A method of determining airway obstruction and related apparatus

Abstract

Embodiments of the present application disclose a method for judging airway obstruction and related devices, which are used to improve the accuracy of judgment. The method comprises: acquiring a to-be-processed image of a diseased part; performing image segmentation processing on the to-be-processed image according to pixel features of pixels in the to-be-processed image based on an image segmentation algorithm to obtain a segmentation mask image including an obstruction region and an airway region; extracting shape regularity of the obstruction region, adhesion degree between the obstruction region and the airway region, gray mean value difference between the obstruction region and the airway region, and gray contrast of the obstruction region from the segmentation mask image; acquiring a trained judgment model, wherein the judgment model includes a foreign matter judgment module; inputting the shape regularity, the adhesion degree, the gray mean value difference and the gray contrast into the foreign matter judgment module to obtain a corresponding foreign matter judgment result, and the foreign matter judgment result is foreign matter obstruction or pathological obstruction.

Description

Technical Field

[0001] This application relates to the field of emergency medical care, and in particular to a method and device for determining airway obstruction. Background Technology

[0002] Airway obstruction refers to a pathological state in which the airway narrows or blocks due to various factors, resulting in a reduction in the airway's cross-sectional area and impaired ventilation function. Mild cases present with increased airflow resistance or difficulty breathing, while severe cases can lead to hypoxia or asphyxiation due to ventilation interruption. It is one of the most life-threatening conditions in clinical emergency care. Common causes include foreign body aspiration (such as food debris, small metal objects, etc.) and pathological factors (such as laryngeal edema, tumor compression, etc.). In current protocols, when a patient experiences airway obstruction, medical staff will determine the specific type of airway obstruction—such as foreign body obstruction or pathological obstruction—within the optimal emergency timeframe (usually five minutes) based on clinical experience, and then perform standardized procedures, such as the Heimlich maneuver.

[0003] However, in the existing approach, emergency treatment relies on the clinical experience of medical staff. If the medical staff lacks experience, it will lead to misjudgment of the specific type of airway obstruction. In addition, the optimal time for emergency treatment is short, and medical staff cannot make an accurate judgment in a short period of time. All of these factors result in low accuracy in the judgment of airway obstruction, which in turn leads to a low success rate of emergency treatment. Summary of the Invention

[0004] To address the technical problem of low accuracy in the prior art, this application provides a method and related apparatus for determining airway obstruction, which improves the accuracy of the determination.

[0005] The first aspect of this application provides a method for determining airway obstruction, characterized in that it includes: Acquire the image of the affected area to be processed; Based on the image segmentation algorithm, the image to be processed is segmented according to the pixel features of the pixels in the image to be processed, and a segmentation mask image including the obstruction region and the airway region is obtained. Extract the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the gray-scale mean difference between the obstruction region and the airway region, and the gray-scale contrast of the obstruction region from the segmentation mask image. Obtain a trained judgment model, the judgment model including a foreign object judgment module; The shape regularity, adhesion, grayscale mean difference, and grayscale contrast are input into the foreign body judgment module to obtain the corresponding foreign body judgment result, wherein the foreign body judgment result is foreign body obstruction or pathological obstruction.

[0006] Optionally, extracting the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the difference in grayscale mean between the obstruction region and the airway region, and the grayscale contrast of the obstruction region from the segmented mask image includes: Based on the segmentation mask image, calculate the area and perimeter of the obstruction region in the segmentation mask image, and substitute the area and perimeter into a preset roundness formula to calculate the roundness of the obstruction region. The curvature of the contour points of the obstruction region in the segmentation mask is calculated based on the segmentation mask image, and the edge smoothness of the obstruction region is calculated based on the obstruction curvature variation coefficient and the preset smoothness coefficient according to the preset smoothness formula. The obstruction curvature variation coefficient is the ratio of the standard deviation to the mean of the curvature of all contour points in the obstruction region. The shape regularity of the obstruction region is obtained by weighted fusion calculation based on the circularity and the edge smoothness. In the segmentation mask image, the area ratio of the contact area to the airway area is determined as the adhesion degree between the obstruction area and the airway area, wherein the contact area is the area where the neighboring pixels of each pixel in the obstruction area are pixels of the airway area. The mean grayscale difference between the obstruction region and the airway region, and the grayscale contrast of the obstruction region are obtained from the segmentation mask image.

