Fundus recognition hyperspectral imaging chip design method, fundus imaging device and equipment

Abstract

The invention discloses a fundus recognition hyperspectral imaging chip design method, a fundus imaging device and fundus imaging equipment, and relates to the field of wisdom medicines.The fundus recognition hyperspectral imaging chip design method comprises the steps that a hyperspectral image sample of clinical hyperspectral leopard-print-shaped fundus in a set waveband range is collected, and waveband splitting is conducted to obtain a single-channel spectral image; performing image scoring on the single-channel spectral images to form a training sample so as to train a neural network model, and performing random combination on the multiple single-channel spectral images corresponding to each patient to obtain multiple single-channel spectral image combinations; scoring each single-channel spectral image combination based on the trained neural network model, and screening to obtain a set number of single-channel spectral image combinations with top scores; and constructing a corresponding hyperspectral imaging chip for each single-channel spectral image combination, and performing clinical evaluation to obtain an optimal hyperspectral imaging chip. The method can effectively improve the efficiency of clinically identifying the leopard print fundus.

Description

Technical Field

[0001] The present application relates to the field of smart medical care, and specifically to a fundus recognition hyperspectral imaging chip design method, fundus imaging device and equipment. Background Art

[0002] With the continuous development of ophthalmology and artificial intelligence technology, fundus image analysis is playing an increasingly important role in early disease screening, diagnosis, and treatment monitoring. Fundus images can reflect the health of structures such as blood vessels, nerves, and the retina. Accurate image analysis is particularly important in the diagnosis of ophthalmic diseases such as diabetic retinopathy, glaucoma, and macular degeneration. In recent years, with the continuous advancement of deep learning and computer vision technologies, automated diagnostic technologies based on fundus images have gradually been applied.

[0003] Leopard-like fundus is a unique fundus lesion, often seen in certain genetic eye diseases or systemic disorders. It is characterized by spotty, irregularly distributed pigmentation or atrophy on the retina. Due to its unique characteristics, leopard-like fundus presents challenges in early diagnosis and clinical monitoring, especially in differential diagnosis from other lesions. Traditional fundus image analysis methods often struggle to achieve high-precision and efficient automatic recognition of such complex images, necessitating further technological development to accurately identify leopard-like fundus.

[0004] In recent years, hyperspectral imaging technology, due to its ability to capture image information across diverse spectral bands, has been widely used in the field of medical imaging, particularly in ophthalmic diagnosis, where it has shown great potential. For the precise identification of fundus lesions with leopard-marked patterns, the application of hyperspectral imaging chips promises to transcend the limitations of traditional band-based image analysis and provide more information. Hyperspectral imaging chips can capture fundus images in multiple bands, including visible light and near-infrared. This multi-angle imaging approach can effectively reveal the different layers and subtle structures of fundus lesions, particularly the complex features of fundus lesions, such as pigmentation and vascular changes. This means that the design and construction of hyperspectral chips can enable refined recognition and real-time analysis of fundus images, potentially enabling accurate identification of fundus lesions with leopard-marked patterns. Therefore, the question of how to construct a hyperspectral imaging chip for accurate identification of fundus lesions with leopard-marked patterns has become a pressing issue. Summary of the Invention

[0005] The present application provides a fundus identification hyperspectral imaging chip design method, fundus imaging device and equipment, which can effectively improve the efficiency of clinical leopard-mark fundus identification.

[0006] In a first aspect, an embodiment of the present application provides a method for designing a hyperspectral imaging chip for fundus recognition, and the method for designing a hyperspectral imaging chip for fundus recognition includes: Collect hyperspectral image samples of clinical hyperspectral tigroid fundus under a set band range, and perform band splitting on the hyperspectral image samples to obtain single-channel spectral images; Perform image scoring on the single-channel spectral images to form training samples for training a neural network model, randomly combine multiple single-channel spectral images corresponding to each patient to obtain multiple single-channel spectral image combinations; Based on the trained neural network model, score each single-channel spectral image combination, and screen and obtain a set number of single-channel spectral image combinations with higher scores according to the scoring results; Construct a corresponding hyperspectral imaging chip for each of the selected single-channel spectral image combinations, and integrate it into a fundus imaging device for clinical evaluation to obtain the best hyperspectral imaging chip.

