Intelligent visual prosthesis system and coding method

Abstract

This invention discloses an intelligent visual prosthesis system and encoding method. The system includes an external sensing processing unit and an internal implanted stimulation unit. The external sensing processing unit includes at least an environment sensing unit and a first data processing unit, and the internal implanted stimulation unit includes at least a stimulation signal generation unit and an electrode array. The environment sensing unit is used to acquire a target image. The first data processing unit is used to match the current scene to a preset lighting scene mode according to the image features or ambient illuminance of the target image, independently configure anchor points for the corresponding image grayscale value boundaries for the matched lighting scene mode, and generate a grayscale-charge density mapping relationship corresponding to the lighting scene mode. Then, the grayscale value of each pixel of the target image is mapped to the corresponding target charge density, and a stimulation command is generated and output to the stimulation signal generation unit. The stimulation signal generation unit is used to generate an electrical stimulation pulse sequence according to the stimulation command and a preset reference pulse width and output it to the electrode array.

Description

Technical Field

[0001] This invention relates to the field of intelligent medical technology, specifically to an intelligent visual prosthesis system and its encoding method. Background Technology

[0002] Diseases such as retinitis pigmentosa and age-related macular degeneration lead to the degeneration and blindness of photoreceptor cells, but some of the inner retinal nerve cells (such as bipolar cells and ganglion cells) often remain. Artificial retinal prostheses (such as Argus II) acquire images through a camera, converting them into electrical stimulation signals to stimulate the remaining ganglion cells, helping patients regain some light perception. However, current technologies suffer from a fundamental flaw in the systematic engineering of natural vision reconstruction: static visual encoding, resulting in a severe disconnect from the environment. The visual encoding and decoding strategies of existing visual prosthesis systems are mostly static linear mappings, directly and fixedly converting video brightness (or image grayscale values) into stimulation current amplitudes. This approach completely ignores the core function of the human retina—light and dark adaptation. A healthy retina can dynamically adjust its sensitivity and response range when faced with drastic changes in ambient light, from starlight to sunlight, through mechanisms evolved over hundreds of millions of years. Current retinal prostheses lack this ability, causing a sharp decline in visual perception quality in changing lighting environments: details are lost in darkness, and bright areas appear as overexposed white light, making it impossible to form stable and usable vision. Summary of the Invention

[0003] Based on the above background, this invention provides an intelligent visual prosthesis system and encoding method that can simulate retinal adaptive function, ensure long-term dynamic safety, and realize at least one function of personalized visual perception. Specifically, the following technical solution is adopted:

[0004] The first aspect of this invention discloses an intelligent visual prosthesis system, including an external sensing and processing unit and an internal implanted stimulation unit connected in communication. The external sensing and processing unit includes at least an environmental sensing unit and a first data processing unit, and the internal implanted stimulation unit includes at least a stimulation signal generation unit and an electrode array.

[0005] The environmental sensing unit is used to acquire target images;

[0006] The first data processing unit is used to match the current scene to a preset lighting scene mode based on the image features of the target image or the ambient illuminance obtained by the environment perception unit, independently configure anchor points for the corresponding image gray value boundaries for the matched lighting scene mode, and generate a gray-charge density mapping relationship corresponding to the lighting scene mode based on the configured anchor points, thereby mapping the gray value of each pixel of the target image to the corresponding target charge density, generating a stimulus instruction and outputting it to the stimulus signal generation unit.

[0007] The stimulation signal generation unit is used to generate an electrical stimulation pulse sequence according to the stimulation command and a preset reference pulse width and output it to the electrode array.

[0008] Furthermore, the data processing unit extracts image features based on the target image and matches lighting scene patterns based on a preset value range according to the extracted image features.

[0009] Furthermore, the implanted stimulation unit also includes an impedance monitoring and compensation unit;

[0010] The stimulation signal generation unit is used to calculate the required ideal stimulation current amplitude based on the stimulation command and the preset reference pulse width.

[0011] The impedance monitoring and compensation unit is used to measure the contact impedance of the current electrode and calculate the compensated current amplitude based on a preset impedance-efficiency compensation function.

[0012] The stimulation signal generation unit is used to generate an electrical stimulation pulse sequence based on the compensated current amplitude and output it to the electrode array.

[0013] Furthermore, independently configuring anchor points for the corresponding image grayscale value boundaries for the matched lighting scene mode includes:

[0014] Configure a black point anchor point, which determines the minimum target charge density value corresponding to the image grayscale value of 0 in the current lighting scene mode. From dark lighting scene mode to bright lighting scene mode, the minimum target charge density value corresponding to this anchor point transitions from the minimum effective value higher than the physiological threshold to 0.

[0015] Configure a white anchor point, which determines the maximum target charge density value corresponding to the current lighting scene mode of the image grayscale value 255. This maximum target charge density value increases as the ambient illuminance increases, and the maximum value is less than the absolute safety upper limit.

[0016] Furthermore, the grayscale-charge density mapping relationship corresponding to the lighting scene mode is generated based on the configured anchor points, including:

[0017] Using black and white anchor points as boundary anchor points, a nonlinear function is used to generate the basic gray-scale-charge density mapping curve for the corresponding lighting scene mode.

[0018] Real-time brightness histogram analysis is performed on the current target image. If the overall brightness of the image is lower or higher than a preset threshold, the basic gray-level-charge density mapping curve is finely adjusted globally or locally to generate a gray-level-charge density mapping lookup table.

[0019] Furthermore, generating an electrical stimulation pulse sequence based on the compensated current amplitude includes:

[0020] Calculate the charge density of the planned electrical stimulation pulse sequence based on the compensated current amplitude:

[0021] The calculated charge density of the planned electrical stimulation pulse sequence is compared with the preset personalized safety threshold of the current electrode:

[0022] If the value is less than the personalized safety threshold, it is considered safe, and a discharge stimulation pulse sequence is generated according to the calculated charge density.

[0023] If the charge density exceeds the personalized safety threshold, dynamic safety control is triggered to ensure that the charge density of the actual electrical stimulation pulse sequence is less than the personalized safety threshold of the current electrode.

[0024] Furthermore, dynamic safety control includes:

[0025] Based on the constant charge density optimization algorithm, the pulse width of the electrical stimulation pulse sequence is adjusted preferentially within the allowable range;

[0026] If the adjusted charge density still cannot meet the personalized safety threshold of the current electrode, the magnitude of the stimulation current signal is limited proportionally so that the charge density of the actual electrical stimulation pulse sequence is less than the personalized safety threshold of the current electrode.