[0007] Optionally, the image segmentation algorithm performs image segmentation processing on the image to be processed based on the pixel features of pixels in the image to be processed, to obtain a segmentation mask image including the obstruction region and the airway region, including: Pixel features of each pixel are extracted from the image to be processed, wherein the pixel features include grayscale features, color features, texture features and semantic features; The grayscale feature, color feature, texture feature, and semantic feature corresponding to each pixel are concatenated and fused to obtain the fused feature; For each pixel, the weighted sum of the fusion features and the preset fusion weights is determined as the probability vector of the pixel, and the category label corresponding to the dimension with the highest probability in the probability vector is determined as the target category label of the pixel, thereby obtaining the segmentation mask map corresponding to the image to be processed, including the obstruction region and the airway region.

[0008] Optionally, the judgment model further includes a type judgment module. After inputting the shape regularity, adhesion, grayscale mean difference, and grayscale contrast into the foreign object judgment module to obtain the corresponding foreign object judgment result, the method further includes: The gray values ​​of all pixels in the obstruction region are obtained according to the segmentation mask image, and the standard deviation of the gray values ​​calculated based on the gray values ​​of all pixels in the obstruction region is determined as the density uniformity. The grayscale values ​​of all pixels in the obstruction region are obtained according to the segmentation mask image, and the average grayscale value calculated based on the grayscale values ​​of all pixels in the obstruction region is determined as the transmittance. The area of ​​the obstruction region and the cross-sectional area of ​​the airway are determined based on the segmentation mask image, and the ratio of the area of ​​the obstruction region to the cross-sectional area of ​​the airway is determined as the airway mucosal redness rate. The minimum effective ventilation diameter of the airway stenosis segment and the airway diameter are determined based on the segmented mask diagram, and the ratio of the minimum effective ventilation diameter of the airway stenosis segment to the airway diameter is determined as the stenosis rate. The minimum diameter of all cross sections within the airway stenosis is determined based on the segmentation mask image, and the ratio of the standard deviation to the mean of the minimum diameter of all cross sections within the airway stenosis is determined as the diameter variation coefficient; the stenosis uniformity is obtained based on the diameter variation coefficient, wherein the diameter variation coefficient is inversely proportional to the stenosis uniformity. The curvature of the narrow segment contour points of the airway region in the segmentation mask is calculated based on the segmentation mask, and the ratio of the standard deviation to the mean of the curvature of the narrow segment contour points of the airway region is determined as the narrow segment edge regularity. The foreign body judgment result, density uniformity, light transmittance, airway mucosal redness and swelling rate, stenosis rate, stenosis uniformity, and stenosis segment edge regularity are input into the type judgment module to obtain the corresponding type judgment result, which is a hard foreign body, a soft foreign body, laryngeal edema, or tumor compression.

[0009] Optionally, the judgment model further includes a location judgment module, and the segmentation mask image further includes a glottic region. After inputting the foreign body judgment result, the density uniformity, the light transmittance, the airway mucosal redness rate, the stenosis rate, the stenosis uniformity, and the stenosis segment edge regularity into the type judgment module to obtain the corresponding type judgment result, the method further includes: Based on the calibration relationship between pixel coordinates and actual size, the target distance between the obstruction region and the glottic region is calculated according to the segmentation mask image; If the type determination result is a hard foreign object, the target distance is input to the position determination module to obtain the corresponding position determination result, which includes high-level obstruction, mid-level obstruction, or low-level obstruction.

[0010] Optionally, after acquiring the image to be processed targeting the affected area, and before performing image segmentation processing on the image to be processed based on the pixel features of the pixels in the image to be processed using an image segmentation algorithm to obtain a segmentation mask image including the obstruction region and the airway region, the method further includes: The image to be processed is subjected to noise removal based on the Gaussian filtering algorithm, and the image after noise removal is subjected to contrast enhancement based on the histogram equalization algorithm to obtain the first image. The first image is processed by an edge detection algorithm to obtain a second image that displays the contour boundaries of different objects. The image segmentation algorithm performs image segmentation processing on the image to be processed based on the pixel features of the pixels in the image to be processed, and obtains a segmentation mask image including the obstruction region and the airway region, including: Based on the image segmentation algorithm, the image to be processed is segmented according to the pixel features of the pixels in the second image to obtain a segmentation mask image including the obstruction region and the airway region.