[0007] Combined with the first aspect, in an implementation manner, the collecting hyperspectral image samples of clinical hyperspectral tigroid fundus under a set band range, and performing band splitting on the hyperspectral image samples to obtain single-channel spectral images specifically includes: Collect hyperspectral image samples of clinical hyperspectral tigroid fundus, and the hyperspectral image samples are imaging data of each specific band under a set band range; Perform band splitting on the collected hyperspectral image samples to obtain single-channel spectral images of specific bands.

[0008] Combined with the first aspect, in an implementation manner, the performing image scoring on the single-channel spectral images to form training samples for training a neural network model specifically includes: Based on Sobel gradient, Laplacian variance, and subjective scoring methods, score the current single-channel spectral image respectively, and perform weight settlement on the scores of each scoring method to obtain the score of the current single-channel spectral image; Based on each single-channel spectral image and the score of each single-channel spectral image, construct training samples for training a neural network model.

[0009] Combined with the first aspect, in an implementation manner, the scoring each single-channel spectral image combination based on the trained neural network model, and screening and obtaining a set number of single-channel spectral image combinations with higher scores according to the scoring results specifically includes: Based on the trained neural network model, score each single-channel spectral image in the single-channel spectral image combination, and use the sum of the scores of each single-channel spectral image as the score of the current single-channel spectral image combination; Sort the combinations of single-channel spectral images according to the scores from high to low, and screen out a set number of combinations of single-channel spectral images with the top scores.

[0010] Combined with the first aspect, in one implementation, for each combination of single-channel spectral images obtained by screening, a corresponding hyperspectral imaging chip is constructed and integrated into a fundus imaging device for clinical evaluation to obtain the best hyperspectral imaging chip. Specifically, it includes: Construct a corresponding hyperspectral imaging chip for each combination of single-channel spectral images obtained by screening; Integrate the hyperspectral imaging chips into the fundus imaging device respectively and conduct clinical evaluations for identifying tigroid fundus in sequence. Determine the best hyperspectral imaging chip according to the clinical evaluation results.

[0011] In the second aspect, an embodiment of the present application provides a fundus imaging device for identifying tigroid fundus, including a fundus camera and a hyperspectral imaging chip camera that work together. Among them, the hyperspectral imaging chip camera is constructed based on the best hyperspectral imaging chip obtained by the above design method.

[0012] Combined with the second aspect, in one implementation, the specific process of the fundus imaging device for identifying tigroid fundus includes: Establish a physical connection between the fundus camera and the hyperspectral imaging chip camera, and keep the field of view and focus of the fundus camera and the hyperspectral imaging chip camera consistent; Control the fundus camera and the hyperspectral imaging chip camera to collect fundus images, and fuse the fundus images collected by the fundus camera and the hyperspectral imaging chip camera to generate a comprehensive image containing structural features and spectral information, thereby realizing the identification of tigroid fundus.

[0013] Combined with the second aspect, in one implementation, the establishment of the physical connection between the fundus camera and the hyperspectral imaging chip camera, and keeping the field of view and focus of the fundus camera and the hyperspectral imaging chip camera consistent specifically includes: Establish a physical connection between the fundus camera and the hyperspectral imaging chip camera through an adapter or a standard interface; Adjust the field of view and focus of the fundus camera and the hyperspectral imaging chip camera so that the field of view and focus of the fundus camera and the hyperspectral imaging chip camera are consistent.

[0014] Combined with the second aspect, in one implementation, the control of the fundus camera and the hyperspectral imaging chip camera to collect fundus images, and the fusion of the fundus images collected by the fundus camera and the hyperspectral imaging chip camera to generate a comprehensive image containing structural features and spectral information specifically includes: Based on the cooperation of software and hardware, the synchronization control of the fundus camera and the hyperspectral imaging chip camera is realized, so that the hyperspectral imaging chip camera can collect fundus images of the same area as the fundus camera in different spectral bands; The fundus images collected by the fundus camera and the hyperspectral imaging chip camera are fused through an image fusion algorithm to generate a comprehensive image containing structural features and spectral information.

[0015] In a third aspect, an embodiment of the present application provides a fundus recognition hyperspectral imaging chip design device, which is characterized in that the fundus recognition hyperspectral imaging chip design device includes a processor, a memory, and a fundus recognition hyperspectral imaging chip design program stored on the memory and executable by the processor. When the fundus recognition hyperspectral imaging chip design program is executed by the processor, the steps of the above-mentioned fundus recognition hyperspectral imaging chip design method are realized.