[0027] Furthermore, independently configuring anchor points for the corresponding image grayscale value boundaries for the matched lighting scene mode, and generating the grayscale-charge density mapping relationship corresponding to the lighting scene mode based on the configured anchor points, also includes:

[0028] The charge density perception threshold and comfort limit of a single electrode are obtained based on psychophysical parameter measurements.

[0029] The charge density perception threshold and comfort upper limit are used as calibration references to calibrate the black anchor point and white anchor point;

[0030] By testing multiple stimulation levels, the charge density-subjective brightness function of the current electrode is fitted, and the basic gray-scale-charge density mapping curve is locally corrected based on this function, and a gray-scale-charge density mapping lookup table is generated.

[0031] Furthermore, using the charge density perception threshold and comfort upper limit as calibration references, the calibration of the black dot anchor point and white dot anchor point includes:

[0032] Each electrode was initially stimulated starting from zero charge density, and the working potential or visual evoked potential generated by the electrode during the initial stimulation was detected.

[0033] The minimum charge density threshold that can generate the correct action potential or visual evoked potential is mapped to grayscale 0.

[0034] Continue to increase the charge density until the potential amplitude no longer increases, then map the maximum charge density value at that moment to a grayscale value of 255.

[0035] The minimum charge density threshold is used as the calibration reference for the black spot anchor point of the electrode, and the maximum charge density value is used as the calibration reference for the white spot anchor point.

[0036] Furthermore, the external sensing processing unit also includes a second data processing unit, which is used to determine whether an ON event or an OFF event is triggered based on the gray value change between the target image and the previous frame image, and synthesize a mixed stimulation instruction containing ON type stimulation and / or OFF type stimulation based on the stimulation instruction according to the determination result and output it to the stimulation signal generation unit.

[0037] Among them, the ON type stimulation executes the electrical stimulation pulse sequence generated by the stimulation signal generation unit, and the OFF type stimulation includes at least a background pulse below the perception threshold and at least a no-stimulation interval.

[0038] A second aspect of this invention discloses an intelligent visual prosthetic encoding method, comprising the following steps:

[0039] Acquire target images;

[0040] Based on the image features of the target image or the ambient illuminance obtained by the environment perception unit, the current scene is matched to a preset lighting scene mode. Anchor points corresponding to the gray value boundaries of the image are independently configured for the matched lighting scene mode. Based on the configured anchor points, a gray-charge density mapping relationship corresponding to the lighting scene mode is generated. Then, the gray value of each pixel of the target image is mapped to the corresponding target charge density, and a stimulus command is generated.

[0041] An electrical stimulation pulse sequence is generated based on the stimulation command and a preset reference pulse width and output to the electrode array.

[0042] The beneficial effects of this invention are as follows:

[0043] 1) Strong environmental robustness and high visual quality: The system of this invention simulates the light and dark adaptation of biological vision through adaptive coding, enabling patients to obtain stable visual perception with natural contrast and rich details in different lighting environments, which greatly expands the applicable scenarios of the device and the user experience.

[0044] 2) Long-term safety and effectiveness are fundamentally improved: By using impedance feedback closed loop, the stimulation safety is changed from "open-loop preset" to "dynamic protection and control", which effectively prevents the risks caused by changes in the biological interface and provides key technical guarantee for lifelong implantation.

[0045] 3) Personalized and precise visual restoration is achieved: Through optional response calibration process, the system of the present invention can be adapted to the unique neural pathway characteristics of patients, so that artificial vision is more in line with the "expectations" of their brains, laying the foundation for restoring high-quality and understandable visual perception.

[0046] 4) Complete technology chain and high degree of systematization: This invention provides a complete solution from light signal sensing and information processing to safe electrical stimulation generation. Each module works together based on a unified charge density quantification model. The design is clear, the controllability is good, and it has important industrialization and clinical translation value.

[0047] 5) Improved contour and motion perception: By simulating ON / OFF antagonism, patients can perceive the edges of objects and changes in brightness more clearly, greatly improving their understanding of shape and direction of motion.

[0048] 6) Strong hardware compatibility: It does not require changes to the physical structure and implantation location of existing implantable electrode arrays. It can be achieved simply by upgrading the external processor algorithm and stimulator firmware, which facilitates clinical translation and upgrading of existing equipment.

[0049] 7) Reduced power consumption and risk of tissue damage: OFF type stimulation uses "stimulus pause" coding, which only requires maintaining low-frequency background stimulation in most static scenarios. Compared with using high amplitude / high frequency stimulation throughout, it is more energy-efficient and may reduce tissue adaptive fatigue or damage caused by long-term strong stimulation.

[0050] 8) Personalized adaptation: The parameters of this coding strategy (threshold, time constant, intensity mapping curve) can be finely electrophysiologically calibrated and adaptively learned and optimized according to the patient's individual residual neurological function, so as to maximize the individualized therapeutic effect. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the overall hardware architecture and data flow of an embodiment of the retinal prosthesis system of the present invention.

[0052] Figure 2 This is a schematic diagram comparing grayscale-charge density mapping curves under different lighting scene modes in an embodiment of the present invention.

[0053] Figure 3 This is a schematic diagram of the dynamic gray-scale-charge density mapping method for ambient light adaptation in an embodiment of the present invention.

[0054] Figure 4 (a) and Figure 4 (b) is a comparison of the visual image obtained by retinal stimulation without using the adaptive coding method and the visual image obtained by retinal stimulation using the adaptive coding method provided by the system of the present invention.

[0055] Figure 5 This is a schematic diagram of different regions where stimulation is applied to simulate the ON / OFF pathway function of the retina in an embodiment of the present invention.

[0056] Figure 6 This is a schematic diagram of the stimulation signal of the OFF type stimulus in an embodiment of the present invention.

[0057] Figure 7 This is a schematic diagram of the structure of the external sensing processing unit in an embodiment of the present invention.

[0058] Figure 8 This is a schematic diagram of the structure of the implanted stimulation part in an embodiment of the present invention. Detailed Implementation

[0059] Embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the invention. It should be understood that the accompanying drawings and embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the invention.

