Diabetic neuropathy auxiliary detection system based on ophthalmic images

Abstract

This invention relates to the field of medical image processing and disease-aided diagnosis technology, specifically to an auxiliary detection system for diabetic neuropathy based on ophthalmic images. The system comprises a fundus image acquisition module, a hysteresis analysis module, a region fusion module, a microaneurysm influence degree acquisition module, and a diabetic neuropathy detection module. It analyzes the grayscale change characteristics of pixels in FFA images to obtain the hysteresis degree of pixel filling with fluorescein. Then, it performs region fusion based on the hysteresis degree, and quantifies the influence degree of microaneurysms based on the average hysteresis degree and area of ​​the fused region, thereby enabling the diagnosis of diabetic neuropathy. This invention quantifies the filling and diffusion patterns of fluorescein in and around microaneurysms to obtain the influence degree of microaneurysms, thus enabling auxiliary detection of diabetic neuropathy. This assists doctors in efficiently and accurately detecting diabetic neuropathy, effectively improving diagnostic efficiency and accuracy.

Description

Technical Field

[0001] This invention relates to the field of medical image processing and disease-aided diagnosis technology, specifically to an auxiliary detection system for diabetic neuropathy based on ophthalmic images. Background Technology

[0002] Diabetes is a leading cause of peripheral neuropathy worldwide, leading to neuropathic pain, impacting quality of life, and potentially causing foot ulcers and amputations. Early diagnosis of diabetic peripheral neuropathy is crucial for preventing progression and subsequent morbidity and mortality. Current screening methods for diabetic peripheral neuropathy rely on the nervous system to detect moderate to severe peripheral neuropathy affecting large nerve fibers. A reference standard for small fiber damage is skin biopsy quantifying intradermal nerve fibers, but this method is not suitable for population-level screening of peripheral neuropathy.

[0003] In diabetic patients, long-term high blood sugar damages blood vessel walls, causing blood components to leak into the retinal layer, interlayer, and vitreous cavity, leading to retinal ischemia and hypoxia, often accompanied by retinopathy. Since the retina and peripheral nerves are related in terms of pathological mechanisms such as microvascular lesions and metabolic abnormalities, the application of ophthalmic imaging technology in the detection of diabetic peripheral neuropathy has gradually attracted attention.

[0004] Fluorescein angiography is a technique used to examine the fine blood vessels, vascular structures, and microcirculation changes in the fundus. The procedure involves injecting sodium fluorescein as a contrast agent via a vein in the forearm. As the sodium fluorescein enters the fundus vessels, a fundus camera with a filter is used to capture the fluorescence morphology of the vessels, allowing for the detection of any lesions. Therefore, fluorescein angiography can be used to detect diabetic neuropathy. However, quantitative analysis of diabetic neuropathy requires reliable extraction of image features from ophthalmic images. This process is reliance on operator experience and is time-consuming and labor-intensive. Therefore, finding a more efficient method for ophthalmic image analysis is a worthwhile research topic. Summary of the Invention

[0005] To address the above technical problems, this invention provides an auxiliary detection system for diabetic neuropathy based on ophthalmic imaging.

[0006] The ophthalmic imaging-based auxiliary detection system for diabetic neuropathy provided in the embodiments of this application specifically includes:

[0007] The fundus image acquisition module is used to acquire multiple frames of FFA images and filter out the pixels with grayscale changes in the FFA images over time.

[0008] The hysteresis analysis module is used to analyze the grayscale change characteristics of each of the grayscale change pixels and obtain the degree of hysteresis of each of the grayscale change pixels being filled with fluorescein.

[0009] The region fusion module is used to perform region fusion on the grayscale change pixels according to the lag degree to obtain several fusion regions;

[0010] The microaneurysm impact degree acquisition module is used to analyze the hysteresis characteristics of each fusion region based on the hysteresis degree, and obtain the microaneurysm impact degree by combining the regional area of ​​the fusion region.

[0011] The diabetic neuropathy detection module is used to determine diabetic neuropathy based on the degree of influence of the microaneurysms.

[0012] In some embodiments of the present invention, the hysteresis analysis module includes:

[0013] The grayscale time-series curve fitting unit is used to perform curve fitting on the grayscale value of each of the grayscale-changing pixels to obtain the grayscale time-series curve of each of the grayscale-changing pixels;

[0014] The hysteresis acquisition unit is used to analyze the occurrence time and location characteristics of the maximum gray value of each gray-level change pixel based on the gray-level time-series curve, and obtain the possible hysteresis degree of each gray-level change pixel.

