Leak detection method, device and medium for insulin

Abstract

The present disclosure describes a method, device and medium for detecting insulin leakage, which comprises: after a patient injects insulin, using an adsorption material to adsorb the insulin leakage at a target position related to the amount of insulin leakage, and collecting an image of the adsorption material with the adsorbed insulin leakage to obtain a leakage image; performing contrast enhancement on the leakage image to obtain an enhanced image; performing binary classification based on the pixels in the enhanced image to obtain two categories of pixels; selecting target category pixels from the two categories of pixels based on a first gray interval before the adsorption material adsorbs insulin and a second gray interval after the adsorption material adsorbs insulin; performing convex hull fitting on the region corresponding to the target category pixels to determine a leakage region; determining a leakage area based on the resolution of the leakage image and the leakage region; and determining the amount of insulin leakage corresponding to the leakage area based on the adsorption capacity and the leakage area. Thus, the amount of insulin leakage can be effectively, conveniently and non-invasively obtained.

Description

Technical Field

[0001] This disclosure generally relates to the medical field of health-related information systems, and specifically to a method, device, and medium for detecting insulin leakage. Background Technology

[0002] In the medical field (such as diabetes or vaccination), strictly controlling injection dosage is often used to treat or prevent diseases. Taking diabetes as an example, diabetes is a lifelong chronic endocrine disease characterized by insulin resistance and reduced pancreatic function. It can trigger a series of metabolic disorders, including high protein and high fat intake. Diabetic patients generally have poor pancreatic function, which easily leads to unstable blood sugar levels. Some patients often need to inject insulin to control their blood sugar levels.

[0003] Currently, insulin injection has become an important way to lower blood sugar in the treatment of diabetes (such as type 2 diabetes). By injecting a certain dose of insulin, patients can control their blood sugar levels.

[0004] However, relevant medical clinical studies have shown that during insulin injection, problems such as drug residue at the outlet of the injection device or leakage of the drug solution can easily occur. This leads to a reduction in the dose of insulin injected into the patient's body, which can easily cause unstable blood sugar control and thus reduce the effectiveness of blood sugar control. However, there is no relevant research on how to quantitatively detect leakage. Summary of the Invention

[0005] This disclosure is made in view of the above-mentioned circumstances, and its purpose is to provide a method, device and medium for detecting leakage in insulin that can effectively, conveniently and non-invasively obtain the amount of leakage.

[0006] Therefore, a first aspect of this disclosure provides a method for detecting insulin leakage, used to detect the amount of leakage during insulin injection in a patient. The method includes: acquiring a leakage image, wherein the leakage image is obtained by acquiring an image of the adsorbent material with the adsorbed leakage at a target location related to the leakage amount after the patient injects insulin, the adsorbent material having the ability to adsorb insulin; enhancing the contrast of the leakage image to obtain an enhanced image; performing binary classification based on pixels in the enhanced image to obtain two categories of pixels; selecting a target category of pixels from the two categories of pixels based on a first grayscale range for the adsorbent material before insulin adsorption and a second grayscale range for the adsorbent material after insulin adsorption; performing connectivity analysis on the region corresponding to the target category of pixels to determine at least one connecting component, obtaining the convex hull of each connecting component in the at least one connecting component, and using the convex hull corresponding to the at least one connecting component as the leakage region; determining the leakage area based on the resolution of the leakage image and the leakage region; and determining the leakage amount corresponding to the leakage area based on the adsorption capacity and the leakage area. In this approach, the amount of leakage is effectively collected using adsorbent materials. Based on the grayscale changes before and after insulin adsorption, target category pixels are identified. Convex hull fitting is then used to adjust the regions corresponding to these target category pixels, intelligently and conveniently determining regular and accurate leakage areas. The leakage situation is then quantitatively determined based on the leakage area and the adsorption capacity of the adsorbent material. This allows for the effective, convenient, and non-invasive acquisition of quantitative leakage information.

[0007] Furthermore, in the leakage detection method according to the first aspect of this disclosure, optionally, when selecting pixels of the target category from the two categories of pixels, if the first grayscale range is greater than the second grayscale range, then the category with the smaller average grayscale value of the pixels in the two categories is selected as the target category; otherwise, the category with the larger average grayscale value of the pixels in the two categories is selected as the target category. Thus, the target category of pixels can be determined based on the grayscale change before and after insulin adsorption by the adsorbent material.

[0008] Furthermore, in the leakage detection method according to the first aspect of this disclosure, optionally, the adsorbent material is a nitrocellulose membrane, the nitrocellulose membrane is white, and the first grayscale range is larger than the second grayscale range. Therefore, it is possible to determine the target category of pixels based on the grayscale change before and after insulin adsorption onto the white nitrocellulose membrane.

[0009] In addition, in the leakage detection method according to the first aspect of this disclosure, optionally, the connectivity analysis is an eight-neighbor connectivity analysis, and the convex hull is the smallest convex polygon corresponding to each connected component.

[0010] Additionally, in the leakage detection method according to the first aspect of this disclosure, optionally, in the contrast enhancement, the color space of the leakage image is converted to a target color space with a lightness component, histogram equalization is performed on the gray values ​​of the lightness component, and the gray values ​​of the lightness component are converted back to the RGB space via the histogram equalized target color space to obtain the enhanced image, wherein the cumulative distribution function corresponding to the histogram equalization satisfies the formula: Where k represents the grayscale value of the brightness component, and n k S represents the number of pixels in the leakage image with a grayscale value of k for the brightness component, N represents the total number of pixels in the leakage image, and S represents the number of pixels in the leakage image. k The gray value of the lightness component, denoted as k, is the gray value after histogram equalization. This improves the contrast between the leaking area and other areas in the enhanced image.

[0011] Furthermore, in the leakage detection method according to the first aspect of this disclosure, optionally, the binary classification is K-means clustering, where K is set to 2, the pixels in the enhanced image are samples, and the grayscale values ​​of the three channels in the RGB space of each sample are features. Unsupervised clustering is performed on the samples to obtain the two categories. Thus, it is possible to obtain pixels of two categories based on K-means clustering.