[0060] This invention discloses an intelligent visual prosthetic system, comprising an external sensing and processing unit and an internally implanted stimulation unit connected by communication. The external sensing and processing unit includes at least an environmental sensing unit and a first data processing unit, and the internally implanted stimulation unit includes at least a stimulation signal generation unit and an electrode array; wherein,

[0061] The environmental sensing unit is used to acquire target images;

[0062] The first data processing unit is used to match the current scene to a preset lighting scene mode based on the image features of the target image or the ambient illuminance obtained by the environment perception unit. It independently configures the anchor points of the corresponding image gray value boundary for the matched lighting scene mode, and generates a gray-charge density mapping relationship corresponding to the lighting scene mode based on the configured anchor points. Then, it maps the gray value of each pixel of the target image to the corresponding target charge density, generates a stimulus instruction, and outputs it to the stimulus signal generation unit.

[0063] The stimulation signal generation unit is used to generate an electrical stimulation pulse sequence according to the stimulation command and the preset reference pulse width and output it to the electrode array.

[0064] See Figure 1 In some implementations, the external sensing processing unit can be set up independently or integrated into, for example, Figure 7In the external wearable device or apparatus shown, the environmental sensing unit may include an ambient light sensor 102 and a camera 101, and the first data processing unit may be a main processor 103 built based on a main control computer or an embedded processor. The implantable stimulation part is one that can be implanted inside the human eyeball, such as... Figure 8 The components shown include a stimulation signal generation unit comprising a microcontroller (ASIC) 106 and a multi-channel stimulation generator 108 integrated on component 11, and an electrode array of a microelectrode array 109. Both are connected via a wireless transmission module 104 configured on the external sensing processing unit and a wireless receiving and power management module 105 configured within the implanted stimulation unit (also integrated on component 11). The wireless receiving and power management module 105 also manages the power distribution of the implanted stimulation unit, which will not be described in detail here.

[0065] It should be noted that in this embodiment of the invention, the camera in the environmental perception unit acquires a continuous sequence of video frames, while the "target image" mentioned in the preceding and following text generally refers to a single video frame image, especially the video frame image corresponding to the current processing cycle. After processing each frame image in the continuous video frame sequence based on the method of this invention, a continuous stimulus signal corresponding to the content acquired by the camera can be obtained.

[0066] In some implementation schemes, the preset lighting scene modes may include at least two lighting scene modes, preferably including three lighting scene modes: strong light scene, weak light scene and normal scene. They may also be classified into more modes according to actual needs, without specific limitations here.

[0067] In some implementations, lighting scene patterns can be matched using real-time captured video frame images (i.e., target images), specifically using the following methods:

[0068] 1) Frame sampling: No need to process every frame. Frame skipping can be fixed (e.g., 5 frames per second) or dynamically adjusted to reduce the amount of data processed and improve processing speed.

[0069] 2) Reduce resolution: Reduce the frame size to 160x200 (the video frame resolution can be an integer multiple of the number of electrode arrays) to significantly reduce the computational load and have minimal impact on illumination statistics.

[0070] 3) Fast feature extraction: Only calculate the grayscale mean, overexposure ratio, underexposure ratio, and standard deviation (which can be ignored if necessary).

[0071] 3.1 Extracted features:

[0072] Mean: Overall brightness level;

[0073] Overexposed Ratio: The percentage of pixels with a grayscale value greater than 250, reflecting the degree of highlight cropping.

[0074] Underexposed Ratio: The percentage of pixels with a grayscale value less than 5, reflecting the degree of shadow cropping;

[0075] Standard deviation (StdDev): Optional, used to distinguish low-contrast scenes (such as foggy days, extremely dark environments).

[0076] 3.2 Judgment Rules (Three-Level Decision-Making):

[0077] A stratified approach is adopted, prioritizing extreme proportions, then combining this with the mean, and finally fine-tuning with the standard deviation to ensure a balance between speed and accuracy.

[0078] Strong light scene (M_bright): If the overexposure ratio is >10% or the average value is >200, it is judged as a strong light scene.

[0079] Low-light scene (M_dark): If the underexposure ratio is >10% or the average value is <50, it is judged as a low-light scene.

[0080] Normal scenario (M_normal): If the above conditions are not met, then it enters the normal scenario range. If the standard deviation is calculated at this time, it can be further subdivided into "normal to strong / too weak".

[0081] If the standard deviation is extremely low (< 30) and the mean is in the middle (50~200), it may be a foggy day or uniform lighting, and is still classified as a normal scene.

[0082] 4) Temporal smoothing: The classification results are smoothed using exponential moving average (EMA) or sliding window, and a majority voting mechanism is used to avoid misjudgment in a single frame, forming a stable lighting scene pattern matching.

[0083] In other implementations, lighting scene patterns can also be matched based on the ambient illuminance obtained by the environmental sensing unit, specifically using the following methods:

[0084] An ambient light sensor acquires the ambient illuminance in real time and matches it to a preset lighting scene mode based on a preset threshold range. In this example, the acquired ambient illuminance E is matched to three lighting scene modes from dark to bright: M_dark, M_normal, and M_bright.

[0085] It should be noted that in other implementation schemes, fuzzy logic or continuous functions can also be used to match lighting scene patterns in order to achieve a smoother pattern transition.

[0086] After obtaining the lighting scene mode, anchor points for the corresponding image grayscale value boundaries can be configured independently for the matched lighting scene mode, using the following method:

[0087] 1) Configure a black point anchor. This anchor determines the minimum target charge density value CD_0(M_x) corresponding to the image grayscale value of 0 in the current lighting scene mode. From dark lighting scene mode to bright lighting scene mode, the minimum target charge density value corresponding to this anchor transitions from a minimum effective value higher than the physiological threshold to 0. For example, under M_dark, CD_0(M_dark) can be set to a minimum effective value slightly higher than the physiological threshold to ensure that "black" can still be faintly perceived in a dark environment; under M_bright, CD_0(M_bright) can be set to 0, because ambient light itself is sufficient to define "black," and for a blind person, no stimulation is black.

[0088] 2) Configure the white point anchor point. This value is defined as the "maximum stimulus of the target" in the current scene and must satisfy CD_255(M_x) < CD_safe_max (absolute safety upper limit). Its value increases with the enhancement of ambient light, i.e., CD_255(M_bright) > CD_255(M_dark), so that the "white" of the optical illusion still has sufficient perceived brightness in bright environments, which has a stretching effect in bright environments and enhances contrast.

[0089] In some implementations, a grayscale-charge density mapping relationship corresponding to the lighting scene mode is generated based on the configured anchor points, specifically using the following method:

[0090] First, using the black anchor point (0, CD_0(M_x)) and white anchor point (255, CD_255(M_x)) configured above as boundary anchor points, the basic gray-scale-charge density mapping curve LUT_base(M_x) under the corresponding lighting scene mode is generated using nonlinear functions (such as S-curves and gamma curves).