[0015] The hysteresis correction unit is used to analyze the number of gray-level maxima of each gray-level changing pixel based on the gray-level time-series curve, correct the possible hysteresis degree, and obtain the hysteresis degree of each gray-level changing pixel being filled with fluorescein.

[0016] In some embodiments of the present invention, the hysteresis acquisition unit is configured as follows:

[0017] Based on the gray-scale time-series curve, the maximum gray value of the gray-scale changing pixel and the maximum value time corresponding to the maximum gray value are obtained, as well as the fluorescein initiation filling time of the gray-scale changing pixel and the initial gray value corresponding to the fluorescein initiation filling time are obtained.

[0018] The average filling rate of each pixel with grayscale change is obtained based on the maximum grayscale value, the initial grayscale value, the time of the maximum value, and the time of the fluorescein initial filling.

[0019] Based on the gray-scale time-series curves of all the gray-scale change pixels, the distribution of the number of gray-scale change pixels reaching the maximum gray-scale value at each time point is analyzed to obtain the filling peak time.

[0020] Based on the peak filling time, combined with the maximum filling time and the average filling rate, the possible lag degree of each grayscale change pixel is obtained.

[0021] In some embodiments of the present invention, based on the gray-scale time-series curves of all the gray-scale change pixels, the distribution of the number of gray-scale change pixels reaching the maximum gray-scale value at each time moment is analyzed to obtain the filling peak time, including:

[0022] Based on the gray-scale time-series curves of all the gray-scale change pixels, the total number of gray-scale change pixels reaching the maximum gray-scale value at each time point is counted, and the time point corresponding to the maximum total number of gray-scale change pixels is determined and denoted as the filling peak time point.

[0023] In some embodiments of the present invention, the hysteresis correction unit is configured as follows:

[0024] Peak detection is performed on the gray-level time-series curve of the gray-level changing pixels to obtain the number of gray-level maxima for each gray-level changing pixel;

[0025] Based on the number of gray-level maxima, the possible lag degree is corrected to obtain the lag degree of each gray-level change pixel being filled with phosphor.

[0026] In some embodiments of the present invention, the region fusion module is configured as follows:

[0027] Calculate the absolute value of the difference between the gray-level change pixel and the gray-level change pixels in its neighborhood;

[0028] Preset absolute value threshold for the difference;

[0029] When the absolute value of the difference in the degree of lag is less than the absolute value threshold of the difference, the neighboring gray-level change pixels and the gray-level change pixels are divided into the same region to obtain several fusion regions.

[0030] In some embodiments of the present invention, the microaneurysm impact degree acquisition module is configured as follows:

[0031] Calculate the average hysteresis of all grayscale-changing pixels within each fusion region;

[0032] Obtain the area of ​​each of the fused regions;

[0033] The degree of influence of microaneurysms is obtained based on the average lag and the area of ​​the region.

[0034] In some embodiments of the present invention, the diabetic neuropathy detection module is configured as follows:

[0035] Preset impact threshold;

[0036] Determine whether the impact of the microaneurysm is greater than the impact threshold;

[0037] If so, the patient is diagnosed with diabetic neuropathy, and the average hysteresis of all said fusion areas is marked.

[0038] In some embodiments of the present invention, the fundus image acquisition module includes:

[0039] The image acquisition unit is used to acquire multiple consecutive FFA images at a fixed sampling frequency after the patient has received intravenous injection of sodium fluorescein using a fundus fluorescein angiography system.

[0040] An image preprocessing unit is used to perform smoothing and grayscale processing on the FFA image to obtain an FFA grayscale image;

[0041] The grayscale change pixel extraction unit is used to analyze the change of grayscale value of each pixel in the FFA grayscale image over multiple consecutive frames as the frame sequence changes, set a grayscale change threshold, and filter out the grayscale change pixels of the FFA grayscale image in the time series.

[0042] In some embodiments of the present invention, the fluorescein initial filling time of the grayscale-changing pixel is the moment when the grayscale value of the grayscale-changing pixel begins to exceed 10% of the corresponding maximum grayscale value.