[0012] Furthermore, in the leakage detection method according to the first aspect of this disclosure, optionally, the leakage image includes a pattern of a reference object of known size; the number of pixels in the leakage region is determined based on the resolution of the leakage image; the area of ​​each pixel is determined based on the pattern of the reference object in the leakage image; and the leakage area is determined based on the number of pixels and the area of ​​each pixel. Thus, the leakage area can be determined based on the reference object.

[0013] Additionally, in the leakage detection method according to the first aspect of this disclosure, optionally, the target location includes the insulin outlet of the injection device for injecting insulin and the skin surface surrounding the patient's injection site. In this case, leakage at the target location can substantially encompass the medication not injected into the patient's body, thereby enabling effective analysis or evaluation based on a more comprehensive leakage volume in some leakage volume applications.

[0014] A second aspect of this disclosure provides an electronic device including at least one processing circuit configured to perform the leakage detection method according to the first aspect of this disclosure.

[0015] A third aspect of this disclosure provides a computer-readable storage medium storing at least one instruction that, when executed by a processor, implements the leakage detection method according to the first aspect of this disclosure.

[0016] According to this disclosure, a method, device, and medium for detecting leakage in insulin can be provided, which can effectively, conveniently, and non-invasively obtain the amount of leakage. Attached Figure Description

[0017] This disclosure will now be explained in further detail by way of example only with reference to the accompanying drawings, in which:

[0018] Figure 1 This is a schematic diagram illustrating an example of the detection environment involved in this disclosure.

[0019] Figure 2 This is an exemplary flowchart illustrating a leakage detection method as described in this disclosure.

[0020] Figure 3 This is an exemplary flowchart illustrating the acquisition of leakage images based on a reference object, as described in this disclosure.

[0021] Figure 4 This is a schematic diagram showing a leakage image using a reference object, as illustrated in the examples of this disclosure.

[0022] Figure 5 This is an exemplary flowchart illustrating the contrast enhancement involved in the examples of this disclosure.

[0023] Figure 6A This shows a histogram of the gray values ​​of the lightness component involved in the example of this disclosure before histogram equalization.

[0024] Figure 6B This is a histogram showing the gray values ​​of the lightness component involved in the example of this disclosure after histogram equalization.

[0025] Figure 7 This is an exemplary flowchart illustrating the convex hull fitting process involved in the examples of this disclosure.

[0026] Figure 8 This is an exemplary flowchart illustrating an analysis based on leakage volume as described in this disclosure. Detailed Implementation

[0027] The preferred embodiments of this disclosure are described in detail below with reference to the accompanying drawings. In the following description, the same reference numerals are used for the same components, and repeated descriptions are omitted. Furthermore, the drawings are merely schematic diagrams, and the proportions of the components or the shapes of the components may differ from actual figures. It should be noted that the terms "comprising" and "having," and any variations thereof, in this disclosure, do not necessarily limit the process, method, system, product, or apparatus to the explicitly listed steps or units, but may include or have other steps or units not explicitly listed or inherent to these processes, methods, products, or apparatuses. All methods described in this disclosure may be performed in any suitable order unless otherwise indicated herein or clearly contradicted by the context.

[0028] As mentioned above, during drug injection (such as insulin or vaccines), problems such as drug residue remaining at the injection device's outlet (i.e., the drug outlet) or leakage can easily occur. The main reasons for this problem are: the outlet of the injection device is relatively narrow, and the drug takes a relatively long time to be injected into the body; moreover, as the drug dosage increases, the absorption rate of the drug at the injection site (also known as the injection location) gradually slows down; and if the injection components (e.g., the needle of a needle-based injection device or the nozzle of a needleless injection device) are not removed in time after the injection, drug leakage may occur at the outlet when the external temperature changes (from cold to hot).

[0029] The inventors discovered through research that leakage volume is closely related to both the injection device and the injection dosage. Therefore, a leakage detection method, device, and medium for insulin are proposed. The leakage detection method, device, and medium for insulin disclosed herein can effectively, conveniently, and non-invasively obtain leakage volume. Leakage detection methods are sometimes also referred to as detection methods or leakage assessment methods. Furthermore, for insulin, leakage volume can affect blood glucose fluctuations. In some examples, the impact of leakage volume on blood glucose fluctuations can also be analyzed.

[0030] The leakage detection method described in this disclosure can be applied to detect the amount of leakage during drug injection. The drug may include, but is not limited to, insulin or vaccines. The following description uses insulin as an example of the drug in this disclosure and does not constitute a limitation on the technical solutions provided in this disclosure. Figure 1 This is a schematic diagram illustrating an example of the detection environment involved in this disclosure.

[0031] like Figure 1 As shown, the detection environment may include an adsorbent material 10 for adsorbing insulin. The adsorbent material 10 can be any material capable of adsorbing insulin, a protein. After the patient has injected insulin, the adsorbent material 10 can be applied to the target location. Figure 1 The diagram shows an adsorbent material 10 attached to the skin surface around the patient's injection site to collect the leakage. The adsorbent material 10 with the collected leakage is placed near a reference object 20 located on the same plane as the adsorbent material 10, and the image of the adsorbent material 10 and the reference object 20 can be acquired by the computing device 30 as a leakage image.

[0032] Continue to refer to Figure 1 The detection environment may include a computing device 30. The computing device 30 can implement a leakage detection method, which can acquire images of the adsorbent material 10 and the reference object 20 as leakage images, and obtain the leakage amount based on the leakage images.

[0033] In some examples, the computing device 30 can also determine the impact of leakage volume on the patient's blood glucose fluctuations. In some examples, the computing device 30 can also receive blood glucose data transmitted from a continuous glucose monitoring system. Thus, the impact of leakage volume on the patient's blood glucose fluctuations can be determined by combining the blood glucose data. The continuous glucose monitoring system can collect the patient's blood glucose data and transmit it to the computing device 30. Specifically, the sensors in the continuous glucose monitoring system can be partially or completely placed under the patient's skin to collect the patient's blood glucose data, which can be transmitted to the computing device 30 via a receiver located outside the patient's body.