[0011] Optionally, before obtaining the trained judgment model, the method further includes: Acquire a set of labeled training images, a set of labeled verification images, and a set of labeled test images according to a preset ratio; The training image set, the verification image set, and the test image set to be processed are respectively subjected to denoising, contrast enhancement, and edge detection processing to obtain the target training image set, the target verification image set, and the target test image set; The semantic segmentation-based image segmentation algorithm performs image segmentation processing on the target training image set, the target verification image set, and the target test image set respectively, and then performs feature extraction to obtain the target features corresponding to each of the target training image set, the target verification image set, and the target test image set. The target features include the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the gray-level mean difference of the airway region, and the gray-level contrast of the obstruction region. The target features corresponding to the target training image set are input into a preset initial model to iteratively train the weight matrix of the initial model. The target features corresponding to the target verification image set are input into the initial model to adjust the hyperparameters and structure of the initial model. The initial model is then tested based on the target features corresponding to the target test image set until the initial model meets the preset convergence condition and becomes the judgment model.

[0012] A second aspect of this application provides an airway obstruction detection device, characterized in that it includes: The acquisition unit is used to acquire the image to be processed of the affected area. The segmentation unit is used to perform image segmentation processing on the image to be processed based on the pixel features of the pixels in the image to be processed, based on the image segmentation algorithm, to obtain a segmentation mask image including the obstruction region and the airway region. The extraction unit is used to extract from the segmentation mask image the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the gray-level mean difference between the obstruction region and the airway region, and the gray-level contrast of the obstruction region. The acquisition unit is also used to acquire a trained judgment model, the judgment model including a foreign object judgment module; The processing unit is used to input the shape regularity, the adhesion, the grayscale mean difference and the grayscale contrast into the foreign body judgment module to obtain the corresponding foreign body judgment result, wherein the foreign body judgment result is foreign body obstruction or pathological obstruction.

[0013] A third aspect of this application provides an airway obstruction detection device, comprising: Central processing unit, memory, and input / output interfaces; The memory is either a short-term storage memory or a persistent storage memory; The central processing unit is configured to communicate with the memory and execute instructions in the memory to perform the aforementioned method.

[0014] A fourth aspect of this application provides a computer-readable storage medium including instructions that, when executed on a computer, cause the computer to perform the aforementioned method.

[0015] As can be seen from the above technical solutions, the embodiments of this application have the following advantages: First, an image of the affected area is acquired. Then, based on an image segmentation algorithm, the image is segmented according to the pixel features of the pixels in the image, resulting in a segmentation mask image including the obstruction region and the airway region. Next, the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the difference in the mean grayscale values ​​between the two regions, and the grayscale contrast of the obstruction region are extracted from the segmentation mask image. Then, a trained judgment model is acquired. Finally, the shape regularity, adhesion, mean grayscale difference, and grayscale contrast are input into the foreign body judgment module of the judgment model to obtain the corresponding foreign body judgment result. The method of this application is applied to an airway obstruction judgment device. After acquiring an image of the affected area using a laryngoscope or similar means, the device can quickly process the image to obtain a foreign body judgment result (foreign body obstruction or pathological obstruction) corresponding to the image. Because it utilizes a judgment model trained with a large number of examples, the device can accurately determine whether it is a foreign body obstruction or a pathological obstruction. Compared with manual judgment, it can quickly provide results in a short time with high accuracy, greatly improving the success rate of emergency treatment.

[0016] Additional aspects and advantages of embodiments of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of embodiments of this application. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application 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 this application. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of an embodiment of the airway obstruction determination method disclosed in this application; Figure 2 This is a schematic diagram of another embodiment of the airway obstruction determination method disclosed in this application; Figure 3 This is a schematic diagram of an embodiment of an airway obstruction detection device disclosed in this application; Figure 4 This is a schematic diagram of another embodiment of the airway obstruction detection device disclosed in this application.

[0019] The realization of the objectives, functional features and advantages of the embodiments of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] Airway obstruction is a life-threatening condition in clinical emergency departments, with foreign body obstruction and pathological obstruction being the most common types. In existing protocols, when a patient experiences airway obstruction, medical staff rely on clinical experience to determine the specific type of obstruction (e.g., foreign body obstruction or pathological obstruction) within the optimal emergency timeframe, and then perform standardized procedures, such as the Heimlich maneuver. However, existing protocols rely heavily on the clinical experience of medical staff. Insufficient experience can lead to misdiagnosis of the specific type of airway obstruction. Furthermore, the short optimal emergency timeframe makes it difficult for medical staff to accurately determine whether it is a foreign body obstruction or a pathological obstruction within a short period. These factors contribute to low accuracy in airway obstruction diagnosis, resulting in a low success rate of emergency treatment. For example, inexperienced medical staff may misdiagnose laryngeal edema (a pathological obstruction) as a foreign body obstruction, blindly applying the Heimlich maneuver and worsening mucosal damage, or fail to diagnose obstruction caused by tumor compression, delaying targeted treatment. To address the aforementioned issues, this application provides a method and related device for determining airway obstruction, which can rapidly obtain the corresponding foreign body assessment result within a short period of time. That is, it can accurately determine whether the obstruction is due to a foreign body or a pathological obstruction. Compared with manual assessment, it can provide results quickly and with high accuracy, greatly improving the success rate of emergency treatment.