[0043] Compared with existing technologies, the ophthalmic imaging-based auxiliary detection system for diabetic neuropathy provided by this invention has the following beneficial effects:

[0044] This invention acquires multiple frames of FFA images through a fundus image acquisition module; analyzes the grayscale change characteristics of each pixel with grayscale changes through a hysteresis analysis module to obtain the degree of hysteresis in which each pixel is filled with fluorescein; performs region fusion on the pixels with grayscale changes based on the degree of hysteresis to obtain several fused regions; quantifies the degree of influence of microaneurysms based on the average hysteresis and area of ​​each fused region; and diagnoses diabetic neuropathy through a diabetic neuropathy detection module. This system quantifies the degree of influence of microaneurysms by quantitatively analyzing the filling and diffusion patterns of fluorescein in and around microaneurysms, combined with hysteresis and superposition effect correction coefficients, thereby enabling auxiliary detection of diabetic neuropathy and assisting doctors in efficiently and accurately detecting diabetic neuropathy, effectively improving diagnostic efficiency and accuracy. Attached Figure Description

[0045] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a schematic diagram of the basic components of an auxiliary detection system for diabetic neuropathy based on ophthalmic imaging, provided in one embodiment of the present invention.

[0047] Figure 2 This is a schematic diagram of the basic components of another auxiliary detection system for diabetic neuropathy based on ophthalmic imaging, provided in one embodiment of the present invention.

[0048] Figure 3 This is an example of an FFA grayscale image provided in one embodiment of the present invention. Detailed Implementation

[0049] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the ophthalmic imaging-based auxiliary detection system for diabetic neuropathy proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Terms such as “comprising,” “including,” or any other variations thereof are intended to cover a non-exclusive inclusion, such that a circuit structure, article, or device comprising a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such article or device. Without further limitation, an element defined by the phrase “comprising one…” does not exclude the presence of additional identical elements in the article or device that includes the element.

[0051] This invention addresses the following scenario: During the detection of diabetic neuropathy using fluorescein fundus angiography, when the fundus vessels are in a normal state, intravascular components do not leak out of the vessels, so fluorescein only changes within the vessels. However, in diabetic patients, due to factors such as damaged vessel walls, microaneurysms can form, allowing intravascular components to leak out. Specifically, fluorescein leaks through the wall of the microaneurysm and then diffuses into the extracellular space, with the brightness change in the extracellular space lagging behind the brightness change in the microaneurysm. Furthermore, the diffusion at multiple microaneurysms exhibits a superposition phenomenon, which manifests as a higher grayscale value in the image of that region compared to the surrounding areas, and multiple grayscale peaks may appear during the filling process of the superimposed region.

[0052] Therefore, the main objective of this invention is to detect diabetic neuropathy by quantifying the temporal changes in images of microaneurysms and their surrounding areas, thereby assisting clinicians in improving diagnostic efficiency and accuracy.

[0053] The overall implementation scheme of this invention is as follows: Based on ophthalmic images, by analyzing the dynamic changes of microaneurysms presented by multiple frames of FFA (Fundus Fluorescein Angiography) images in the dynamic angiography sequence, the filling and diffusion patterns of fluorescein in the microaneurysm and its surrounding areas, as well as the changes in the diffusion superposition area, are quantitatively analyzed. The degree of lag in filling each area with fluorescein is quantified, thereby calculating the degree of influence of microaneurysms, judging the possibility of diabetic neuropathy in patients, and marking the average lag degree of each area to help doctors judge diabetic neuropathy.

[0054] The following description, in conjunction with the accompanying drawings, details the specific solution of an auxiliary detection system for diabetic neuropathy based on ophthalmic imaging provided by the present invention.

[0055] Please see Figure 1 This illustrates the basic components of an ophthalmic imaging-based auxiliary detection system for diabetic neuropathy provided in one embodiment of the present invention.