[0016] The beneficial effects brought by the technical solution provided by the embodiment of the present application include: By designing the optimal hyperspectral imaging chip, a hyperspectral imaging chip camera is obtained based on the optimal hyperspectral imaging chip, and then a fundus imaging device is obtained to perform fundus imaging on clinical patients, realizing the accurate recognition of the tigroid fundus image and having a recognition accuracy rate, providing intuitive and convenient diagnostic support for ophthalmologists, thereby improving the clinical recognition efficiency of the tigroid fundus and providing important support for the accurate clinical recognition of tigroid fundus lesions, having significant clinical value and significance. Description of the Drawings

[0017] Figure 1 It is a schematic flowchart of the fundus recognition hyperspectral imaging chip design method of the present application; Figure 2 It is a schematic diagram of the functional modules of the fundus imaging device of the present application; Figure 3 It is a schematic diagram of the hardware structure of the fundus recognition hyperspectral imaging chip design device of the present application. Detailed Embodiments

[0018] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0019] To make the purpose, technical solution and advantages of the present application clearer, the embodiments of the present application will be further described in detail below in conjunction with the drawings.

[0020] In a first aspect, an embodiment of the present application provides a method for designing a hyperspectral imaging chip for fundus recognition.

[0021] In one embodiment, referring to Figure 1 , Figure 1 is a schematic flowchart of the method for designing a hyperspectral imaging chip for fundus recognition in the present application. As Figure 1 shown, the method for designing a hyperspectral imaging chip for fundus recognition includes: S1: Collect hyperspectral image samples of clinical hyperspectral tigroid fundus within a set wavelength band range, and split the hyperspectral image samples into single-channel spectral images. S2: Perform image scoring on the single-channel spectral images to form training samples for training a neural network model, and randomly combine multiple single-channel spectral images corresponding to each patient to obtain multiple combinations of single-channel spectral images. S3: Based on the trained neural network model, score each combination of single-channel spectral images, and screen out a set number of combinations of single-channel spectral images with higher scores according to the scoring results. S4: Construct a corresponding hyperspectral imaging chip for each screened combination of single-channel spectral images, and integrate it into a fundus imaging device for clinical evaluation to obtain the best hyperspectral imaging chip.

[0022] Further, in one embodiment, collecting hyperspectral image samples of clinical hyperspectral tigroid fundus within a set wavelength band range, and splitting the hyperspectral image samples into single-channel spectral images specifically includes: S101: Collect hyperspectral image samples of clinical hyperspectral tigroid fundus, where the hyperspectral image samples are imaging data of each specific wavelength band within a set wavelength band range. S102: Split the collected hyperspectral image samples into single-channel spectral images of specific wavelength bands.

[0023] Specifically, in actual applications, first collect hyperspectral image samples of clinical hyperspectral tigroid fundus, covering imaging data in the wavelength band of 623nm to 794nm. The specific wavelength bands include 623nm, 650nm, 663nm, 685nm, 709nm, 730nm, 751nm, 773nm, and 794nm. Then split the collected hyperspectral image samples into single-channel spectral images of 623nm, 650nm, 663nm, 685nm, 709nm, 730nm, 751nm, 773nm, and 794nm.

[0024] Further, in one embodiment, performing image scoring on the single-channel spectral images to form training samples for training a neural network model specifically includes: S201: Score the current single-channel spectral image based on Sobel gradient, Laplacian variance, and subjective scoring methods respectively, and calculate the weights of the scores of each scoring method to obtain the score of the current single-channel spectral image; S202: Construct training samples based on each single-channel spectral image and the scores of each single-channel spectral image to train the neural network model.

[0025] Specifically, for the multiple single-channel spectral images obtained after splitting, each single-channel spectral image is obtained in sequence. The current single-channel spectral image is scored using Sobel gradient, Laplacian variance, and doctor's subjective scoring methods respectively. Then, the weights of the Sobel gradient score, Laplacian variance score, and doctor's subjective evaluation score are calculated. In a possible real-time method, the weights of the Sobel gradient score and Laplacian variance score are 0.25, and the weight of the doctor's subjective evaluation score is 0.5. Finally, the score of the current single-channel spectral image is obtained. Using the above method, the scores of each single-channel spectral image are calculated, and then training samples are constructed based on the single-channel spectral images and the corresponding scores of each single-channel spectral image to train the neural network model, so that the trained neural network model has the ability to score single-channel spectral images.