[0021] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the embodiments of this application, and should not be construed as limiting the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of the embodiments of this application.

[0022] In the description of the embodiments of this application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "circumferential", "radial", etc., indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of this application.

[0023] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "first," "second," "third," "fourth," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. In the description of embodiments of this application, "a plurality of" means two or more, unless otherwise expressly specified.

[0024] In the embodiments of this application, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0025] In the embodiments of this application, unless otherwise expressly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature being directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature being directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0026] The following describes a method for diagnosing airway obstruction according to this application. Please refer to... Figure 1 An embodiment of the airway obstruction determination method of this application includes: 101. Obtain the image to be processed for the affected area; The device acquires images of the affected area. Specifically, it is equipped with a visualization laryngoscope module, which directly acquires images of the affected area. The camera of the visualization laryngoscope module has a resolution of at least 1080P and a lens diameter of ≤5mm to adapt to narrow airway environments. The lens can also have a built-in supplementary lighting module, which can automatically turn on supplementary lighting in low-light environments (such as ambulances and emergency rooms in primary hospitals) to ensure that the image clarity is at least 720P. Since the visualization laryngoscope module has a flexible and adjustable function, it can adapt well to the environment of the affected area.

[0027] 102. Based on the image segmentation algorithm, the image to be processed is segmented according to the pixel features of the pixels in the image to be processed, and a segmentation mask image including the obstruction region and the airway region is obtained. Based on image segmentation algorithms, the image to be processed is segmented according to the pixel features of the pixels in the image to be processed, resulting in a segmentation mask map including the obstruction region and the airway region. Specifically, image segmentation algorithms based on semantic segmentation can be used for processing, and pixel features can be grayscale features, color features, texture features, etc. The segmentation mask map has at least two semantically labeled regions, namely the obstruction region and the airway region.

[0028] 103. Extract the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the difference in the mean gray value between the obstruction region and the airway region, and the gray-level contrast of the obstruction region from the segmentation mask image. The shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the difference in grayscale mean between the obstruction region and the airway region, and the grayscale contrast of the obstruction region are extracted from the segmentation mask image. Specifically, the shape regularity of the obstruction region indicates the regularity of the obstruction; the adhesion indicates the viscosity of the obstruction; a higher grayscale mean difference indicates a greater probability of foreign body obstruction; and a higher grayscale contrast also indicates a greater probability of foreign body obstruction.

[0029] 104. Obtain the trained judgment model, which includes a foreign object judgment module; Obtain the trained judgment model. The judgment model includes at least a foreign object judgment module, whose input is features and output is the judgment result.

[0030] 105. Input the shape regularity, adhesion, grayscale mean difference, and grayscale contrast into the foreign object detection module to obtain the corresponding foreign object detection results.

[0031] The shape regularity, adhesion, gray-level mean difference, and gray-level contrast are input into the foreign body detection module to obtain the corresponding foreign body detection result. The foreign body detection result is either foreign body obstruction or pathological obstruction. Specifically, the weight matrix of the foreign body detection module itself is used to calculate the corresponding foreign body detection result with the shape regularity, adhesion, gray-level mean difference, and gray-level contrast, thus determining whether it is foreign body obstruction or pathological obstruction.

[0032] In this embodiment, an image of the affected area is first acquired. Then, based on an image segmentation algorithm, the image is segmented according to the pixel features of the pixels in the image to be processed, resulting in a segmentation mask image including the obstruction region and the airway region. Next, the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the gray-level mean difference between the obstruction region and the airway region, and the gray-level contrast of the obstruction region are extracted from the segmentation mask image. Then, a trained judgment model is acquired. Finally, the shape regularity, adhesion, gray-level mean difference, and gray-level contrast are input into the foreign body judgment module of the judgment model to obtain the corresponding foreign body judgment result. The method of this application is applied to an airway obstruction detection device. After obtaining an image of the affected area through a laryngoscope or other means, the device can quickly process the image to obtain a foreign body detection result (foreign body obstruction or pathological obstruction) corresponding to the image. Because it uses a judgment model trained with a large number of examples, the device can accurately determine whether it is a foreign body obstruction or a pathological obstruction. Compared with manual judgment, it can quickly provide results in a short time with high accuracy, which greatly improves the success rate of emergency treatment.