[0056] like Figure 1 As shown in the figure, an embodiment of the present invention provides an auxiliary detection system for diabetic neuropathy based on ophthalmic imaging. The system mainly includes a fundus image acquisition module 10, a hysteresis analysis module 20, a region fusion module 30, a microaneurysm influence degree acquisition module 40, and a diabetic neuropathy detection module 50. Wherein:

[0057] The fundus image acquisition module 10 is used to acquire multiple frames of FFA images and filter out pixels with grayscale changes over time in the FFA images. Further, such as... Figure 2As shown, the fundus image acquisition module 10 includes an image acquisition unit 11, an image preprocessing unit 12, and a grayscale variation pixel extraction unit 13, specifically:

[0058] Image acquisition unit 11 is used to acquire multiple consecutive FFA images at a fixed sampling frequency after the patient receives an intravenous injection of sodium fluorescein using a fundus fluorescein angiography system. Specifically, after the patient receives an intravenous injection of sodium fluorescein, the fundus fluorescein angiography system is used to acquire multiple consecutive FFA images at a fixed sampling frequency. In this embodiment, the sampling frequency can be set to 60 times per second (the specific sampling frequency can be set according to the parameters and performance of the fundus fluorescein angiography system). The multiple consecutive FFA images are then arranged in chronological order to obtain a sequence of consecutive FFA images (time series).

[0059] Image preprocessing unit 12 is used to smooth and grayscale the FFA images to obtain an FFA grayscale image. Specifically, it preprocesses all FFA images acquired by image acquisition unit 11, including: firstly, smoothing the FFA images using a Gaussian filtering algorithm to remove high-frequency noise; then, converting the smoothed FFA images to grayscale to obtain the FFA grayscale image. Figure 3 As shown.

[0060] The grayscale change pixel extraction unit 13 is used to analyze the change of grayscale value of each pixel in a continuous multi-frame FFA grayscale image with the frame sequence, set the grayscale change threshold, and filter out the grayscale change pixels of the FFA grayscale image in the time series. The specific configuration is as follows: First, based on the SIFT algorithm (Scale-Invariant Feature Transform), feature point detection is performed to extract blood vessel bifurcation points in each frame of FFA grayscale image. Registration is then performed on multiple frames of FFA grayscale images to obtain the corresponding pixel for each pixel in the multiple frames. Next, the average grayscale value difference of each pixel in all adjacent frames of FFA grayscale images is calculated (the grayscale value of the pixel in the next frame minus the grayscale value of the corresponding pixel in the previous frame). A grayscale change threshold is set. If the average grayscale value difference of each pixel in adjacent frames of FFA grayscale images is greater than the grayscale change threshold, it indicates that the grayscale value of that pixel has changed significantly over time. All pixels with significant grayscale value changes over time are selected, resulting in grayscale change pixels in the FFA grayscale image over time. These grayscale change pixels are the pixels that may represent blood vessels, microaneurysms, and the range of fluorescein leakage and diffusion, and are used as the objects of subsequent analysis.

[0061] The hysteresis analysis module 20 is used to analyze the grayscale change characteristics of each grayscale change pixel and obtain the degree of hysteresis of each grayscale change pixel being filled with fluorescein.

[0062] Because fluorescein needs to infiltrate from the wall of the microaneurysm and then diffuse into the extracellular space, the brightness change in the extracellular space lags behind the brightness change in the microaneurysm. This manifests as the fluorescein filling rate of the microaneurysm being greater than that of the surrounding extracellular space, and the fluorescein reaching its peak brightness lags behind that of the microaneurysm. Therefore, the presence of a microaneurysm can be determined by the filling rate and lag time.

[0063] Based on the above analysis, in an embodiment of the present invention, a hysteresis analysis module 20 is provided to analyze the grayscale change characteristics of each pixel with grayscale change, and to obtain the degree of hysteresis in which each pixel with grayscale change is filled with fluorescein. Further, as... Figure 2 As shown, the hysteresis analysis module 20 includes a gray-scale time-series curve fitting unit 21, a hysteresis acquisition unit 22, and a hysteresis correction unit 23, wherein:

[0064] The grayscale temporal curve fitting unit 21 is used to perform curve fitting on the grayscale value of each grayscale-changing pixel to obtain the grayscale temporal curve of each grayscale-changing pixel. The specific configuration is as follows: using the frame sequence (time series) as the X-axis and the grayscale value of the grayscale-changing pixel as the Y-axis, points are marked on a two-dimensional sample space. Then, the sample points of the grayscale value of each grayscale-changing pixel are fitted. The least squares method is used to make the fitted curve approximate the grayscale value data sample points of each grayscale-changing pixel, finally obtaining the grayscale temporal curve of the first pixel. The gray-level time-series curve of the gray-level value change of a pixel over time is denoted as: .