[0026] Further, in one embodiment, based on the trained neural network model, score each combination of single-channel spectral images, and screen out a set number of combinations of single-channel spectral images with higher scores according to the scoring results. Specifically, it includes: S301: Score each single-channel spectral image in the combination of single-channel spectral images based on the trained neural network model, and use the sum of the scores of each single-channel spectral image as the score of the current combination of single-channel spectral images; S302: Sort according to the scores of each combination of single-channel spectral images from high to low, and screen out a set number of combinations of single-channel spectral images with higher scores.

[0027] Specifically, for the single-channel spectral images obtained after the previous splitting, multiple single-channel spectral images corresponding to each patient are regarded as a single group. The multiple single-channel spectral images within a single group are randomly recombined. For each single group, multiple combinations of single-channel spectral images with a set number are obtained. For all patients, multiple combinations of single-channel spectral images are finally obtained. Then, the trained neural network model is used to score each single-channel spectral image within a single combination of single-channel spectral images. The sum of the scores of each single-channel spectral image is used as the score of the current combination of single-channel spectral images. The combinations of single-channel spectral images are sorted according to the scores from high to low, and a set number of combinations of single-channel spectral images with the top scores are selected. In a possible real-time manner, the top 20 combinations of single-channel spectral images are selected.

[0028] Further, in one embodiment, for each combination of single-channel spectral images obtained by screening, a corresponding hyperspectral imaging chip is constructed and integrated into a fundus imaging device for clinical evaluation to obtain the best hyperspectral imaging chip, which specifically includes: S401: For each combination of single-channel spectral images obtained by screening, a corresponding hyperspectral imaging chip is constructed; S402: The hyperspectral imaging chips are respectively integrated into the fundus imaging device and are successively subjected to clinical evaluations for identifying tigroid fundus. The best hyperspectral imaging chip is determined according to the clinical evaluation results.

[0029] Specifically, for the combinations of single-channel spectral images obtained by screening, each combination of single-channel spectral images is designed and customized with a hyperspectral imaging chip to obtain multiple hyperspectral imaging chips, which are respectively integrated into the fundus imaging device. Then, the fundus imaging device integrated with the current hyperspectral imaging chip is successively subjected to clinical use evaluations, and finally, the professional doctor determines the hyperspectral imaging chip with the best effect for identifying tigroid fundus to obtain the best hyperspectral imaging chip.

[0030] Further, for the construction of the hyperspectral imaging chip based on the spectral images, it specifically includes: S4011: Obtain the spatial position information corresponding to the spectral images, decompose the input optical signal through a micro-spectral separation device to generate an initial spectral-spatial data set, and use a positioning system to record the spatial coordinates corresponding to each spectrum to obtain a preliminary mapping relationship matrix; S4022: For the preliminary mapping relationship matrix, adopt an adaptive spectral-spatial mapping algorithm to extract feature vectors from the initial spectral-spatial data set, and determine whether the spectral resolution and spatial resolution of the feature vectors meet the preset threshold to obtain an optimized mapping feature set; S4023: According to the optimized mapping feature set, adjust the spectral decomposition parameters through the micro-spectral separation device. If the spectral resolution is lower than the threshold, increase the sampling frequency of the micro-spectral separation device to generate a high-precision spectral-spatial correspondence table; S4024: Obtain the high-precision spectral-spatial correspondence table, calibrate the spatial coordinate deviation in real time in combination with the positioning system, update the mapping relationship through the dynamic change adaptation algorithm. If coordinate drift is detected, recalculate the position offset to determine the calibrated mapping data set; S4025: For the calibrated mapping data set, adopt a real-time mapping quality evaluation mechanism to calculate the spectral-spatial matching error, extract the data distortion degree index from the error distribution, judge whether the distortion degree exceeds the preset threshold to obtain the mapping result set with optimized error; S4026: According to the mapping result set with optimized error, adjust the parameters of the spectral-spatial mapping algorithm through adaptive technology, and update the parameters of the mapping algorithm according to the scene change trend to generate an adaptively adjusted mapping output set; S4027: For the mapping data set, verify the stability of the spatial coordinates through the positioning system. If it is detected that the positioning error accumulates in a dynamic scene, perform optimization processing to generate a stable spectral-spatial mapping result set, thereby obtaining the chip data mapping set and realizing the construction of the hyperspectral imaging chip.

[0031] The following combines an example to specifically illustrate the design method of the fundus recognition hyperspectral imaging chip of the present application.