[0033] Please see Figure 2 Another embodiment of the airway obstruction determination method of this application includes: 201. Obtain the image to be processed for the affected area; The device acquires images of the affected area. Specifically, it is equipped with a visualization laryngoscope module, which directly acquires images of the affected area. The camera of the visualization laryngoscope module has a resolution of at least 1080P and a lens diameter of ≤5mm to adapt to narrow airway environments. The lens can also have a built-in supplementary lighting module, which can automatically turn on supplementary lighting in low-light environments (such as ambulances and emergency rooms in primary hospitals) to ensure that the image clarity is at least 720P. Since the visualization laryngoscope module has a flexible and adjustable function, it can adapt well to the environment of the affected area.

[0034] 202. The image to be processed is denoised using a Gaussian filtering algorithm, and the image after denoising is contrast-enhanced using a histogram equalization algorithm to obtain the first image. The image to be processed is denoised using a Gaussian filtering algorithm, and then contrast-enhanced using a histogram equalization algorithm to obtain the first image. It is understood that other types of filtering algorithms besides Gaussian filtering can be used, depending on the specific needs; no particular limitation is made here. Histogram equalization is also one implementation method; other contrast-enhancing algorithms are also available, and no particular limitation is made here.

[0035] 203. Perform edge detection processing on the first image based on the edge detection algorithm to obtain a second image that displays the contour boundaries of different objects; Edge detection is performed on the first image using an edge detection algorithm to obtain a second image that displays the boundaries of different objects. Specifically, the edge detection algorithm can be the Canny edge detection algorithm or others, and can be replaced according to actual needs; no specific limitation is made here. The edge detection algorithm can highlight the boundary differences between the obstruction area and the airway area, laying the foundation for subsequent feature extraction.

[0036] 204. Based on the image segmentation algorithm, the image to be processed is segmented according to the pixel features of the pixels in the second image to obtain a segmentation mask image including the obstruction region, the airway region and the glottis region. Based on image segmentation algorithms, the image to be processed is segmented according to the pixel features of pixels in the second image, resulting in a segmentation mask image including the obstruction region, airway region, and glottis region. The image segmentation algorithm can be semantic segmentation-based, or other types; no specific limitation is made here. The segmentation mask image contains semantic annotations and boundary delineations, distinguishing the obstruction region, airway region, and glottis region. The following describes the processing procedure of the semantic segmentation-based image segmentation algorithm in detail.

[0037] In one implementation, pixel features for each pixel are first extracted from the image to be processed. These pixel features include grayscale features, color features, texture features, and semantic features. If a convolutional network is used, at least three convolutional layers are employed, each with a different kernel. Shallow convolutional layers extract grayscale and color features, mid-level convolutional layers extract texture features, and deep convolutional layers extract semantic features. The size of the convolutional kernels can be selected according to actual needs and is not limited here.

[0038] Then, the grayscale features, color features, texture features, and semantic features corresponding to each pixel are concatenated and fused to obtain the fused features. In other words, the shallow image, the medium image, and the deep image are fused, and the information of each pixel can be superimposed, resulting in fused features.

[0039] Finally, for each pixel, the weighted sum of the fusion features and preset fusion weights is used to determine the pixel's probability vector. The category label corresponding to the dimension with the highest probability in the probability vector is then determined as the pixel's target category label, thus obtaining a segmentation mask map corresponding to the image to be processed, including the obstruction region and the airway region. For example, if the fusion features are a 27-dimensional vector and the probability vector obtained after calculating the fusion weights is 3-dimensional (0.4, 0.9, 0.78), and assuming the first dimension represents the obstruction region category label, the second dimension represents the airway region category label, and the third dimension represents the glottic region category label, then the pixel corresponding to this vector is labeled with the airway region category label. This process continues until all pixels are labeled, resulting in the segmentation mask map.

[0040] 205. Extract the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the difference in the mean gray level between the obstruction region and the airway region, and the gray level contrast of the obstruction region from the segmentation mask image. Extract the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the difference in the mean grayscale value between the obstruction region and the airway region, and the grayscale contrast of the obstruction region from the segmented mask image. See the detailed steps below.