[0065] The hysteresis acquisition unit 22 is used to analyze the occurrence time and location characteristics of the maximum gray value of each gray-level change pixel based on the gray-level time-series curve, and obtain the possible hysteresis degree of each gray-level change pixel. The specific configuration is as follows:

[0066] First, based on the gray-scale time-series curve, the maximum gray value of the pixel with gray-scale changes and the time of its corresponding maximum value are obtained. Additionally, since the diffusion of fluorescein within blood vessels takes time, the closer to the optic nerve, the earlier the fluorescein begins to fill. Therefore, to avoid the influence of positional relationships, it is necessary to calculate and determine the initial filling time of fluorescein for each pixel with gray-scale changes, i.e., the moment when the curve begins to rise significantly from the baseline gray value (the background gray value before fluorescein injection). This moment is recorded as the initial filling time of fluorescein for the pixel with gray-scale changes. The initial filling time of fluorescein for a pixel with gray-scale changes can be set as the moment when the gray value of the pixel with gray-scale changes begins to exceed the corresponding maximum gray value by 10%. Thus, based on the gray-scale time-series curve, the initial filling time of fluorescein for the pixel with gray-scale changes and the initial gray value corresponding to this initial filling time are obtained.

[0067] Then, based on the maximum gray value, initial gray value, time of maximum value, and time of fluorescein initiation filling, the average filling rate of each pixel with gray value change is obtained; then the first... The formula for calculating the average filling rate of pixels with grayscale changes is:

[0068]

[0069] In the formula, Indicates the first Average filling rate of pixels with grayscale changes; Indicates the first The grayscale time curve of each pixel with grayscale change The maximum grayscale value on; Indicates the first The grayscale time curve of each pixel with grayscale change The initial grayscale value on, i.e., the first The initial gray value corresponding to the gray value of each pixel with gray-level change; Indicates the first The grayscale time curve of each pixel with grayscale change The maximum value time on, i.e. the first time The time corresponding to the maximum gray value of a pixel with gray level change; Indicates the first The grayscale time curve of each pixel with grayscale change The initial filling time of fluorescein on the surface.

[0070] No. The average filling rate of each grayscale pixel from the initial filling time of the fluorescein to the maximum filling time, average filling rate The smaller the value, the more likely it is to be an extracellular space caused by the exudation and diffusion of the tumor wall of a microaneurysm.

[0071] Furthermore, based on the grayscale time-series curves of all grayscale-changing pixels, the distribution of the number of grayscale-changing pixels reaching the maximum grayscale value at each time step is analyzed to obtain the filling peak time. Specifically, based on the grayscale time-series curves of all grayscale-changing pixels, the total number of grayscale-changing pixels reaching the maximum grayscale value at each time step is counted, and the time corresponding to the maximum total number of grayscale-changing pixels is determined and denoted as the filling peak time.

[0072] Finally, since fluorescein needs to infiltrate from the wall of the microaneurysm and then diffuse into the extracellular space, the filling rate of fluorescein in the microaneurysm is greater than that in the surrounding extracellular space. Furthermore, the brightness change in the extracellular space near the microaneurysm lags behind the brightness change in the microaneurysm. Therefore, based on the peak filling time, combined with the maximum filling time and the average filling rate, the possible lag degree of each grayscale change pixel is obtained, and the first... The formula for calculating the possible lag of a pixel with grayscale changes is:

[0073]

[0074] In the formula, Indicates the first The possible lag of a pixel with a grayscale change; Indicates the first Average filling rate of pixels with grayscale changes; Indicates the first The grayscale time curve of each pixel with grayscale change The maximum value time on, i.e. the first time The time corresponding to the maximum gray value of a pixel with gray level change; Indicates the first The peak filling moment of each pixel with grayscale variation; Indicates taking the absolute value; This represents the linear normalization function.

[0075] For the first The average filling rate of each grayscale change pixel. The smaller the average filling rate, the stronger the lag, and the more likely the filling process is formed by exudation and diffusion from the tumor wall of the microaneurysm. Indicates the first The difference between the peak grayscale value of a pixel and the time corresponding to the moment with the highest number of pixels exhibiting grayscale peak values ​​during the entire filling process. When fluorescein propagates in blood vessels, there is no penetration phenomenon; its propagation is rapid, therefore... This is manifested as multiple pixels (blood vessels, microaneurysms) simultaneously reaching their peak values, through... The value reflects the degree of time lag. The larger the value, the more obvious the lag, indicating that the grayscale change of the pixel is more likely to be formed by the exudation and diffusion of the tumor wall of the microaneurysm.

