; the proximal execution end transmits the inspection image and the instrument operation status image to the remote control end through the proximal interface unit for display.

11. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 10, characterized in that: The slit lamp microscope inspection camera is an electronic eyepiece.

12. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 10, characterized in that: The proximal execution end also includes a video unit, which merges the inspection image of the microscope and the instrument operation status image as the main screen and secondary screen of the picture-in-picture respectively into one image and transmits it to the instrument end interface unit, and transmits the inspection image to the remote control end through the video interface; the remote control end acquisition unit also obtains the switching instruction for the main screen and secondary screen of the picture-in-picture and transmits it to the remote interface unit, the remote interface unit transmits the switching instruction and audio signal to the proximal interface unit and then separates the audio signal and the switching instruction, the video unit sets the main screen of the original picture-in-picture as the secondary screen of the new picture-in-picture according to the received switching instruction, sets the main and secondary screens of the original picture-in-picture to the main screen of the new picture-in-picture, and then transmits the new picture-in-picture to the proximal interface unit and then sends it to the remote control end.

13. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 1, characterized in that: The acquisition unit also includes a microphone module for acquiring the doctor's voice communication signal to the patient, and transmitting the voice communication signal to the remote interface unit. The remote interface unit of the control end transmits the reverberation signal obtained by reverberating the audio signal and the voice communication signal to the proximal interface unit. The proximal interface unit separates the reverberation signal to obtain an audio signal and a voice communication signal, and the voice communication signal is played through the voice playback device.

14. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 12, characterized in that: The remote control end also includes a camera unit for capturing the doctor's communication with the patient, transmitting the communication image to the remote interface unit, transmitting the communication image to the instrument end interface unit via the data interface, and the instrument end interface unit transmitting the communication image to the instrument end display for display.

15. The ophthalmic disease diagnosis system based on remote slit lamp according to claim 12, characterized in that: The remote control terminals include multiple terminals; one of the multiple remote control terminals is a main control terminal used to control the actions of the near-end execution terminal, and the near-end execution terminal transmits the picture-in-picture to multiple remote control terminals for display, and the multiple remote control terminals are connected to each other through video conferencing.

Images