[0034] In some examples, the leakage detection method may be stored as computer program instructions on the computing device 30 and executed by the computing device 30. In some examples, the computing device 30 may include one or more processors and one or more memories. The processor can execute the computer program instructions. The memory can be used to store the computer program instructions. The processor of the computing device 30 can implement the leakage detection method by executing the computer program instructions on the memory.

[0035] In some examples, computing device 30 may include, but is not limited to, laptops, tablets, mobile phones, desktop computers, or virtual computers (virtual computers can refer to virtual machines that are simulated by software, have full hardware system functionality, and run in a completely isolated environment). In some examples, the number of computing devices 30 may be multiple.

[0036] This disclosure discloses a method, device, and medium for detecting leakage of insulin. The method involves acquiring a leakage image of an adsorbent material 10 with adsorbed leakage fluid, enhancing the contrast of the leakage image to obtain an enhanced image, classifying the pixels in the enhanced image to obtain pixel categories, selecting target category pixels from different categories based on grayscale changes before and after insulin adsorption, determining the leakage region based on the area corresponding to the target category pixels, determining the leakage area based on the leakage region, and then determining the leakage amount corresponding to the leakage area based on the adsorption capacity of the adsorbent material 10 for adsorbing insulin and the leakage area. In this configuration, the adsorbent material 10 effectively collects the leakage fluid, and the target category pixels are determined based on the characteristics of the grayscale changes before and after insulin adsorption to intelligently and conveniently identify accurate leakage regions. The leakage situation is then quantitatively determined based on the leakage region and the adsorption capacity of the adsorbent material 10. Therefore, a quantitative assessment of leakage (i.e., leakage amount) can be obtained effectively, conveniently, and non-invasively.

[0037] Furthermore, the leakage volume described in this disclosure can be applied to a wide range of analyses. Taking leakage volume for insulin as an example, its application can include, but is not limited to, analyzing the impact of leakage volume on blood glucose fluctuations, analyzing leakage volume of different injection devices, analyzing leakage volume of different injection doses, and analyzing the effectiveness of injection behavior, at least one of these.

[0038] The adsorbent material 10 disclosed herein may be loaded with an adsorbent. The adsorbent can be any substance capable of adsorbing insulin. Preferably, the adsorbent can be nitrocellulose. In some examples, the nitrocellulose-loaded adsorbent material 10 can be a nitrocellulose membrane (also referred to as an NC membrane). A nitrocellulose membrane is a microporous filter membrane capable of adsorbing proteins, and insulin is a protein hormone. In this case, the nitrocellulose membrane has a strong adsorption capacity for insulin and can effectively adsorb it. In some examples, the nitrocellulose membrane may have a substantially uniform pore size. In this case, insulin can be adsorbed uniformly, thereby reducing the difficulty of subsequent calculation of leakage volume based on leakage area.

[0039] In some examples, a nitrocellulose membrane can be used as a carrier for the C / T line in colloidal gold test paper. In this case, the adsorbent 10 can be colloidal gold test paper, and the leakage image can be specific to the colloidal gold test paper.

[0040] Furthermore, the readily available nitrocellulose membrane further enhances the ease of obtaining leakage data. Through research and verification, the inventors have demonstrated that the application of nitrocellulose membranes to the adsorption of insulin on the human skin surface is safe and non-invasive, and effectively adsorbs leakage generated after insulin injection (i.e., leakage at the target location described later).

[0041] The leakage detection method disclosed herein can quantitatively assess leakage and effectively, conveniently, and non-invasively obtain the leakage volume. In some examples, the leakage detection method disclosed herein can be used to detect the leakage volume during insulin injection in a patient. The following description of the example using a nitrocellulose membrane as the adsorbent material 10 is not intended to limit the scope of this disclosure. Figure 2 This is an exemplary flowchart illustrating a leakage detection method as described in this disclosure.

[0042] like Figure 2 As shown, in some examples, the leakage detection method may include acquiring a leakage image (step S102), performing contrast enhancement on the leakage image to acquire an enhanced image (step S104), classifying the pixels in the enhanced image to determine the category of the pixels (step S106), determining the target category of pixels based on the category of the pixels (step S108), performing convex hull fitting based on the target category of pixels to determine the leakage region (step S110), determining the leakage area based on the leakage region (step S112), and determining the leakage amount based on the leakage area (step S114).

[0043] refer to Figure 2 In this embodiment, in step S102, a leakage image can be acquired.

[0044] In some examples, the leak image can be a color image or a grayscale image. Preferably, the leak image can be a color image. A color leak image can capture more details of the leak.

[0045] In some examples, the leakage image can be acquired by an acquisition device, which can be any device with imaging capabilities. For example, the acquisition device can include, but is not limited to, a mobile phone, computer, or camera. In some examples, the acquisition device can be the computing device 30 described above. Preferably, the acquisition device can be a mobile phone. In this case, because mobile phones are easy to operate and readily accepted by patients or those performing the injection, the convenience of obtaining the leakage volume can be further improved. In some examples, the mobile phone can integrate an application for implementing the leakage detection method according to the examples of this disclosure to acquire leakage images and obtain the leakage volume based on the leakage images. In some examples, the application integrated on the mobile phone can also be used to determine the impact of the leakage volume on the patient's blood glucose fluctuations.

[0046] In this embodiment, the leakage image can be an image of the adsorbent material 10 that has absorbed the leakage. In some examples, the leakage image can be an image of the adsorbent material 10 after the patient has injected insulin, showing the leakage absorbed by the adsorbent material 10. Specifically, after the patient has injected insulin, the adsorbent material 10 can be used to absorb the leakage at a target location related to the amount of leakage, and an image of the adsorbent material 10 with the absorbed leakage can be acquired as the leakage image.

[0047] In some examples, after the adsorbent material 10 has continuously adsorbed the liquid for a first preset time to determine that adsorption is complete and leakage has occurred, a leakage image can be acquired within a second preset time. Furthermore, the first preset time can be determined by the adsorption capacity of the adsorbent material 10. The second preset time can be determined by the evaporation time of insulin relative to the adsorbent material 10. In this case, based on stricter time control, leakage can be effectively adsorbed and the probability of leakage evaporation can be reduced, thereby obtaining a more accurate leakage image representing the leakage situation, which is beneficial to improving the accuracy of leakage area calculation.