[0041] a. Based on the segmentation mask, calculate the area and perimeter of the obstructed region in the segmentation mask, and substitute the area and perimeter into a preset roundness formula to calculate the roundness of the obstructed region. Specifically, the dimensions on the segmentation mask can be measured and mapped to the actual dimensions to obtain the actual values.

[0042] The curvature of the contour points of the obstruction region in the segmentation mask is calculated based on the segmentation mask image. The edge smoothness of the obstruction region is calculated based on the obstruction curvature variation coefficient and the preset smoothness coefficient, according to the preset smoothness formula. The obstruction curvature variation coefficient is the ratio of the standard deviation to the mean of the curvature of all contour points in the obstruction region, while the smoothness coefficient can be set according to actual needs.

[0043] The shape regularity of the obstruction region is obtained by weighted fusion calculation based on roundness and edge smoothness. The weight allocation is generally 1:1, but other weights can also be used, depending on the importance attached to the obstruction. The higher the shape regularity, the more likely it is to be a foreign body obstruction; the lower the shape regularity, the more likely it is to be a pathological obstruction.

[0044] b. In the segmentation mask image, the area ratio of the contact area to the airway area is determined as the adhesion degree between the obstruction area and the airway area. The contact area is defined as the region where every pixel in the obstruction area has neighboring pixels in the airway area. Simply put, if a pixel in the obstruction area has eight adjacent pixels in its neighborhood that are all pixels in the airway area, then this pixel is part of the contact area. A higher adhesion degree indicates a larger contact area, and the more likely the obstruction is integrated with the airway, i.e., a more likely pathological obstruction. Conversely, a lower adhesion degree indicates a more likely foreign body obstruction.

[0045] c. Obtain the mean grayscale difference and grayscale contrast of the obstruction region and the airway region from the segmentation mask image. A larger mean grayscale difference indicates a greater difference between the obstruction region and the airway region, suggesting a higher probability of foreign body obstruction. Conversely, a smaller mean grayscale difference indicates a smaller difference between the obstruction region and the airway region, suggesting a higher probability of pathological obstruction. Similarly, a larger mean grayscale difference generally indicates a higher probability of foreign body obstruction, while a smaller mean grayscale contrast generally indicates a higher probability of pathological obstruction.

[0046] 206. Obtain the trained judgment model, which includes a foreign object judgment module, a type judgment module, and a location judgment module; A trained judgment model is obtained, which includes a foreign body judgment module, a type judgment module, and a location judgment module. In one implementation, the foreign body judgment module is the first level, the type judgment module is the second level, and the location judgment module is the third level. The foreign body judgment module determines whether the obstruction is a foreign body, the type judgment module determines the specific type of obstruction (laryngeal edema, tumor compression, hard foreign body, and soft foreign body), and the location judgment module determines the actual distance of the hard foreign body from the glottis.

[0047] The training of the judgment model is described below.

[0048] a. Obtain labeled training images, validation images, and test images according to a preset ratio. The preset ratio can be 7:2:1, or other ratios; no specific limit is specified here. The image sets are pre-labeled by experienced medical personnel.

[0049] b. Perform denoising, contrast enhancement, and edge detection processing on the training image set, verification image set, and test image set respectively to obtain the target training image set, target verification image set, and target test image set. The algorithms involved in this subdivision step can be referred to the relevant steps above, and will not be repeated here.

[0050] c. An image segmentation algorithm based on semantic segmentation is used to segment the target training image set, target verification image set, and target test image set respectively, and then feature extraction is performed to obtain the target features corresponding to each of the target training image set, target verification image set, and target test image set. Among them, the target features include the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the gray-level mean difference of the airway region, the gray-level contrast, density uniformity, light transmittance of the obstruction region, the airway mucosal redness and swelling rate, stenosis rate, stenosis uniformity, stenosis segment edge regularity, and the target distance between the obstruction region and the glottis region.