[0091] Then, real-time brightness histogram analysis is performed on the current target image. If the image is too dark or too bright overall, the basic gray-level-charge density mapping curve LUT_base(M_x) is globally or locally fine-tuned to generate the final gray-level-charge density mapping lookup table LUT_final. This process maps the gray value G_i of each input pixel in the target image to a preliminary target charge density CD_target_i, generating stimulus instructions.

[0092] After the above mapping is completed, the external sensing and processing unit can send the stimulation command containing the preliminary target charge density CD_target_i to the implanted stimulation unit in vivo via wireless communication.

[0093] Figure 2 The mapping strategies under different lighting scene modes are compared intuitively.

[0094] In the figure, the horizontal axis represents the grayscale value of the input image (0-255), and the vertical axis represents the target charge density (0 to the maximum safe value CD_safe_max).

[0095] The three curves are as follows:

[0096] Curve D (dark environment, corresponding to M_dark): CD_0_D > 0 (black points have slight stimulation), CD_255_D value is low, the curve is steep in the low grayscale area, emphasizing the enhancement of dark details.

[0097] Curve N (normal environment, corresponding to M_normal): CD_0_N is close to 0, CD_255_N is a medium value, and the curve is close to linear.

[0098] Curve B (bright environment, corresponding to M_bright): CD_0_B = 0 (no stimulation from black points), CD_255_B value is very high (close to CD_safe_max), the curve is steep in the high grayscale area, ensuring that "white" is bright enough.

[0099] The gray horizontal line above represents CD_safe_max, and the CD_255 point of all curves is located below it to ensure safety.

[0100] The following is combined Figure 3 Here is a specific example illustrating the steps described above:

[0101] ① Begin processing the new frame in the video frame sequence captured by the camera, i.e., the target image Img;

[0102] ②Simultaneously acquire the current ambient light intensity E and the target image Img;

[0103] ③ Match the lighting scene mode M_x (e.g., dark, normal, bright);

[0104] ④ Based on the lighting scene mode M_x, read the pre-stored key mapping parameters: black anchor point CD_0(M_x) (grayscale 0 corresponds to charge density) and white anchor point CD_255(M_x) (grayscale 255 corresponds to charge density);

[0105] ⑤ Using (0, CD_0) and (255, CD_255) as anchor points, generate the base mapping curve LUT_base;

[0106] ⑥ Analyze the grayscale histogram of the target image Img to obtain its overall brightness and contrast characteristics;

[0107] ⑦ Based on image features, fine-tune LUT_base. For example, if the image is generally dark, appropriately increase the mapping slope of the low grayscale area to generate the final mapping table LUT_final.

[0108] ⑧ Apply LUT_final to map the gray value G_i of each pixel / region in the target image Img to the target charge density CD_target_i;

[0109] ⑨ Output the stimulation command containing CD_target_i to the wireless transmission module;

[0110] ⑩ End processing of this frame and enter the next loop.

[0111] After receiving the stimulation command containing the target charge density CD_target_i sent by the external sensing and processing unit, the implanted stimulation unit can calculate the required ideal stimulation current amplitude I_theory_i based on the target charge density CD_target_i and the preset reference pulse width PW_ref, and generate an electrical stimulation pulse sequence to output to the electrode array.

[0112] See Figure 4 , Figure 4 (a) in the image is the visual image obtained by retinal stimulation without using an adaptive coding method. Figure 4 Image (b) shows a visual image obtained by retinal stimulation using the adaptive coding method provided by the system of the present invention. It can be seen that the adaptive coding method provided by the system of the present invention can significantly improve the clarity and resolution of the visual image obtained by stimulation.

[0113] On the other hand, existing visual prosthetic systems suffer from drawbacks such as open-loop stimulation safety and a mismatch with the dynamic changes at the biological interface. Generally, the safety of stimulation is determined by whether the charge density exceeds the tissue damage safety threshold. Existing visual prosthetic systems use preset, constant stimulation parameters (current I, pulse width PW). However, electrode-contact impedance is a crucial dynamic variable that continuously changes with tissue response, electrolyte environment, and electrode electrochemical state. Under a fixed voltage, increased impedance leads to a reduction in the effective charge actually delivered to the neuron, rendering the stimulation ineffective; conversely, decreased impedance may result in excessive charge density, posing a risk of damaging both the neuron and the electrode. Existing visual prosthetic systems lack real-time perception and closed-loop control of this dynamic process, compromising long-term safety and effectiveness.

[0114] To ensure the safety of the stimulation, as a further preferred embodiment, the visual prosthesis system of the present invention further includes an impedance monitoring and compensation unit in its implanted stimulation part for real-time impedance monitoring and compensation. Accordingly, the stimulation signal generation unit calculates the required ideal stimulation current amplitude based on the target charge density and a preset reference pulse width; the impedance monitoring and compensation unit measures the contact impedance of the current electrode and calculates the compensated current amplitude based on a preset impedance-efficiency compensation function; the stimulation signal generation unit generates an electrical stimulation pulse sequence based on the compensated current amplitude and outputs it to the electrode array.

[0115] refer to Figure 1 The impedance monitoring and compensation unit can specifically be an impedance measurement circuit 107. In some implementations, the contact impedance Z_i(t) of the current electrode is measured by the impedance measurement circuit 107 before stimulation is delivered. When the impedance is too high, the effective current acting on the neuron will decrease. The system pre-stores an impedance-efficiency compensation function β(Z) and calculates the compensated current amplitude I_comp_i = I_theory_i / β(Z_i(t)) to counteract the impedance effect and stabilize the electrically evoked response.

[0116] As a preferred implementation, impedance measurement can be performed during the interval of the stimulation pulse or by superimposing subthreshold test signals to achieve dynamic online monitoring.

[0117] Finally, a closed-loop verification and dynamic adjustment of the safety threshold are performed. Specifically, the charge density CD_plan_i = I_comp_i * PW_ref *η (η is the waveform coefficient, which defaults to 1 and can be adjusted via the host computer according to actual conditions) of the planned electrical stimulation pulse sequence is calculated and compared with the personalized safety threshold CD_safe_i of the electrode stored in the system.

[0118] If CD_plan_i ≤ CD_safe_i, it is considered safe, and the stimulus is delivered according to (I_comp_i, PW_ref).