[0048] In some examples, the first preset time may be a first fixed value (e.g., an empirical value) related to the adsorption capacity of the adsorbent material 10. Alternatively, the first fixed value may be a range. In some examples, the first preset time can be obtained through multiple tests of the adsorbent material 10 adsorbing leakage. In other examples, the first preset time can be estimated based on the adsorption capacity of the adsorbent material 10.

[0049] In some examples, the second preset time may be a second fixed value (e.g., an empirical value) related to the evaporation time of insulin adsorbed by the adsorbent material 10. Alternatively, the second fixed value may be a range. In some examples, the second preset time can be obtained through multiple trials verifying the evaporation time of insulin adsorbed by the adsorbent material 10. In some examples, the second preset time may be less than 30 seconds. In some examples, the requirement for the second preset time can be reduced by improving the stability of insulin adsorption by the adsorbent material 10.

[0050] As described above, an image of the adsorbent material 10 after the leakage can be acquired as a leakage image. In some examples, an image of the adsorbent material 10 after the leakage and a reference object 20 can be acquired as a leakage image. That is, the leakage image can simultaneously include the leakage pattern (i.e., the pattern for the leakage area) and the pattern of the reference object 20 (which can also be simply referred to as the reference object pattern). In this case, the leakage area can be easily determined based on the reference object 20. In addition, since the reference object 20 is present, the consistency requirements for acquisition conditions (e.g., the resolution, focal length, and shooting distance of the acquisition device) are lower, which improves the convenience of acquiring leakage images.

[0051] In some examples, the size of the reference object 20 may be known. That is, the reference object 20 can be any object of known size. For example, the reference object 20 may include, but is not limited to, a coin and a ruler. Preferably, the reference object 20 can be a ruler. In this case, the area of ​​a pixel (i.e., the area of ​​a single pixel) can be conveniently determined by the scale on the ruler.

[0052] In some examples, during data acquisition, images of the adsorbent material 10 after leakage and a reference object 20 of known size can be taken to obtain images of the leakage. In some examples, the adsorbent material 10 and the reference object 20 can be located on the same plane during image acquisition. In this case, the distances of the adsorbent material 10 and the reference object 20 from the acquisition device are approximately the same, which reduces the difficulty of subsequently calculating the leakage area.

[0053] Figure 3 This is an exemplary flowchart illustrating the acquisition of leakage images based on reference 20, as described in this disclosure example. Figure 4 This is a schematic diagram showing a leakage image using reference object 20 as described in the example of this disclosure.

[0054] Specifically, Figure 3 The process of acquiring leakage images based on reference object 20 is shown, which may include the following steps:

[0055] Step S202: The patient injects insulin.

[0056] Step S204: After insulin injection, the absorbent material 10 can be used to collect any leakage at the target site. For example, the absorbent material 10 can be attached to the target site to collect the leakage.

[0057] Step S206: The collected leaked absorbent material 10 can be placed near a reference object 20 of known size, located on the same plane as the absorbent material 10. The vicinity of the reference object 20 can be, for example, above, below, to the left, or to the right of the reference object 20.

[0058] Step S208: Images of the collected leaked adsorbent material 10 and reference object 20 can be taken to obtain leak images. As an example, Figure 4 This is a schematic diagram showing a leakage image with reference object 20 as a scale, adsorbent material 10 as a nitrocellulose membrane, and the nitrocellulose membrane placed above the scale. Figure 4 The diagram shows a nitrocellulose membrane pattern 11 representing a nitrocellulose membrane, a leakage pattern 111 within the nitrocellulose membrane pattern 11, a scale pattern 21 representing a scale, and a background cloth pattern 50 representing a background cloth. The background cloth can be used to suppress noise from the shooting environment, which is beneficial for distinguishing the areas corresponding to the adsorbent material 10 and the reference object 20 from the leakage image.

[0059] In some examples, there may be a gap between the leakage pattern and the reference pattern in the leakage image. In this case, the mutual influence between the leakage pattern and the reference pattern can be reduced, and it is beneficial to obtain images of the corresponding areas for separate processing.

[0060] However, the examples disclosed herein are not limited to this. In other examples, reference object 20 may not be used (that is, the reference object pattern may not be included in the leakage image), which will be described in detail later in the section on determining the leakage area.

[0061] As mentioned above, the absorbent material 10 can be used to absorb the leakage at the target location. In some examples, the target location can be any one or more locations associated with the leakage. In some examples, the target location can include the insulin outlet of an insulin injection device and the skin surface around the patient's injection site. In this case, the leakage at the target location can substantially encompass the medication that was not injected into the patient's body, thus enabling effective analysis or evaluation based on a more comprehensive leakage volume in some applications (e.g., the impact on the patient's blood glucose fluctuations and / or assessing the effectiveness of the injection).

[0062] In some examples, the leakage corresponding to different locations within the target location can be distributed across different locations on the adsorbent material 10. In this case, the leakage situation at different locations can be distinguished in a single leak image acquisition, thereby enabling the simultaneous acquisition of the leakage volume at each location and the total leakage volume at multiple locations.

[0063] In other examples, the target location can be either the insulin outlet of the injection device or a location on the skin surface surrounding the patient's injection site. This allows for the acquisition of leakage volume at each location. The specific target location can be selected based on the intended use of the leakage measurement.

[0064] In some examples, the injection device may include at least one of a needle-free injection device (also known as a jet injection device) and a needle-based injection device. In this case, the leakage rate corresponding to different injection devices can be obtained, thus helping the patient choose the appropriate device. A needle-free injection device can utilize a specific device to generate instantaneous high pressure, which propels the medication through a very fine nozzle onto the skin surface, forming micropores and allowing the medication to diffuse directly into the subcutaneous tissue. Furthermore, for needle-free injection devices, the insulin outlet can be a nozzle. For needle-based injection devices, the insulin outlet can be a needle tip.

[0065] Furthermore, the color of the adsorbent material 10 can be any color that creates a color difference between the areas where insulin leakage occurs and the areas where there is no leakage after insulin adsorption. Preferably, the adsorbent material 10 can be a white nitrocellulose membrane. In this case, after the white nitrocellulose membrane adsorbs insulin, its appearance will darken, which is beneficial for distinguishing leakage areas in subsequent image analysis.