[0076] The hysteresis correction unit 23 is used to analyze the number of gray-level maxima of each gray-level changing pixel based on the gray-level time-series curve, correct the possible hysteresis, and obtain the hysteresis degree of each gray-level changing pixel being filled with fluorescein.

[0077] Fluorescein seeps from the wall of a microaneurysm and then diffuses into the extracellular space. This diffusion is directional, with the microaneurysm being the brightest and the brightness decreasing as it diffuses outwards, showing a decreasing trend from the center to the periphery. Multiple microaneurysms may exist on the vessel wall, causing overlapping features during the diffusion process. This overlapping results in a larger gray value in the overlapping area of ​​the FFA grayscale image, which in turn increases the calculated average filling rate and reduces the hysteresis. Therefore, it is necessary to identify the overlapping area and correct its hysteresis based on the characteristics of the overlapping area.

[0078] For pixels with non-overlapping grayscale changes, the grayscale temporal curve formed during the filling process... It exhibits a single peak value, while the grayscale changes of pixels in the superimposed region are affected by the exudation of multiple microaneurysms, thus its grayscale time-series curve... The phenomenon is characterized by multiple peaks. The more peaks there are, the greater the degree of diffusion overlap, indicating that the pixel with the grayscale change is more affected by the diffusion of multiple microaneurysms.

[0079] Based on the above analysis, in some embodiments of the present invention, a hysteresis correction unit 23 is set up to analyze the number of gray-scale maxima of each gray-scale changing pixel based on the gray-scale time-series curve, correct the possible hysteresis degree, and obtain the hysteresis degree of each gray-scale changing pixel being filled with fluorescein.

[0080] Because the pixels in the overlay area are affected by multiple microaneurysms, their grayscale change curves will exhibit multiple peaks, leading to potential lag in previously calculated values. This may be inaccurate (the spread and superposition of multiple microaneurysms will cause the gray value of the area to rise faster, thus affecting the average filling rate). Increased size may lead to a lag in the degree of lag. (Become smaller). Therefore, it is necessary to correct for the possible lag based on the number of gray-level maxima. The more superimposed microaneurysms there are (i.e., the larger the number of gray-level maxima), the stronger the superposition effect and the higher the average filling rate. The more it is overestimated, the greater the correction required. Therefore, the hysteresis correction unit 23 is specifically configured as follows: Peak detection is performed on the gray-level time-series curve of the pixels with gray-level changes to obtain the number of gray-level maxima for each pixel with gray-level changes; based on the number of gray-level maxima, the possible hysteresis is corrected to obtain the hysteresis degree of each pixel with gray-level changes being filled with fluorescein. Constructing the first... The formula for calculating the lag degree of phosphor filling in each grayscale-changing pixel is as follows:

[0081]

[0082] In the formula, Indicates the first The degree of lag in filling the grayscale change pixel with phosphor, i.e., the lag of the first grayscale change pixel. The possible lag of a pixel with grayscale changes after being filled with phosphor; Indicates the first The possible lag of a pixel with a grayscale change; Indicates the first The number of gray-level maxima of pixels with gray-level changes; This represents the linear normalization function.

[0083] Based on the number of grayscale maxima Correcting the degree of lag, i.e. The larger the value, the stronger the superposition effect; the corrected value... The larger.

[0084] The region fusion module 30 is used to perform region fusion on pixels with grayscale changes according to the degree of lag, so as to obtain several fused regions.

[0085] The region fusion module 30 is used to perform region fusion on pixels with grayscale changes based on the degree of lag, obtaining several fusion regions. Specifically, it is configured as follows: A region growing algorithm is used to calculate the absolute value of the difference in lag between a pixel with grayscale changes and its eight neighboring pixels with grayscale changes; a preset threshold for the absolute value of the difference is set, which can be 0.2; when the absolute value of the difference in lag is less than the threshold, the neighboring pixels with grayscale changes are grouped into the same region, and this process is repeated for all pixels with grayscale changes to obtain several fusion regions. Since the specific implementation process of using the region growing algorithm to fuse pixels with grayscale changes to obtain multiple fusion regions is existing technology, it will not be elaborated here.