[0051] d. Input the target features corresponding to the target training image set into the preset initial model to iteratively train the weight matrix of the initial model. Input the target features corresponding to the target verification image set into the initial model to adjust the hyperparameters and structure of the initial model. Test the initial model based on the target features corresponding to the target test image set until the initial model meets the preset convergence conditions and becomes the judgment model. Specifically, the initial model also has a foreign object judgment module, a type judgment module, and a location judgment module. Shape regularity, adhesion, gray-level mean difference, and gray-level contrast are used to iteratively train the weight matrix of the foreign object judgment module. Density uniformity, transmittance, airway mucosal redness and swelling rate, stenosis rate, stenosis uniformity, and stenosis segment edge regularity are used to iteratively train the weight matrix of the type judgment module, and target distance is used to train the location judgment module. In addition, the target verification image set is used to select the optimal model, and the target test image set is used to test the formed model. In one implementation, the convergence criteria are an accuracy greater than 98%, a recall greater than 97%, and a latency less than 0.3 seconds. The specific criteria can be set according to actual needs and are not limited here.

[0052] 207. Input the shape regularity, adhesion, grayscale mean difference, and grayscale contrast into the foreign object detection module to obtain the corresponding foreign object detection results; The shape regularity, adhesion, gray-level mean difference, and gray-level contrast are input into the foreign body detection module to obtain the corresponding foreign body detection result, which is either foreign body obstruction or pathological obstruction. Specifically, the weight matrix in the foreign body detection module includes multiple foreign body detection weight coefficients. These weight coefficients include at least a shape regularity threshold, an adhesion threshold, a gray-level mean difference threshold, and a gray-level contrast threshold. The foreign body detection result is obtained by calculating these four thresholds with the shape regularity, adhesion, gray-level mean difference, and gray-level contrast, thus determining whether it is foreign body obstruction or pathological obstruction. Specifically, in one implementation, after receiving the shape regularity, adhesion, gray-level mean difference, and gray-level contrast, the foreign body detection module determines whether the foreign body detection conditions are met. Specifically, the gray-level mean difference must be greater than a gray-level mean difference threshold, the gray-level contrast must be greater than a gray-level contrast threshold, the shape regularity must be greater than or equal to the shape regularity threshold, and the adhesion must be less than or equal to the adhesion threshold. If all four conditions are met simultaneously, the foreign body detection conditions are satisfied. If the foreign body detection conditions are satisfied, the obstruction is determined to be caused by a foreign body; otherwise, it is determined to be a pathological obstruction. Preferably, the shape regularity threshold, adhesion threshold, gray-level mean difference threshold, and gray-level contrast threshold can be 0.7, 0.3, 50, and 0.3, respectively, or other values ​​are not limited here. It is understood that other foreign body detection conditions and corresponding foreign body detection weight coefficients can also be set, which can be set according to actual needs and are not limited here.

[0053] 208. Extract density uniformity, transmittance, airway mucosal redness and swelling rate, stenosis rate, stenosis uniformity, and stenosis segment edge regularity from the segmentation mask image; Density uniformity, transmittance, airway mucosal redness and swelling rate, stenosis rate, stenosis uniformity, and stenosis segment edge regularity are extracted from the segmentation mask image. The subdivision steps are described in detail below.

[0054] a. Obtain the grayscale values ​​of all pixels in the obstructed region from the segmentation mask image, and determine the density uniformity by calculating the standard deviation of the grayscale values ​​based on the grayscale values ​​of all pixels in the obstructed region. Specifically, after converting the obstructed region of the segmentation mask image into a grayscale image, extract the grayscale value of each pixel, and then calculate the standard deviation to obtain the density uniformity. Density uniformity is used to describe the degree of density uniformity of the foreign object. The higher the density uniformity, the more likely the foreign object is to be a hard foreign object; the lower the density uniformity, the more likely the foreign object is to be a soft foreign object.

[0055] b. Obtain the grayscale values ​​of all pixels in the obstructed region based on the segmentation mask image, and determine the average grayscale value calculated based on the grayscale values ​​of all pixels in the obstructed region as the transmittance. Transmittance, for foreign objects, represents the degree to which the foreign object transmits light. A higher transmittance indicates that the foreign object is likely a soft foreign object, while a lower transmittance indicates that the foreign object is likely a hard foreign object.

[0056] c. Determine the area of ​​the obstruction region and the cross-sectional area of ​​the airway based on the segmentation mask image, and determine the ratio of the area of ​​the obstruction region to the cross-sectional area of ​​the airway as the airway mucosal redness and swelling rate. Specifically, the airway mucosal redness and swelling rate is relevant to pathological obstruction; a higher rate indicates a higher probability of laryngeal edema, while a lower rate indicates a lower probability of laryngeal edema.