[0066] Furthermore, for the adsorbent material 10, which creates a color difference after adsorbing insulin, a color difference can exist between the non-leaking area (which can be simply referred to as the background area) and the leaking area (i.e., the leaking area) in the leakage image. That is, the grayscale values ​​of pixels in the background area and the leaking area in the leakage image can belong to different grayscale ranges. In this case, distinguishing the leaking area using the color change after the adsorbent material 10 adsorbs insulin is more convenient than methods that require reagent reactions to generate color intensity.

[0067] Figure 5 This is an exemplary flowchart illustrating the contrast enhancement involved in the examples of this disclosure. Figure 6A This shows a histogram of the gray values ​​of the lightness component involved in the example of this disclosure before histogram equalization. Figure 6B This is a histogram showing the gray values ​​of the lightness component involved in the example of this disclosure after histogram equalization.

[0068] Return to reference Figure 2 In this embodiment, in step S104, the leakage image can be contrast-enhanced to obtain an enhanced image. This improves the contrast between the leakage area and other areas in the enhanced image.

[0069] In some examples, contrast enhancement can be performed on the leakage image based on histogram equalization to obtain an enhanced image. In some examples, contrast enhancement can be performed on the grayscale values ​​of a preset component in the leakage image based on histogram equalization to obtain an enhanced image. In some examples, the preset component can be a luminance component (also known as a brightness component).

[0070] Taking contrast enhancement based on grayscale values ​​of the luminance component as an example, such as Figure 5 As shown, it may include:

[0071] Step S302 can convert the color space of the leakage image to a target color space with a lightness component.

[0072] In some examples, the target color space may also include hue and saturation components. For instance, the target color space can be an HSV (Hue, Saturation, Value) space or an HSI (Hue, Saturation, Intensity) space. Thus, enhanced images can be generated using either the HSV or HSI color space.

[0073] Step S304 can perform histogram equalization on the grayscale values ​​of the lightness component in the target color space.

[0074] In some examples, for the gray values ​​of the brightness component, the cumulative distribution function corresponding to histogram equalization can satisfy the formula:

[0075]

[0076] Where k can represent the gray value of the lightness component, and n k S can represent the number of pixels with a gray value of k in the brightness component of the leakage image, and N can represent the total number of pixels in the leakage image. k It can be represented as the gray value of the lightness component of k after histogram equalization. In some examples, for an 8-bit grayscale image, k∈[0,255].

[0077] Step S306: The gray values ​​of the brightness component can be converted back to the RGB (Red, Green, Blue) space after histogram equalization of the target color space to obtain an enhanced image.

[0078] As an example, Figure 6A and Figure 6B The histograms of the lightness component's grayscale values ​​before and after histogram equalization are shown. It can be seen that before histogram equalization, Figure 6A Two peaks are clearly visible (corresponding to the background area and the leaking area), and the grayscale distribution is within a narrow range (meaning the contrast of the leaking image is not high). After histogram equalization, Figure 6B The more uniform grayscale distribution increases the range of grayscale value differences between pixels, thereby enhancing the contrast in the image.

[0079] In some examples, for leak images that contain other patterns (e.g., reference patterns or background fabric patterns), an image of the leaking area within the leak image can be contrast-enhanced to obtain an enhanced image. For example, an image of the leaking area can be cropped from the leak image and its contrast enhanced to obtain an enhanced image.

[0080] Return to reference Figure 2 In this embodiment, in step S106, the pixels in the enhanced image can be classified to determine the category of the pixels.

[0081] In some examples, the pixel category can be the category of different objects in the augmented image. In some examples, the pixels in the augmented image can be binary classified to obtain pixels of two categories.

[0082] In some examples, the binary classification method described above can be K-means clustering. Specifically, in K-means clustering, K can be set to 2, the pixels in the enhanced image can be considered as samples, and the grayscale values ​​of the three channels (or three spaces) in the RGB space of each sample can be used as features. Unsupervised clustering is then performed on the samples to obtain pixels of two categories. Thus, it is possible to obtain pixels of two categories based on K-means clustering.

[0083] In some examples, the objective function J of K-means clustering can be expressed as:

[0084]

[0085] Where M can represent the number of samples, K can represent the number of classes, and μ k x can represent the cluster center of the k-th category. i J can represent the feature value of the i-th sample. In some examples, if the difference between two consecutive J values ​​is less than a preset value (e.g., 0.0001), clustering can be considered to have stopped and training to be complete (even if clustering is complete).

[0086] In some examples, the enhanced image can be normalized before classification. This reduces the differences in grayscale values ​​between pixels, further improving classification accuracy. Specifically, the grayscale values ​​of each channel of the enhanced image can be normalized using the following formula:

[0087]

[0088]

[0089]

[0090] Where R, G, and B can represent the gray values ​​of the three channels of the enhanced image, R', G', and B' can represent the gray values ​​of the three channels of the normalized enhanced image, max can represent the maximum value function, and min can represent the minimum value function.

[0091] Continue to refer to Figure 2 In this embodiment, in step S108, the target category of pixels can be determined based on the category of the pixels.

[0092] As shown above, pixels in an enhanced image can be binary classified to obtain two categories of pixels. In some examples, pixels of the target category can be selected from the two categories.

[0093] In some examples, pixels of a target category can be selected from two categories of pixels based on a first grayscale range before insulin adsorption by the adsorbent material 10 and a second grayscale range after insulin adsorption by the adsorbent material 10. This allows the determination of the target category of pixels based on the grayscale changes before and after insulin adsorption by the adsorbent material 10. Specifically, if the first grayscale range is larger than the second grayscale range, the category with the smaller average grayscale value of the pixels in both categories can be selected as the target category; otherwise, the category with the larger average grayscale value of the pixels in both categories can be selected as the target category. This allows the determination of the target category and the pixels within that target category.