[0119] If CD_plan_i > CD_safe_i, dynamic safety control is triggered: the "constant charge density" optimization algorithm is activated, prioritizing the adjustment of the pulse width PW within the allowable physiological range. If the adjusted charge density CD_plan_i' still cannot satisfy CD_plan_i' ≤ CD_safe_i, the current I is proportionally limited, ultimately ensuring that the actual delivered stimulus has a charge density CD_actual_i = min(CD_target_i, CD_safe_i). This closed loop guarantees absolute safety and strives for effectiveness under any impedance state.

[0120] It should be noted that the aforementioned safety threshold CD_safe_i is not a fixed value, but rather an upper limit of a range that is finely adjusted based on a dynamic risk assessment of the electrode material, size, usage history, and long-term impedance trends.

[0121] The following is a specific example illustrating the steps described above:

[0122] ① Begin processing the received stimulus instruction containing CD_target_i;

[0123] ② Process each electrode i sequentially;

[0124] ③ Calculate the theoretical stimulus current I_theory_i = CD_target_i / (PW_ref * η);

[0125] ④ Measure the current contact impedance Z_i(t) of the electrode in real time;

[0126] ⑤ Perform impedance compensation and calculate the compensated current I_comp_i = I_theory_i / β(Z_i(t));

[0127] ⑥ Calculate the planned charge density CD_plan_i = I_comp_i * PW_ref * η;

[0128] ⑦ Compare with the safety threshold CD_safe_i;

[0129] ⑧ Decision Branch:

[0130] If it is safe (CD_plan_i <= CD_safe_i), then directly output the stimulus parameters (I_comp_i, PW_ref).

[0131] If it is not safe (CD_plan_i > CD_safe_i), then proceed to the "Safety Restriction Sub-process":

[0132] First, try increasing the pulse width PW to the maximum allowable value PW_max, recalculate the current I', and verify the new CD_plan_i'; if it is still not safe, then force the output charge density to be limited to CD_safe_i, and deduce the final I_final and PW_final.

[0133] ⑨ Stimulus delivery: Drive the stimulator with the finalized (I_final, PW_final) parameters.

[0134] ⑩ Process the next electrode until all stimulation commands for the current target image have been executed.

[0135] On the other hand, existing visual prosthetic systems suffer from a crude perceptual mapping that is detached from individual neural response characteristics: the grayscale range of digital images (e.g., image grayscale values ​​from 0 to 255) needs to be mapped to the physiologically effective range of electrical stimulation. Current technologies simply map grayscale 0 to a threshold stimulus and grayscale 255 to a maximum safe stimulus. This approach presents a double problem: first, it fails to consider the significant impact of ambient light on the subjective perception of "black" (grayscale 0) and "white" (grayscale 255); second, it ignores the significant individual differences in electrical evoked responses among different patients and at different sites. Since each electrode has a different perceptual threshold, dynamic range, and saturation characteristics, a uniform mapping inevitably leads to insufficient stimulation of some electrodes and premature saturation of others, resulting in overall visual distortion and an unnatural appearance.

[0136] To overcome the aforementioned problems and improve visual perception quality, as a further preferred embodiment, the visual prosthesis system of the present invention also supports personalized calibration in clinical trial mode. Specifically, when independently configuring boundary anchor points based on image grayscale value boundaries for each lighting scene mode, and generating a grayscale-charge density mapping relationship corresponding to the corresponding lighting scene mode based on the configured boundary anchor points, personalized parameter calibration and optimization operations based on electro-evoked response are also performed. The specific method is as follows:

[0137] First, psychophysical parameters are measured: by stimulating individual electrodes and recording the patient's subjective feedback, the charge density perception threshold CD_th_i at which just beginning to be perceived photic hallucinations is accurately determined, and the upper limit of charge density comfort CD_comfort_i at which maximum comfortable brightness is achieved is determined. Specifically, action potential or visual evoked potential measurements can be used as feedback methods to calibrate each electrode: each electrode is initially stimulated starting from 0 charge density, and the working potential or visual evoked potential generated by the electrode in the initial stimulation is detected. The minimum charge density threshold at which a correct action potential or visual evoked potential is generated is mapped to grayscale 0. Then, the charge density is continuously increased until the potential amplitude no longer increases; at this point, the charge density value is mapped to grayscale 255.

[0138] Then, using the calibrated charge density sensing threshold and comfort upper limit as calibration references, the aforementioned black and white anchor points are calibrated. Specifically, CD_th_i can be used as the calibration reference for mapping the electrode's "black point" to CD_0, and CD_comfort_i can be used as the calibration reference for mapping the "white point" to CD_255. For example, a lower CD_255 is set for a response-sensitive electrode, and a higher CD_255 is set for a response-insensitive electrode.

[0139] Finally, the perceptual model is fitted and the encoding is optimized: by testing multiple stimulus levels, the charge density-subjective brightness function of the current electrode is fitted. This function is used to locally correct the previously fine-tuned basic gray-level-charge density mapping curve, and a final gray-level-charge density mapping lookup table is generated, making the final brightness perception change more consistent with the patient's psychophysical laws, and realizing personalized visual reconstruction of "what you see is what you feel".

[0140] The implementation of the above-mentioned solution of the present invention will be further illustrated below through two specific examples.

[0141] Example 1: Ambient light-based coding and dynamic security control process based on impedance measurement.

[0142] The patient, wearing a device equipped with the system of this invention, enters an indoor corridor (approximately 300 lux). The ambient light sensor measures the illuminance, the system determines the mode to be M_normal, and calls the parameter CD_0_N = 15 μC / cm². 2 CD_255_N = 360 μC / cm 2 The camera captures an image of the door frame in front. The main processor analyzes the image and finds that the area inside the door frame is relatively dark, so it slightly enhances the low grayscale area of ​​the base mapping curve. Finally, the dark areas (grayscale ~30) of the door frame are mapped to CD_target ≈ 60 μC / cm. 2 Map the white wall (grayscale ~240) to CD_target ≈ 330 μC / cm. 2 The command is sent to the implanted stimulation unit inside the body.

[0143] The microcontroller of the implanted stimulation unit receives the CD_target command. For the electrode corresponding to the gate frame, I_theory is calculated. At this time, the impedance measurement circuit measures the electrode impedance Z = 20 kΩ (within the normal range). The compensation function β(20kΩ) = 0.95, and I_comp is calculated. The CD_plan is verified to be approximately 58.5 μC / cm. 2 The value is less than the electrode safety threshold CD_safe=450μC / cm. 2 Stimuli were then delivered according to these parameters. The patient steadily perceived the outline of the door frame.