[0086] The microaneurysm impact degree acquisition module 40 is used to analyze the hysteresis characteristics of each fusion region based on the hysteresis degree, and obtain the microaneurysm impact degree by combining the regional area of ​​the fusion region.

[0087] During the filling process, fluorescein needs to leak from the wall of the microaneurysm and then diffuse into the extracellular space. This is manifested in the degree of lag and the extent of diffusion. The greater the lag and the larger the diffusion area, the greater the impact of the microaneurysm and the higher the likelihood of diabetic neuropathy in the patient. Therefore, the degree of impact of the microaneurysm is calculated based on the lag of the corrected fusion region. The greater the lag and the larger the area of ​​the fusion region, the greater the impact of the microaneurysm.

[0088] Based on the above analysis, in this embodiment of the invention, a microaneurysm influence degree acquisition module 40 is set up to analyze the hysteresis characteristics of each fusion region based on the hysteresis degree, and obtain the microaneurysm influence degree by combining the area of ​​the fusion region. Specifically, the configuration is as follows: calculate the average hysteresis degree of all grayscale change pixels within each fusion region; obtain the area of ​​each fusion region; and obtain the microaneurysm influence degree based on the average hysteresis degree and the area. Therefore, the formula for calculating the microaneurysm influence degree is constructed as follows:

[0089]

[0090] In the formula, Indicates the degree of impact of microaneurysms; Indicates the first The average hysteresis of all grayscale-changing pixels within the fusion region; Indicates the first The area of ​​each integration zone; Indicates the total number of merged regions; This represents the linear normalization function.

[0091] The overall impact of microaneurysms is obtained by summing the lag degree and the area occupied by each fusion region. The larger the lag degree and the larger the area affected by the microaneurysms, the higher the possibility of diabetic neuropathy.

[0092] The diabetic neuropathy detection module 50 is used to determine diabetic neuropathy based on the degree of influence of microaneurysms.

[0093] The diabetic neuropathy detection module 50, based on the microaneurysm influence level obtained by the microaneurysm influence level acquisition module 40, determines the likelihood of a patient having diabetic neuropathy. A higher microaneurysm influence level value indicates a higher likelihood of diabetic neuropathy. It also marks the average hysteresis degree of each fusion region to assist doctors in diagnosing diabetic neuropathy. The specific configuration of the diabetic neuropathy detection module 50 is as follows: a preset influence level threshold, which can be set to 0.7; it determines whether the microaneurysm influence level exceeds the influence level threshold; if so, i.e. If the hysteresis is positive, the patient is diagnosed with diabetic neuropathy, and the average hysteresis of all fusion areas is marked to assist the doctor in diagnosing diabetic neuropathy; if not, then... This indicates that the patient has not developed diabetic neuropathy, and also marks the average lag degree of all fusion areas, assisting doctors in diagnosing diabetic neuropathy.

[0094] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0095] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A system for assisting in the detection of diabetic neuropathy based on ophthalmic imaging, characterized in that, The system includes: The fundus image acquisition module is used to acquire multiple frames of FFA images and filter out the pixels with grayscale changes in the FFA images over time. The hysteresis analysis module is used to analyze the grayscale change characteristics of each of the grayscale change pixels and obtain the degree of hysteresis of each of the grayscale change pixels being filled with fluorescein. The region fusion module is used to perform region fusion on the grayscale change pixels according to the lag degree to obtain several fusion regions; The microaneurysm impact degree acquisition module is used to analyze the hysteresis characteristics of each fusion region based on the hysteresis degree, and obtain the microaneurysm impact degree by combining the regional area of ​​the fusion region. A diabetic neuropathy detection module is used to determine diabetic neuropathy based on the degree of influence of the microaneurysms; the hysteresis analysis module includes: The grayscale time-series curve fitting unit is used to perform curve fitting on the grayscale value of each of the grayscale-changing pixels to obtain the grayscale time-series curve of each of the grayscale-changing pixels; The hysteresis acquisition unit is used to analyze the occurrence time and location characteristics of the maximum gray value of each gray-level change pixel based on the gray-level time-series curve, and obtain the possible hysteresis degree of each gray-level change pixel. The hysteresis correction unit is used to analyze the number of gray-level maxima of each gray-level changing pixel based on the gray-level time-series curve, correct the possible hysteresis degree, and obtain the hysteresis degree of each gray-level changing pixel being filled with fluorescein.