[0094] Taking a white nitrocellulose membrane as an example, as mentioned above, after the white nitrocellulose membrane adsorbs insulin, its appearance darkens, and it is displayed with a lower grayscale value in the enhanced image. Therefore, areas with lower grayscale values ​​in the enhanced image can be leakage areas. For a white nitrocellulose membrane, the first grayscale range can be larger than the second grayscale range. Thus, it is possible to determine the target category of pixels based on the grayscale changes before and after insulin adsorption on the white nitrocellulose membrane.

[0095] Specifically, pixels in the enhanced image can be binary classified to obtain two categories of pixels. The category with the smaller gray-scale mean is designated as the target category, and the other as the background category. In this case, the target category can be determined based on the color change after insulin is adsorbed by the nitrocellulose membrane, thus enabling a more accurate identification of the leakage area.

[0096] In some examples, the leakage area can be determined based on the region corresponding to the pixel of the target category. In some examples, the region corresponding to the pixel of the target category can be directly used as the leakage area. In some examples, further processing can be performed based on the region corresponding to the pixel of the target category to obtain the leakage area (see step S110 described later).

[0097] Figure 7 This is an exemplary flowchart illustrating the convex hull fitting process involved in the examples of this disclosure.

[0098] Continue to refer to Figure 2 In this embodiment, in step S110, convex hull fitting can be performed based on the pixels of the target category to determine the leakage area.

[0099] Specifically, convex hull fitting can be used to adjust the regions corresponding to pixels of the target category. This reduces the influence of isolated pixels, further improving the accuracy of calculating the leakage area. Additionally, it makes the leakage area as regular as possible, which is beneficial for observing the leakage situation.

[0100] In some examples, such as Figure 7As shown, the convex hull fitting process may include:

[0101] Step S402: Connectivity analysis can be performed on the regions corresponding to the pixels of the above target categories to determine at least one connected component.

[0102] Furthermore, each connectivity component can include multiple interconnected pixels. In some examples, connectivity analysis can be eight-neighbor connectivity analysis. Specifically, within the region corresponding to a pixel of the target category, the connectivity between that pixel and its eight neighboring pixels can be analyzed to identify multiple interconnected pixels. In other examples, connectivity analysis can also be four-neighbor connectivity analysis.

[0103] Step S404: The convex hull of each connection component in at least one connection component can be obtained.

[0104] In some examples, the convex hull can be the smallest convex polygon that encloses each connected component. Specifically, the smallest convex polygon that encloses each connected component can be computed as the convex hull of each connected component.

[0105] Step S406: At least one convex hull corresponding to a connecting component can be used as a leakage area.

[0106] In some examples, Laplacian filtering can be applied to the regions corresponding to pixels of the target class before convex hull fitting to adjust the edges of these regions. This reduces the computational cost of subsequent convex hull fitting. In some examples, the Laplacian filtering can satisfy the formula:

[0107]

[0108] Where U can represent the image of the region corresponding to the pixels of the target category. It can represent the result obtained. It can represent convolution operations.

[0109] The process of obtaining the leak region has been described above. However, the examples disclosed herein are not limited to this; in other examples, the leak image can be segmented to obtain the leak region. For example, an image segmentation algorithm can be used to segment the leak image to extract the leak region. Image segmentation algorithms can include, but are not limited to, active contouring, Grabcut, or thresholding methods.

[0110] Return to reference Figure 2 In this embodiment, in step S112, the leakage area (that is, the actual area of ​​insulin adsorbed by the adsorbent material 10) can be determined based on the leakage area.

[0111] In some examples, the leakage area can be determined based on the resolution of the leakage image and the leakage region. The resolution of the leakage image can be used to determine the number of pixels in the corresponding region of the leakage image.

[0112] As described above, images of the adsorbent material 10 and the reference object 20 after the liquid has leaked can be acquired as leakage images. In some examples, the leakage area can be determined based on the resolution of the leakage image, the leakage region, and the reference object pattern in the leakage image.

[0113] Specifically, for a reference object 20 with known dimensions, the number of pixels in the leakage region can be determined based on the resolution of the leakage image, the area of ​​each pixel (i.e., the area of ​​a single pixel) can be determined based on the pattern of the reference object in the leakage image, and the leakage area can be determined based on the number and area of ​​each pixel in the leakage region. Thus, the leakage area can be determined based on the reference object 20.

[0114] Taking reference object 20 as a ruler as an example, since the ruler has scales, edge detection can be performed on the portion of the leakage image including the ruler pattern using image processing algorithms to obtain areas with clear scales on the ruler pattern. This allows us to obtain the scale range corresponding to that area. Based on the number of pixels within and between this scale range, we can obtain the side length of each pixel. Based on the side length of each pixel, we can obtain its area. Therefore, based on the number of pixels and the area of ​​each pixel in the leakage area, we can determine the leakage area. In some examples, the side length of each pixel can be obtained by dividing the scale range by the number of pixels within that scale range.

[0115] As mentioned above, in other examples, reference 20 may not be used. For example, acquisition conditions can be fixed to ensure that the pixel area of ​​each pixel in all leakage images is roughly fixed, thus enabling the leakage region to be converted into a leakage area based on the known pixel area and number of pixels. As another example, the pixel interval (the actual size represented by the distance between two pixels in the leakage image) can be obtained from a DICOM (Digital Imaging and Communications in Medicine) format leakage image. Using the pixel interval as the side length of a pixel, the pixel area can be obtained based on the pixel's side length, and then the leakage area can be determined based on the number and area of ​​pixels in the leakage region.

[0116] Continue to refer to Figure 2 In this embodiment, in step S114, the leakage amount can be determined based on the leakage area.

[0117] In some examples, the leakage amount corresponding to the leakage area can be determined based on the adsorption capacity of the adsorbent material 10 for insulin (hereinafter referred to as adsorption capacity) and the leakage area. Generally, the particle size, pore size, and porosity of the adsorbent material 10 can affect the adsorption capacity. In some examples, the adsorption capacity can be a fixed value. In some examples, the adsorption capacity can be obtained experimentally. In some examples, the adsorption capacity can be a known parameter of the adsorbent material 10. In some examples, the adsorption capacity can be expressed using the dose of insulin adsorbed per unit area of ​​the adsorbent material 10.