[0032] First, collect hyperspectral image samples of the clinical hyperspectral leopard-patterned fundus. The collected image sample data covers the spectral band of 623nm - 794nm. These bands include multiple spectral points such as 623nm, 650nm, 663nm, 685nm, 709nm, 730nm, 751nm, 773nm, 794nm, etc. Through the collection of hyperspectral data in this wavelength range, rich information of the fundus image can be obtained from multiple levels and multiple bands, providing comprehensive data support for subsequent image processing and analysis. Then, split the collected hyperspectral image samples into nine single-channel spectral images, corresponding to the bands of 623nm, 650nm, 663nm, 685nm, 709nm, 730nm, 751nm, 773nm, 794nm respectively. Each single-channel spectral image will be evaluated through multiple indicators to ensure the quality and information effectiveness. The evaluation indicators of the image include the Sobel gradient score, Laplacian variance score, and doctor's subjective evaluation score. These evaluation indicators will comprehensively analyze each single-channel spectral image to evaluate its practicality and effect in fundus disease diagnosis.

[0033] Then, develop a neural network model through AI technology, and combine the image evaluation results to train and optimize the leopard pattern fundus recognition of the neural network model. Import the constructed training samples into the neural network model for learning and optimization. Take multiple single-channel spectral images corresponding to each patient as a single group, and randomly recombine the multiple single-channel spectral images within a single group. Considering the complementarity of different spectral band information, in the obtained single-channel spectral image combinations, each combination must contain at least three different band information, and each combination of spectral bands is allowed to appear at most three times.

[0034] During the model training process, the evaluation index is crucial for the training effect. The score of each combination will comprehensively consider the Sobel gradient score, the Laplacian variance score, and the physician evaluation score. Among them, the weights of the Sobel gradient score and the Laplacian variance score each account for 0.25, while the weight of the physician evaluation score is 0.5. The neural network model scores each single-channel spectral image within a single single-channel spectral image combination, and takes the sum of the scores of each single-channel spectral image as the score of the current single-channel spectral image combination. Sort the single-channel spectral image combinations according to the scores from high to low, and select the top 20 single-channel spectral image combinations to obtain the most effective spectral combination, thereby improving the recognition accuracy of the leopard pattern fundus.

[0035] Furthermore, in order to improve the recognition accuracy, select the top 3 single-channel spectral image combinations from the 20 selected single-channel spectral image combinations, and develop a corresponding hyperspectral imaging chip based on the neural network model and the scoring results of the neural network model. Compared with traditional color photos and other hyperspectral chips, it has stronger leopard pattern fundus recognition ability.

[0036] In a second aspect, the embodiments of the present application also provide a fundus imaging device.

[0037] In one embodiment, refer to Figure 2 , Figure 2 is a schematic diagram of the functional modules of the fundus imaging device of the present application. As Figure 2 shown, the fundus imaging device includes a fundus camera and a hyperspectral imaging chip camera that work together. Among them, the hyperspectral imaging chip camera is constructed based on the best hyperspectral imaging chip obtained by the above design method. By combining the fundus camera and the hyperspectral imaging chip camera, fundus imaging of clinical patients can be achieved. In a possible implementation manner, the fundus camera is a TopCon NW8 fundus camera.

[0038] In the present application, the specific process of the fundus imaging device for leopard pattern fundus recognition includes: a: Establish a physical connection between the fundus camera and the hyperspectral imaging chip camera, and keep the field of view range and focus of the fundus camera and the hyperspectral imaging chip camera consistent; Further, establish a physical connection between the fundus camera and the hyperspectral imaging chip camera, and keep the field of view and focus of the fundus camera and the hyperspectral imaging chip camera consistent. Specifically, it includes: a1: Establish a physical connection between the fundus camera and the hyperspectral imaging chip camera through an adapter or a standard interface; a2: Adjust the field of view and focus of the fundus camera and the hyperspectral imaging chip camera so that the field of view and focus of the fundus camera and the hyperspectral imaging chip camera are consistent.

[0039] b: Control the fundus camera and the hyperspectral imaging chip camera to collect fundus images, and fuse the fundus images collected by the fundus camera and the hyperspectral imaging chip camera to generate a comprehensive image containing structural features and spectral information, so as to realize the identification of the tessellated fundus.