[0144] When the patient steps outdoors (>20000 lux), the ambient light sensor immediately triggers a mode switch to M_bright, and the parameter changes to CD_0_B=0μC / cm. 2 CD_255_B=435μC / cm 2 The system was remapped, and the target charge density on the white wall was increased to CD_target≈420μC / cm². 2When processing this instruction, the internal unit found that due to physiological changes such as pupil constriction caused by strong light, the electrode impedance was slightly reduced to Z=18kΩ. After compensation, CD_plan≈426μC / cm 2 The brightness was still below CD_safe, so it passed safely. The patient perceived the wall brightness as stronger than indoors, consistent with the actual visual experience, successfully maintaining the scene's contrast.

[0145] Example 2: Personalized calibration process.

[0146] During the postoperative adjustment phase, the doctor used specialized software. The system sequentially activated each electrode, delivering a charge density stimulus that increased stepwise from 0. The patient reported whether they could see the stimulus (threshold measurement) and whether they felt discomfort (upper limit of comfort measurement) by pressing buttons. The measured CD_th_A of electrode A was 24 μC / cm. 2 CD_comfort_A=390 μC / cm 2 ;

[0147] The CD_th_B of electrode B is 45 μC / cm. 2 CD_comfort_B=300 μC / cm 2 And so on. The software automatically updates these values ​​as personalized parameters to the mapping parameters of the M_normal mode, and sets a more conservative CD_255 for electrode B. In subsequent use, patients reported a significant improvement in visual uniformity.

[0148] To further improve visual perception quality, the system of this invention also supports simulating the ON / OFF pathway function of the retina. The following is a brief explanation of the principles:

[0149] Retinal photoreceptors (rods / cones) are specialized cells that undergo photohyperpolarization: hyperpolarizing in the presence of light and depolarizing in the absence of light (the opposite of the "stimulus-induced depolarization" of ordinary neurons). They are the starting point of the entire pathway and the root of ON / OFF cell polarity differentiation. ON-type cells respond most strongly to increased light (bright stimulation), while OFF-type cells respond most strongly to decreased light (dark stimulation). Behind this is a sophisticated central-peripheral antagonistic neural computation mechanism. Specifically, these two types of cells refer to ganglion cells in the retina (responsible for transmitting processed visual signals to the brain). Their "receptive field" (i.e., the area of ​​the retina that influences the cell's activity) is not homogeneous but consists of a central region and a peripheral region. The responses of these two regions to light are antagonistic.

[0150] ON-type cells, short for ON-central / OFF-peripheral cells, work as follows:

[0151] When light shines on the central area of ​​its receptive field, it becomes excited (the firing frequency increases).

[0152] When light shines on the peripheral area of ​​its receptive field, it inhibits (reduces the discharge frequency).

[0153] The strongest reaction occurs when the center is bright and the periphery is dark (i.e., a bright spot falls on a dark background), which produces the dual effect of central excitation and peripheral inhibition relief. Conversely, if the center is dark and the periphery is bright, it will be subject to the strongest inhibition.

[0154] OFF-type cells, also known as OFF-center / ON-peripheral cells, work as follows:

[0155] When light shines on the central region of its receptive field, it inhibits (reduces the firing frequency).

[0156] When light shines on the periphery of its receptive field, it becomes excited (the firing frequency increases).

[0157] The strongest reaction occurs when the center is dark and the periphery is bright (i.e., a dark spot falls on a bright background), which produces the dual effect of central inhibition relief and peripheral excitation. Conversely, if the center is bright and the periphery is dark, it will be subject to the strongest inhibition.

[0158] This ON / OFF separation of channels has significant biological advantages:

[0159] Efficient coding and energy saving: Instead of simply reporting the absolute brightness of each point (which would be very redundant and energy-consuming), the nervous system specifically reports changes in brightness and contrast, thus greatly compressing the amount of information.

[0160] Enhanced edge detection: The boundaries of an object are precisely where the contrast between light and dark is strongest. ON-type cells react strongly to bright edges (bright sides), while OFF-type cells react strongly to dark edges (dark sides). Working together, they greatly enhance our perception of object contours and textures.

[0161] Adapting to different lighting environments: Whether in bright daylight or dim night, the nervous system always focuses on relative changes rather than absolute brightness, thus maintaining stable shape perception under varying lighting conditions.

[0162] Improved temporal resolution: When the lighting changes (such as when an object moves across the field of view), the ON and OFF channels are activated alternately to provide a more accurate signal of temporal change, which is crucial for motion detection.

[0163] Current mainstream retinal prosthesis systems typically employ a simple linear mapping strategy of "brightness-current amplitude." That is, the brighter the environment, the greater the amplitude or frequency of the electrical pulses emitted by the corresponding electrodes. This strategy only provides coarse point localization perception and cannot simulate the crucial parallel neural information processing mechanism of the retina—the ON / OFF antagonistic pathway. In a healthy retina, the ON pathway detects "brightness increase" (light exposure), while the OFF pathway detects "darkness increase" (light removal). They work together to efficiently encode contrast and motion edges, which is the foundation for forming clear contours, textures, and motion perception.

[0164] The drawback of existing technology is that its single-encoded stimulation can simultaneously and non-specifically activate ON-type and OFF-type ganglion cells in the electrode area, generating chaotic neural signals. This results in blurred images perceived by patients, lack of edges and details, and difficulty in understanding dynamically changing scenes.

[0165] Based on the above background, the external perception processing unit of the intelligent visual prosthesis system of the present invention is further configured with a second data processing unit, which is used to determine whether an ON event or an OFF event is triggered based on the gray value change between the target image and the previous frame image, and synthesize a mixed stimulation instruction containing ON type stimulation and / or OFF type stimulation based on the stimulation instruction generated by the first data processing unit according to the judgment result and output it to the stimulation signal generation unit.

[0166] The ON-type stimulus is executed based on the stimulus instruction, while the OFF-type stimulus includes at least one background pulse below the perception threshold and at least one no-stimulation interval. During the background pulse, neither OFF-type nor ON-type cells are activated. When the background pulse pauses, the cells enter the no-stimulation interval, at which point the OFF-type cells are activated and generate nerve impulses.

[0167] It should be noted that the names "second data processing unit" and "first stimulation unit" are used to distinguish their functions for ease of explanation. In practice, the first stimulation unit and the second data processing unit can be integrated into a single main processor 103 based on different software functional modules, or they can be implemented separately on different processors; this is not limited here.