[0118] In some examples, the adsorption capacity can be adjusted based on the pixel values ​​of the pixels in the leakage area. For instance, a weight can be determined based on the grayscale value of the saturation component, which can then be used to adjust the adsorption capacity. In this case, adjusting the adsorption capacity by incorporating the pixel values ​​of the pixels takes into account subtle differences in insulin dosage that may exist in the leakage area, thus enabling a more accurate determination of the leakage volume.

[0119] In some examples, a formula can be established based on adsorption capacity to convert leakage area to leakage volume, thus automatically converting the leakage area to leakage volume. This allows for convenient conversion of leakage area to leakage volume. In other examples, a formula can be established based on adsorption capacity and the pixel values ​​of pixels within the leakage area to automatically convert the leakage area to leakage volume.

[0120] Figure 8 This is an exemplary flowchart illustrating an analysis based on leakage volume as described in this disclosure.

[0121] As described above, the leakage volume examples in this disclosure can be applied to many aspects of analysis. This disclosure also provides some examples of analysis based on leakage volume. Specifically, such as... Figure 8 As shown, the leakage amount can be determined (step S502), and analysis can be performed based on the leakage amount (step S504). In step S502, the leakage amount can be determined using the leakage detection method disclosed in this example.

[0122] In some examples, in step S504, the impact of the leakage amount on the patient's blood glucose fluctuations (also known as blood glucose variability) can be determined. In other examples, the impact of the leakage condition on the patient's blood glucose fluctuations can be determined based on the leakage area or leakage area. Because diabetic patients have a lower ability to control their blood glucose, blood glucose fluctuations often exceed the correct range. During insulin injection, if the dose of insulin injected into the patient's body is insufficient, it can also easily cause unstable blood glucose control and blood glucose fluctuations.

[0123] In some examples, changes in blood glucose levels before and after insulin injection can be analyzed to determine the impact of leakage on blood glucose fluctuations. Furthermore, the inventors considered that as the insulin dosage increases, the absorption rate of the medication at the injection site gradually slows down, thus affecting leakage. In other words, leakage can be correlated with the dosage of insulin injected each time. In some examples, the impact of injection dosage on blood glucose fluctuations can be analyzed comprehensively. This allows for a more complete analysis of the influence on blood glucose fluctuations.

[0124] Specifically, the impact of leakage on a patient's blood glucose fluctuations can be determined based on the injection dose, leakage volume, and the patient's blood glucose levels before and after insulin injection. In this case, it is possible to obtain the effect of leakage volume on blood glucose fluctuations at the corresponding injection dose (i.e., insulin dosage), thereby enabling timely intervention (e.g., adjusting the dose or guiding standardized injection behavior).

[0125] In addition, a patient's blood glucose levels before and after insulin injection can be divided into a first blood glucose level and a second blood glucose level. The first blood glucose level is the blood glucose level before insulin injection, and the second blood glucose level is the blood glucose level after insulin injection. Blood glucose level can also be referred to as blood glucose concentration.

[0126] In some examples, a patient's blood glucose levels (i.e., the first blood glucose level and / or the second blood glucose level) can be determined by a finger-prick blood collection device and / or a continuous glucose monitoring system. For example, a patient can obtain their blood glucose levels over time by wearing a continuous glucose monitoring system. In this case, the patient's blood glucose data can be obtained conveniently and promptly, allowing for better analysis of the impact of leakage on blood glucose fluctuations.

[0127] In some examples, in determining the impact of leakage volume on a patient's blood glucose fluctuations, multiple target data points (i.e., multiple target data points corresponding to multiple insulin injections) can be obtained for each insulin injection. The correlation between changes in leakage volume and the patient's blood glucose fluctuations can then be determined based on these multiple target data points. The target data points for each insulin injection (i.e., each target data point) can include the injection dose, leakage volume, first blood glucose level, and second blood glucose level. In this case, the impact of leakage volume on blood glucose fluctuations can be determined by comparing the leakage volume corresponding to multiple injections with the corresponding blood glucose indicators.

[0128] In some examples, determining the impact of leakage on a patient's blood glucose fluctuations can involve acquiring multiple target data points corresponding to several days of insulin injections. Based on these target data points, multi-day blood glucose levels are obtained, and the impact of leakage on blood glucose fluctuations is analyzed based on daily injection doses, leakage volume, and blood glucose levels. In this case, the impact of leakage volume on multi-day blood glucose fluctuations at a given injection dose can be obtained. That is, the sustained impact of leakage volume on blood glucose fluctuations can be determined. In some examples, continuous monitoring for a preset number of days (e.g., 7 days) can be used to obtain multi-day blood glucose levels.

[0129] In some examples, multi-day blood glucose indicators may include, but are not limited to, standard deviation (SD), mean of daily differences (MODD), mean amplitude of glycemic excursion (MAGE), and time within the target range for blood glucose (TIR).

[0130] In some examples, in step S504, injection procedures can be standardized based on the amount of leakage. In other examples, the effectiveness of the injection procedure can be determined based on the amount of leakage, and a corresponding guidance message can be created to standardize the injection procedure. Specifically, the effectiveness of the injection procedure can be determined based on the amount of leakage, and a first guidance message can be created and output to the recipient performing the injection procedure to standardize the injection procedure. In this case, the risk of patients not injecting sufficient insulin can be reduced, the occurrence of adverse reactions can be reduced, and by timely standardizing the injection procedure, the effectiveness of subsequent injection procedures can be improved, reducing the economic problems caused by insufficient insulin injection, which has important clinical value.

[0131] In some examples, in step S504, the correlation between the leakage volume and the patient's blood glucose fluctuations can be determined based on the leakage volume, thereby creating a corresponding guidance message to regulate the injection behavior. Specifically, the correlation between the leakage volume and the patient's blood glucose fluctuations can be determined based on the leakage volume, thereby creating a second guidance message and outputting it to the object performing the injection behavior (i.e., injecting insulin) to regulate the injection behavior. In this case, it can have practical guidance significance for the blood glucose fluctuations of patients with poor pancreatic function (e.g., patients with type 1 diabetes).