[0040] Further, control the fundus camera and the hyperspectral imaging chip camera to collect fundus images, and fuse the fundus images collected by the fundus camera and the hyperspectral imaging chip camera to generate a comprehensive image containing structural features and spectral information. Specifically, it includes: b1: Realize the synchronous control of the fundus camera and the hyperspectral imaging chip camera based on the cooperation of software and hardware, so that the hyperspectral imaging chip camera collects fundus images of the same area as the fundus camera under different spectral bands; b2: Fuse the fundus images collected by the fundus camera and the hyperspectral imaging chip camera through an image fusion algorithm to generate a comprehensive image containing structural features and spectral information, so as to analyze fundus lesions more accurately.

[0041] Further, in one embodiment, fuse the fundus images collected by the fundus camera and the hyperspectral imaging chip camera to generate a comprehensive image containing structural features and spectral information. Specifically, it includes: b201: Obtain the fundus image collected by the fundus camera, denoted as the first image, and obtain the fundus image collected by the hyperspectral imaging chip camera, denoted as the second image. Extract features from the first image through a convolutional neural network to obtain the preliminary structural feature map and spectral information map corresponding to the first image. Extract features from the second image through a convolutional neural network to obtain the preliminary structural feature map and spectral information map corresponding to the second image; b202: For the obtained preliminary structural feature map, calculate the edge intensity distribution based on the edge detection algorithm to generate an edge-enhanced feature map, and for the generated edge-enhanced feature map, calculate the significance score of each region through a significance evaluation model to generate a significance distribution map; b203: Calculate the ratio between the saliency score and the preset threshold based on the spectral information map and the saliency distribution map corresponding to the first image. If the ratio is greater than 1, adjust the structural feature weight coefficient according to the proportion of the ratio to generate an initial fusion weight matrix. Calculate the ratio between the saliency score and the preset threshold based on the spectral information map and the saliency distribution map corresponding to the second image. If the ratio is greater than 1, adjust the structural feature weight coefficient according to the proportion of the ratio to generate an initial fusion weight matrix; b204: Perform weighted fusion on the preliminary structural feature map and the spectral information map corresponding to the first image through the initial fusion weight matrix corresponding to the first image to generate preliminary comprehensive image data. Perform weighted fusion on the preliminary structural feature map and the spectral information map corresponding to the second image through the initial fusion weight matrix corresponding to the second image to generate preliminary comprehensive image data; b205: Based on the image fusion algorithm, fuse the preliminary comprehensive image data corresponding to the first image and the preliminary comprehensive image data corresponding to the second image to obtain a comprehensive image, and apply a generative adversarial network to optimize the texture region of the comprehensive image to generate a texture-enhanced comprehensive image; b206: Extract the fusion quality index from the texture-enhanced comprehensive image. If the fusion quality index is lower than the preset standard, adjust the structural feature weight coefficient in the initial fusion weight matrix according to the deviation ratio between the fusion quality index and the preset standard, and perform weighted fusion again to generate a comprehensive image again; b207: Detect the edge details and spectral information integrity of the comprehensively generated image again through a multi-scale analysis method to generate an integrity evaluation report. According to the integrity evaluation report, if edge details or spectral information deviation is detected, perform weighted correction on the pixel values of the deviation region through a local adaptive adjustment algorithm, and finally generate an optimized comprehensive image.

[0042] Furthermore, an integrated operation interface can be constructed to combine the comprehensive image generated by the fundus imaging device to realize image processing, analysis, and display functions, providing intuitive and convenient diagnostic support for ophthalmologists, thereby improving the clinical recognition efficiency of tigroid fundus.

[0043] The following describes the actual application effect of the fundus imaging device of the present application in combination with an example.

[0044] Clinical tests were conducted, and 6 clinical patients were selected for a double-blind test of identifying tigroid fundus. Four senior medical technicians operated the fundus imaging device of the present application and the traditional fundus color photography system, and four intermediate medical technicians also operated the fundus imaging device of the present application and the traditional fundus color photography system. For the results, in the senior medical technician inspection group, the correct identification rates of the traditional fundus color photography system and the fundus imaging device of the present application for tigroid fundus were 95.8% and 100% respectively. In the intermediate medical technician inspection group, the correct identification rates of the traditional fundus color photography system and the fundus imaging device of the present application for tigroid fundus were 83.3% and 95.8% respectively. Thus, it can be shown that the fundus imaging device of the present application greatly improves the accuracy of ordinary experts in identifying tigroid fundus.