[0168] In some implementations, determining whether to trigger an ON or OFF event based on the change in grayscale value between the target image and the previous frame includes:

[0169] Calculate the degree or rate of change of the grayscale value of each pixel in the target image compared to the corresponding pixel in the previous frame image;

[0170] If the grayscale value increases and the degree or rate of change exceeds the first preset threshold, it is determined that the ON event has been triggered.

[0171] If the grayscale value decreases and the degree or rate of change exceeds the second preset threshold, then the OFF event is determined to be triggered.

[0172] In some implementations, synthesizing a mixed stimulus instruction comprising ON-type stimuli and / or OFF-type stimuli includes:

[0173] If the ON event is triggered, the current stimulation cycle will execute the ON type stimulus at time T1 and the OFF type stimulus at time T2, and T1 > T2;

[0174] If the ON / OFF event is not triggered, the ON type stimulus will be executed in the current stimulus cycle;

[0175] If an OFF event is triggered, the current stimulation cycle will execute an ON-type stimulus at time T1 and an OFF-type stimulus at time T2, where T1 < T2.

[0176] Through the above scheme, the visual prosthesis system of the present invention separates the dynamic information of the acquired visual image into two independent feature streams, "brightening" (ON) and "darkening" (OFF), and maps them into two types of stimulation coupled in time and space. It encodes the directional information of brightness change through the temporal structure difference of electrical stimulation, rather than just the amplitude difference, thereby achieving a better stimulation effect.

[0177] In one example shown, see Figure 5 The image represents the grayscale image corresponding to the pixelated target image. When viewing a human figure, the pixels in region A have uniform overall brightness. The difference in grayscale values ​​between corresponding pixels in two adjacent frames over time is small, preventing the triggering of ON or OFF events. Therefore, only ON-type stimuli are executed in the current stimulation cycle. The specific stimulus amplitude is determined by the grayscale value of the corresponding pixel and is limited by the M_dark, M_normal, and M_bright modes. Similarly, the pixels in region D have relatively dark and uniform overall brightness. The difference in grayscale values ​​between two adjacent frames over time is also small, preventing the triggering of ON or OFF events. Only ON-type stimuli are executed in the current stimulation cycle. Studies have shown that ON-type cells are the majority in the macula, which can reduce computational load, improve operational efficiency, and lower power consumption.

[0178] In regions B and C, which contain image edge information, the difference in grayscale values ​​between adjacent pixels is large. Furthermore, when the edge position changes, the difference in grayscale values ​​between corresponding pixels in two adjacent frames is significant over time. For example, if a person moves to the right in the field of vision, the grayscale value of pixel C (e.g., 100) will approach the grayscale value of pixel B in the previous stimulus cycle (e.g., 50), resulting in a significant darkening and triggering an OFF event. Therefore, within this stimulus cycle, the stimulus strategy for point C executes an ON-type stimulus at time T1 and an OFF-type stimulus at time T2, with T1 < T2 (T1 + T2 constitutes one stimulus cycle). Conversely, if a person moves to the left in the field of vision, the grayscale value of pixel B (e.g., 50) will approach the grayscale value of pixel C in the previous stimulus cycle (e.g., 100), resulting in a significant brightening and triggering an ON event. Therefore, within this stimulus cycle, the stimulus strategy for point B executes an ON-type stimulus at time T1 and an OFF-type stimulus at time T2, with T1 > T2. Therefore, during retinal stimulation, the electrodes corresponding to the edge regions of an object can simultaneously activate both ON-type and OFF-type ganglion cells, generating stronger nerve impulses. These impulses are then interpreted by the visual cortex of the brain, resulting in higher contrast and clarity, making it easier to distinguish objects.

[0179] In one example shown, see Figure 6 The diagram illustrates a stimulation cycle following the triggering of an OFF event. ON-type stimuli execute the stimulation command generated by the first data processing unit (the portion corresponding to the horizontal axis T_on); OFF-type stimuli (the portion corresponding to the horizontal axis T_off) include at least one background pulse (green line segment, including two segments corresponding to positive and negative signals respectively located above and below the horizontal axis for charge balance) below the perception threshold (shown by the dashed line above the horizontal axis) and a no-stimulation interval (the portion coinciding with the horizontal axis). The stimulation cycle following the triggering of an ON event is similar, except that the duration of ON-type stimuli is longer than the duration of OFF-type stimuli.

[0180] As a further preferred embodiment, synthesizing a mixed stimulus instruction comprising ON-type stimuli and / or OFF-type stimuli also includes:

[0181] If the ON event is triggered, the current stimulation cycle will execute the ON type stimulus at time T1 and the OFF type stimulus at time T2, and T1 > T2;

[0182] If the OFF event is triggered, the current stimulation cycle will execute the ON type stimulus at time T1 and the OFF type stimulus at time T2, and T1 < T2;

[0183] Where T1 + T2 equals one stimulation period, the values ​​or ratios of T1 and T2 are calculated based on the degree or rate of change of the grayscale value of the corresponding pixel when the ON or OFF event is triggered (spatiotemporal derivative), and the greater the degree or rate of change, the larger the corresponding value or ratio of T1 or T2. The specific calculation method can be calculated using a preset linear or nonlinear function, or by taking or matching values ​​according to a preset range, without specific limitations here.

[0184] As a further preferred embodiment, after implantation, the visual prosthesis system of this embodiment can obtain subjective feedback information from the target subject (patient) viewing specific dynamic patterns (such as flashing gratings, moving edges). The physician can then adjust the ON / OFF event trigger threshold, maximum stimulation intensity, and / or time constant (stimulation period and T1, T2 magnitudes) of each electrode or electrode area through a configuration system (software interface), enabling the target subject (patient) to achieve a preset resolution target (such as clearly distinguishing between "bright bars appearing" and "dark bars appearing"). The above process can also be achieved by detecting visual evoked potentials and action potentials.

[0185] Another embodiment of the present invention also illustrates an intelligent visual prosthesis encoding method, comprising the following steps:

[0186] Acquire target images;

[0187] Based on the image features of the target image or the ambient illuminance obtained by the environment perception unit, the current scene is matched to a preset lighting scene mode. Anchor points corresponding to the gray value boundaries of the image are independently configured for the matched lighting scene mode. Based on the configured anchor points, a gray-charge density mapping relationship corresponding to the lighting scene mode is generated. Then, the gray value of each pixel of the target image is mapped to the corresponding target charge density, and a stimulus command is generated.