[0132] In some examples, leakage rates of different injection devices can be analyzed. In this case, the analysis can identify injection devices with relatively low leakage rates, thereby guiding patient injection practices. Specifically, the analysis can obtain and compare the leakage rates of different injection devices. In some examples, the leakage rates of different injection devices can be analyzed under the condition of the same injection dose.

[0133] In some examples, leakage can be analyzed for different injection doses. In this case, the relationship between injection dose and leakage can be obtained, thus guiding the patient's injection dosage. This can improve the efficacy of insulin injections. Specifically, the analysis can obtain the leakage corresponding to different injection doses of insulin and compare the leakage corresponding to different injection doses. In some examples, leakage can be analyzed for different injection doses under the condition of the same injection device.

[0134] This disclosure also relates to an electronic device that may include at least one processing circuit. The at least one processing circuit is configured to perform one or more steps of the above-described leakage detection method.

[0135] This disclosure also relates to a computer-readable storage medium that may store at least one instruction, which, when executed by a processor, implements one or more steps in the above-described leakage detection method.

[0136] The leakage detection method, analysis system, device, and medium for insulin disclosed herein involve acquiring a leakage image of an adsorbent material 10 with adsorbed leakage, enhancing the contrast of the leakage image to obtain an enhanced image, classifying the pixels in the enhanced image to obtain two categories of pixels, selecting a target category of pixels from the two categories based on the grayscale changes before and after insulin adsorption, adjusting the region corresponding to the target category of pixels using convex hull fitting to determine the leakage region, determining the leakage area based on the leakage region, and further determining the leakage amount corresponding to the leakage area based on the adsorption capacity of the adsorbent material 10 for adsorbing insulin and the leakage area. In this case, the leakage is effectively collected using the adsorbent material 10, the target category of pixels is determined based on the characteristics of the grayscale changes before and after insulin adsorption by the adsorbent material 10, and the region corresponding to the target category of pixels is adjusted using convex hull fitting to intelligently and conveniently determine a regular and accurate leakage region, and the leakage situation is quantitatively determined based on the leakage region and the adsorption capacity of the adsorbent material 10. Thus, quantitative leakage information can be obtained effectively, conveniently, and non-invasively.

[0137] Furthermore, in some of the embodiments disclosed herein, using nitrocellulose membranes as a tool to measure the leakage of insulin injected by the injection device is beneficial for establishing a standardized insulin leakage assessment process, laying a solid foundation for clinical application, and enhancing the benefits of good blood glucose management for patients.

[0138] While the present invention has been specifically described above in conjunction with the accompanying drawings and embodiments, it is to be understood that the above description does not limit the present invention in any way. Those skilled in the art can make modifications and variations to the present invention as needed without departing from the essential spirit and scope of the invention, and all such modifications and variations fall within the scope of the present invention.

Claims

1. A method for detecting insulin leakage, characterized in that, A leakage detection method for detecting leakage during insulin injection includes: acquiring a leakage image, wherein the leakage image is obtained after the patient injects insulin, by using an adsorbent material to adsorb leakage at a target location related to the leakage amount to collect the leakage amount and acquiring an image of the adsorbent material with the adsorbed leakage, the adsorbent material having the adsorption capacity for insulin; enhancing the contrast of the leakage image to obtain an enhanced image; performing binary classification based on pixels in the enhanced image to obtain two categories of pixels; selecting a target category of pixels from the two categories of pixels based on a first grayscale range for the adsorbent material before insulin adsorption and a second grayscale range for the adsorbent material after insulin adsorption; performing connectivity analysis on the region corresponding to the target category of pixels to determine at least one connecting component, obtaining the convex hull of each connecting component in the at least one connecting component, and using the convex hull corresponding to the at least one connecting component as the leakage region; determining the leakage area based on the resolution of the leakage image and the leakage region; and determining the leakage amount corresponding to the leakage area based on the adsorption capacity and the leakage area.

2. The leakage detection method according to claim 1, characterized in that: In selecting the target category of pixels from the two categories of pixels, if the first grayscale range is greater than the second grayscale range, then the category with the smaller average grayscale value of the pixels in the two categories is taken as the target category; otherwise, the category with the larger average grayscale value of the pixels in the two categories is taken as the target category.

3. The leakage detection method according to claim 2, characterized in that: The adsorbent material is a nitrocellulose membrane, which is white in color, and the first grayscale range is larger than the second grayscale range.

4. The leakage detection method according to claim 1, characterized in that: The connectivity analysis is an eight-neighbor connectivity analysis, and the convex hull is the smallest convex polygon corresponding to each connection component.

5. The leakage detection method according to claim 1, characterized in that: In the contrast enhancement, the color space of the leak image is converted to a target color space with a lightness component. Histogram equalization is performed on the grayscale values ​​of the lightness component, and the grayscale values ​​of the lightness component are converted back to the RGB space via the histogram-equalized target color space to obtain the enhanced image. The cumulative distribution function corresponding to the histogram equalization satisfies the formula: Where k represents the grayscale value of the brightness component, and n k S represents the number of pixels in the leakage image with a grayscale value of k for the brightness component, N represents the total number of pixels in the leakage image, and S represents the number of pixels in the leakage image. k The gray value of the brightness component, denoted as k, is the gray value after histogram equalization.

6. The leakage detection method according to claim 1, characterized in that: The binary classification is K-means clustering. In K-means clustering, K is set to 2, the pixels in the enhanced image are samples, and the gray values ​​of the three channels in the RGB space of each sample are features. Unsupervised clustering is performed on the samples to obtain the two categories.

7. The leakage detection method according to claim 1, characterized in that: The leakage image includes a pattern of a reference object of known size. The number of pixels in the leakage region is determined based on the resolution of the leakage image. The area of ​​each pixel is determined based on the pattern of the reference object in the leakage image. The leakage area is determined based on the number of pixels and the area of ​​each pixel.

8. The leakage detection method according to claim 1, characterized in that: The target locations include the insulin outlet of the injection device used to inject insulin and the skin surface surrounding the patient's injection site.

9. An electronic device, characterized in that, It includes at least one processing circuit, which is configured to perform the leakage detection method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction, which, when executed by a processor, implements the leakage detection method as described in any one of claims 1 to 8.

Images