[0045] In a third aspect, an embodiment of the present application provides a device for designing a high-spectral imaging chip for fundus identification. The device for designing a high-spectral imaging chip for fundus identification may be a device with data processing functions such as a personal computer (PC), a laptop computer, a server, etc.

[0046] Refer to Figure 3 , Figure 3 which is a schematic diagram of the hardware structure of the device for designing a high-spectral imaging chip for fundus identification involved in the solution of the embodiment of the present application. In the embodiment of the present application, the device for designing a high-spectral imaging chip for fundus identification may include a processor, a memory, a communication interface, and a communication bus.

[0047] Among them, the communication bus can be of any type and is used to interconnect the processor, the memory, and the communication interface.

[0048] The communication interface includes an input / output (I / O) interface, a physical interface, and a logical interface, etc., which are interfaces for interconnecting the components inside the device for designing a high-spectral imaging chip for fundus identification, and interfaces for interconnecting the device for designing a high-spectral imaging chip for fundus identification with other devices (such as other computing devices or user devices). The physical interface can be an Ethernet interface, an optical fiber interface, an ATM interface, etc.; the user device can be a display screen (Display), a keyboard (Keyboard), etc.

[0049] The memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical memory, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0050] The processor can be a general-purpose processor, which can call the fundus recognition hyperspectral imaging chip design program stored in the memory and execute the fundus recognition hyperspectral imaging chip design method provided by the embodiments of the present application. For example, the general-purpose processor can be a central processing unit (CPU). Among them, for the method executed when the fundus recognition hyperspectral imaging chip design program is called, reference can be made to the various embodiments of the fundus recognition hyperspectral imaging chip design method of the present application, which will not be elaborated here.

[0051] Those skilled in the art can understand that Figure 3 the hardware structure shown in

[0052] does not constitute a limitation to the present application, and may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements. The terms "including" and "having" and any variations thereof in the description of the embodiments of the present application, as well as in the claims and the above-mentioned drawings, are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices. The descriptions with terms such as "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequential order, nor do they limit that "first", "second" and "third" are of different types.

[0053] In the description of the embodiments of the present application, terms such as "exemplary", "for example" or "for instance" are used to indicate examples, illustrations or explanations. Any embodiment or design solution described as "exemplary", "for example" or "for instance" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of terms such as "exemplary", "for example" or "for instance" is intended to present relevant concepts in a specific manner.

[0054] In the description of the embodiments of the present application, unless otherwise specified, " / " means "or". For example, A / B may mean A or B. The "and / or" in the text is merely a description of the relationship between associated objects, indicating that there can be three relationships. For example, A and / or B may mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of the present application, "a plurality of" means two or more than two.

[0055] In some of the processes described in the embodiments of the present application, there are a plurality of operations or steps that appear in a specific order. However, it should be understood that these operations or steps may not be executed in the order in which they appear in the embodiments of the present application or may be executed in parallel. The serial numbers of the operations are only used to distinguish different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations or steps may be executed in sequence or in parallel, and these operations or steps may be combined.

[0056] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product is stored in a storage medium as described above (such as ROM / RAM, magnetic disk, optical disc), and includes several instructions for causing a terminal device to execute the methods described in the various embodiments of the present application.

[0057] The above are only the preferred embodiments of the present application, and do not limit the patent scope of the present application accordingly. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be similarly included in the patent protection scope of the present application.

Claims

1. A design method for a fundus recognition hyperspectral imaging chip, characterized in that, The method for designing a fundus recognition hyperspectral imaging chip includes: Collecting hyperspectral image samples of clinical hyperspectral tigroid fundus within a set wavelength range, and splitting the hyperspectral image samples into single-channel spectral images; Performing image scoring on the single-channel spectral images to form training samples for training a neural network model, randomly combining multiple single-channel spectral images corresponding to each patient to obtain multiple combinations of single-channel spectral images; Based on the trained neural network model, scoring each combination of single-channel spectral images, and screening a set number of combinations of single-channel spectral images with higher scores according to the scoring results; Constructing a corresponding hyperspectral imaging chip for each screened combination of single-channel spectral images, integrating it into a fundus imaging device for clinical evaluation, and obtaining the optimal hyperspectral imaging chip.