2. The auxiliary detection system for diabetic neuropathy based on ophthalmic imaging according to claim 1, characterized in that, The hysteresis acquisition unit is configured as follows: Based on the gray-scale time-series curve, the maximum gray value of the gray-scale changing pixel and the maximum value time corresponding to the maximum gray value are obtained, as well as the fluorescein initiation filling time of the gray-scale changing pixel and the initial gray value corresponding to the fluorescein initiation filling time are obtained. The average filling rate of each pixel with grayscale change is obtained based on the maximum grayscale value, the initial grayscale value, the time of the maximum value, and the time of the fluorescein initial filling. Based on the gray-scale time-series curves of all the gray-scale change pixels, the distribution of the number of gray-scale change pixels reaching the maximum gray-scale value at each time point is analyzed to obtain the filling peak time. Based on the peak filling time, combined with the maximum filling time and the average filling rate, the possible lag degree of each grayscale change pixel is obtained.

3. The auxiliary detection system for diabetic neuropathy based on ophthalmic imaging according to claim 2, characterized in that, Based on the gray-level time-series curves of all the aforementioned gray-level change pixels, the distribution of the number of gray-level change pixels reaching the maximum gray-level value at each time step is analyzed to obtain the filling peak time, including: Based on the gray-scale time-series curves of all the gray-scale change pixels, the total number of gray-scale change pixels reaching the maximum gray-scale value at each time point is counted, and the time point corresponding to the maximum total number of gray-scale change pixels is determined and denoted as the filling peak time point.

4. The auxiliary detection system for diabetic neuropathy based on ophthalmic imaging according to claim 1, characterized in that, The hysteresis correction unit is configured as follows: Peak detection is performed on the gray-level time-series curve of the gray-level changing pixels to obtain the number of gray-level maxima for each gray-level changing pixel; Based on the number of gray-level maxima, the possible lag degree is corrected to obtain the lag degree of each gray-level change pixel being filled with phosphor.

5. The auxiliary detection system for diabetic neuropathy based on ophthalmic imaging according to claim 1, characterized in that, The region fusion module is configured as follows: Calculate the absolute value of the difference between the gray-level change pixel and the gray-level change pixels in its neighborhood; Preset absolute value threshold for the difference; When the absolute value of the difference in the degree of lag is less than the absolute value threshold of the difference, the neighboring gray-level change pixels and the gray-level change pixels are divided into the same region to obtain several fusion regions.

6. The auxiliary detection system for diabetic neuropathy based on ophthalmic imaging according to claim 1, characterized in that, The module for obtaining the degree of impact of microaneurysms is configured as follows: Calculate the average hysteresis of all grayscale-changing pixels within each fusion region; Obtain the area of ​​each of the fused regions; The degree of influence of microaneurysms is obtained based on the average lag and the area of ​​the region.

7. The auxiliary detection system for diabetic neuropathy based on ophthalmic imaging according to claim 1, characterized in that, The diabetic neuropathy detection module is configured as follows: Preset impact threshold; Determine whether the impact of the microaneurysm is greater than the impact threshold; If so, the patient is diagnosed with diabetic neuropathy, and the average hysteresis of all said fusion areas is marked.

8. The auxiliary detection system for diabetic neuropathy based on ophthalmic imaging according to claim 1, characterized in that, The fundus image acquisition module includes: The image acquisition unit is used to acquire multiple consecutive FFA images at a fixed sampling frequency after the patient has received intravenous injection of sodium fluorescein using a fundus fluorescein angiography system. An image preprocessing unit is used to perform smoothing and grayscale processing on the FFA image to obtain an FFA grayscale image; The grayscale change pixel extraction unit is used to analyze the change of grayscale value of each pixel in the FFA grayscale image over multiple consecutive frames as the frame sequence changes, set a grayscale change threshold, and filter out the grayscale change pixels of the FFA grayscale image in the time series.

9. The ophthalmic imaging-based auxiliary detection system for diabetic neuropathy according to claim 2, characterized in that, The initial filling time of the phosphosome of the grayscale-changing pixel is the moment when the grayscale value of the grayscale-changing pixel begins to exceed 10% of the corresponding maximum grayscale value.

Images