[0188] An electrical stimulation pulse sequence is generated based on the stimulation command and a preset reference pulse width and output to the electrode array.

[0189] The specific implementation methods of each step of the above method can be found in the above-described intelligent vision prosthetic system embodiments, and will not be repeated here.

[0190] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples; within the framework of the invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of the invention as described above, which are not provided in the details for the sake of brevity.

[0191] The embodiments of this invention are intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of this invention should be included within the protection scope of this invention.

Claims

1. An intelligent visual prosthetic system, characterized in that, The device includes an external sensing and processing unit and an internal implanted stimulation unit with communication connection. The external sensing and processing unit includes at least an environmental sensing unit and a first data processing unit, and the internal implanted stimulation unit includes at least a stimulation signal generation unit and an electrode array. The environmental sensing unit is used to acquire target images; The first data processing unit is used to match the current scene to a preset lighting scene mode based on the image features of the target image or the ambient illuminance obtained by the environment perception unit, independently configure anchor points for the corresponding image gray value boundaries for the matched lighting scene mode, and generate a gray-charge density mapping relationship corresponding to the lighting scene mode based on the configured anchor points, thereby mapping the gray value of each pixel of the target image to the corresponding target charge density, generating a stimulus instruction and outputting it to the stimulus signal generation unit. The stimulation signal generation unit is used to generate an electrical stimulation pulse sequence according to the stimulation command and a preset reference pulse width and output it to the electrode array; Among them, the anchor points for independently configuring the corresponding image grayscale value boundaries for the matched lighting scene mode include: Configure a black point anchor point, which determines the minimum target charge density value corresponding to the image gray value 0 in the current lighting scene mode, and from the dark lighting scene mode to the bright lighting scene mode, the minimum target charge density value corresponding to the anchor point transitions from the minimum effective value higher than the physiological threshold to 0. Configure a white anchor point, which determines the maximum target charge density value corresponding to the current lighting scene mode of the image grayscale value 255. This maximum target charge density value increases as the ambient illuminance increases, and the maximum value is less than the absolute safety upper limit.

2. The intelligent visual prosthesis system according to claim 1, characterized in that, The data processing unit extracts image features based on the target image and matches lighting scene modes based on a preset value range according to the extracted image features.

3. The intelligent visual prosthesis system according to claim 1, characterized in that, The implanted stimulation unit also includes an impedance monitoring and compensation unit. The stimulation signal generation unit is used to calculate the required ideal stimulation current amplitude based on the stimulation command and the preset reference pulse width. The impedance monitoring and compensation unit is used to measure the contact impedance of the current electrode and calculate the compensated current amplitude based on a preset impedance-efficiency compensation function. The stimulation signal generation unit is used to generate an electrical stimulation pulse sequence based on the compensated current amplitude and output it to the electrode array.

4. The intelligent visual prosthesis system according to claim 1, characterized in that, The gray-scale-charge density mapping relationship generated based on the configured anchor points for this lighting scene mode includes: Using the black and white anchor points as boundary anchor points, a nonlinear function is used to generate the basic gray-scale-charge density mapping curve under the corresponding lighting scene mode. Real-time brightness histogram analysis is performed on the current target image. If the overall brightness of the image is lower or higher than a preset threshold, the basic gray-level-charge density mapping curve is globally or locally fine-tuned to generate a gray-level-charge density mapping lookup table.

5. The intelligent visual prosthesis system according to claim 3, characterized in that, Generating an electrical stimulation pulse sequence based on the compensated current amplitude includes: Calculate the charge density of the planned electrical stimulation pulse sequence based on the compensated current amplitude: The calculated charge density of the planned electrical stimulation pulse sequence is compared with the preset personalized safety threshold of the current electrode: If the value is less than the personalized safety threshold, it is considered safe, and a discharge stimulation pulse sequence is generated according to the calculated charge density. If the charge density exceeds the personalized safety threshold, dynamic safety control is triggered to ensure that the charge density of the actual electrical stimulation pulse sequence is less than the personalized safety threshold of the current electrode.

6. The intelligent visual prosthesis system according to claim 5, characterized in that, The dynamic security control includes: Based on the constant charge density optimization algorithm, the pulse width of the electrical stimulation pulse sequence is adjusted preferentially within the allowable range; If the adjusted charge density still cannot meet the personalized safety threshold of the current electrode, the magnitude of the stimulation current signal is limited proportionally so that the charge density of the actual electrical stimulation pulse sequence is less than the personalized safety threshold of the current electrode.

7. The intelligent visual prosthesis system according to claim 1, characterized in that, The process of independently configuring anchor points for the corresponding image grayscale value boundaries for the matched lighting scene mode, and generating a grayscale-charge density mapping relationship corresponding to the lighting scene mode based on the configured anchor points, also includes: The charge density perception threshold and comfort limit of a single electrode are obtained based on psychophysical parameter measurements. The charge density perception threshold and comfort upper limit are used as calibration references to calibrate the black anchor point and white anchor point; By testing multiple stimulation levels, the charge density-subjective brightness function of the current electrode is fitted, and the basic gray-scale-charge density mapping curve is locally corrected based on this function, and a gray-scale-charge density mapping lookup table is generated.

8. The intelligent visual prosthesis system according to claim 7, characterized in that, Using the charge density perception threshold and comfort upper limit as calibration references, the calibration of the black dot anchor and white dot anchor includes: Each electrode was initially stimulated starting from zero charge density, and the working potential or visual evoked potential generated by the electrode during the initial stimulation was detected. The minimum charge density threshold that can generate the correct action potential or visual evoked potential is mapped to grayscale 0. Continue to increase the charge density until the potential amplitude no longer increases, then map the maximum charge density value at that moment to a grayscale value of 255. The minimum charge density threshold is used as the calibration reference for the black spot anchor point of the electrode, and the maximum charge density value is used as the calibration reference for the white spot anchor point.

9. The intelligent visual prosthesis system according to claim 1, characterized in that, The external sensing processing unit further includes a second data processing unit, which is used to determine whether an ON event or an OFF event is triggered based on the gray value change between the target image and the previous frame image, and to synthesize a mixed stimulation instruction containing ON type stimulation and / or OFF type stimulation based on the stimulation instruction according to the determination result and output it to the stimulation signal generation unit. The ON type stimulus is executed based on the stimulus instruction, and the OFF type stimulus includes at least one background pulse below the perception threshold and at least one no-stimulation interval.

Images