2. The method for designing a fundus recognition hyperspectral imaging chip according to claim 1, wherein, The step of collecting hyperspectral image samples of clinical hyperspectral tigroid fundus within a set wavelength range and splitting the hyperspectral image samples into single-channel spectral images specifically includes: Collecting hyperspectral image samples of clinical hyperspectral tigroid fundus, where the hyperspectral image samples are imaging data of specific wavelengths within a set wavelength range; Splitting the collected hyperspectral image samples into single-channel spectral images of specific wavelengths.

3. The method for designing a fundus recognition hyperspectral imaging chip according to claim 1, characterized in that, The step of performing image scoring on the single-channel spectral images to form training samples for training a neural network model specifically includes: Based on Sobel gradient, Laplacian variance, and subjective scoring methods, respectively scoring the current single-channel spectral image, and calculating the weights of the scores of each scoring method to obtain the score of the current single-channel spectral image; Based on each single-channel spectral image and its score, constructing training samples for training a neural network model.

4. A method for designing a fundus recognition hyperspectral imaging chip according to claim 1, characterized in that, The step of scoring each combination of single-channel spectral images based on the trained neural network model and screening a set number of combinations of single-channel spectral images with higher scores according to the scoring results specifically includes: Based on the trained neural network model, scoring each single-channel spectral image within the combination of single-channel spectral images, and taking the sum of the scores of each single-channel spectral image as the score of the current combination of single-channel spectral images; Sorting according to the scores of each combination of single-channel spectral images from high to low, and screening a set number of combinations of single-channel spectral images with higher scores.

5. The design method of a fundus recognition hyperspectral imaging chip according to claim 1, characterized in that The step of constructing a corresponding hyperspectral imaging chip for each screened combination of single-channel spectral images, integrating it into a fundus imaging device for clinical evaluation, and obtaining the optimal hyperspectral imaging chip specifically includes: Constructing a corresponding hyperspectral imaging chip for each screened combination of single-channel spectral images; Integrating the hyperspectral imaging chips into a fundus imaging device respectively and successively performing clinical evaluations for tigroid fundus recognition, and determining the optimal hyperspectral imaging chip according to the clinical evaluation results.

6. An fundus imaging device for identifying tessellated fundus, characterized in that: It includes a fundus camera and a hyperspectral imaging chip camera that work together, where the hyperspectral imaging chip camera is constructed based on the optimal hyperspectral imaging chip obtained by the design method according to any one of claims 1 to 5.

7. The fundus imaging device according to claim 6, wherein, The specific process of the fundus imaging device for identifying tessellated fundus includes: Establish a physical connection between the fundus camera and the hyperspectral imaging chip camera, and keep the field of view and focus of the fundus camera and the hyperspectral imaging chip camera consistent; Control the fundus camera and the hyperspectral imaging chip camera to collect fundus images, and fuse the fundus images collected by the fundus camera and the hyperspectral imaging chip camera to generate a comprehensive image containing structural features and spectral information, so as to realize the identification of tessellated fundus.

8. The fundus imaging device according to claim 7, wherein, The establishment of the physical connection between the fundus camera and the hyperspectral imaging chip camera, and keeping the field of view and focus of the fundus camera and the hyperspectral imaging chip camera consistent specifically includes: Establish a physical connection between the fundus camera and the hyperspectral imaging chip camera through an adapter or a standard interface; Adjust the field of view and focus of the fundus camera and the hyperspectral imaging chip camera so that the field of view and focus of the fundus camera and the hyperspectral imaging chip camera are consistent.

9. The fundus imaging device according to claim 7, characterized in that, The control of the fundus camera and the hyperspectral imaging chip camera to collect fundus images, and the fusion of the fundus images collected by the fundus camera and the hyperspectral imaging chip camera to generate a comprehensive image containing structural features and spectral information specifically includes: Based on the cooperation of software and hardware, realize the synchronous control of the fundus camera and the hyperspectral imaging chip camera, so that the hyperspectral imaging chip camera collects fundus images of the same area as the fundus camera at different spectral bands; Fuse the fundus images collected by the fundus camera and the hyperspectral imaging chip camera through an image fusion algorithm to generate a comprehensive image containing structural features and spectral information.

10. An apparatus for designing a hyperspectral imaging chip for fundus recognition, characterized in that, The fundus identification hyperspectral imaging chip design device includes a processor, a memory, and a fundus identification hyperspectral imaging chip design program stored on the memory and executable by the processor. When the fundus identification hyperspectral imaging chip design program is executed by the processor, the steps of the fundus identification hyperspectral imaging chip design method according to any one of claims 1 to 5 are realized.

Images