[0057] d. Determine the minimum effective ventilation diameter and airway diameter of the airway stenosis segment based on the segmentation mask diagram, and define the ratio of the minimum effective ventilation diameter to the airway diameter as the stenosis rate. A higher stenosis rate indicates a greater likelihood of laryngeal edema, while a lower stenosis rate indicates a lower likelihood of laryngeal edema.

[0058] e. Determine the minimum diameter of all cross-sections within the airway stenosis segment based on the segmentation mask image, and determine the ratio of the standard deviation to the mean of the minimum diameter of all cross-sections within the airway stenosis segment as the diameter coefficient of variation; obtain the stenosis uniformity based on the diameter coefficient of variation, where the diameter coefficient of variation is inversely proportional to the stenosis uniformity. Simply put, stenosis uniformity is to see if the airway is narrowed uniformly; the greater the stenosis uniformity, the more likely it is to be caused by tumor compression in pathological obstruction.

[0059] f. Calculate the curvature of the narrowing contour points in the segmentation mask image, and determine the stenosis edge regularity as the ratio of the standard deviation to the mean of the curvature of the narrowing contour points in the airway region. A higher stenosis edge regularity indicates greater likelihood of tumor compression.

[0060] 209. Input the foreign body judgment result, density uniformity, light transmittance, airway mucosal redness and swelling rate, stenosis rate, stenosis uniformity, and stenosis segment edge regularity into the type judgment module to obtain the corresponding type judgment result; The foreign body identification result, density uniformity, transmittance, airway mucosal redness and swelling rate, stenosis rate, stenosis uniformity, and stenosis segment edge regularity are input into the type identification module to obtain the corresponding type identification result. The type identification result is either a hard foreign body, a soft foreign body, laryngeal edema, or tumor compression. The type identification module's weight matrix includes multiple type identification weight coefficients such as density uniformity threshold, transmittance threshold, airway mucosal redness and swelling rate threshold, stenosis rate threshold, stenosis uniformity threshold, and stenosis segment edge regularity threshold. These are calculated to obtain the type identification result. Specifically, in one embodiment, the type identification module determines the next step based on the foreign body identification result. If it is a foreign body obstruction, it determines whether it is a hard or soft obstruction. If the density uniformity is greater than or equal to the density uniformity threshold and the transmittance is less than or equal to the transmittance threshold, it is determined to be a hard obstruction. If the density uniformity is less than the density uniformity threshold and the transmittance is greater than the transmittance threshold, it is determined to be a soft obstruction. On the other hand, if it is a pathological obstruction, it is necessary to determine whether it is laryngeal edema or tumor compression. If the airway mucosal redness and swelling rate is greater than or equal to the airway mucosal redness and swelling rate threshold and the stenosis rate is greater than or equal to the stenosis rate threshold, it is determined to be laryngeal edema. If the stenosis uniformity is less than the stenosis uniformity threshold and the stenosis segment edge regularity is less than the stenosis segment edge regularity threshold, it is determined to be tumor compression. Preferably, the density uniformity threshold, transmittance threshold, airway mucosal redness and swelling rate threshold, stenosis rate threshold, stenosis uniformity threshold, and stenosis segment edge regularity threshold are 0.85, 0.3, 0.5, 0.4, 0.5, and 0.5, respectively. It is understood that these thresholds can be set according to actual needs, and no specific limitation is made here.

[0061] 210. Based on the calibration relationship between pixel coordinates and actual size, calculate the target distance between the obstruction region and the glottis region according to the segmentation mask image; Based on the calibration relationship between pixel coordinates and actual dimensions, the target distance between the obstruction region and the glottis region is calculated according to the segmentation mask image. Specifically, the distance on the segmentation mask image is first calculated based on the pixel coordinates, and then scaled up according to the calibration relationship to obtain the target distance.

[0062] 211. If the type judgment result is a hard foreign object, the target distance is input to the position judgment module to obtain the corresponding position judgment result.

[0063] If the type determination result is a hard foreign body, the target distance is input to the location determination module to obtain the corresponding location determination result. The location determination result includes high obstruction, mid-obstruction, or low obstruction. High obstruction indicates that the hard foreign body is 0-3cm away from the glottis, indicating that the hard foreign body is in the pharynx and near the glottis. Mid-obstruction indicates that the hard foreign body is 3-7cm away from the glottis, indicating that the hard foreign body is in the upper part of the upper trachea. Low obstruction indicates that the hard foreign body is 7-10cm away from the glottis, indicating that the hard foreign body is in the lower part of the upper trachea.

[0064] Ultimately, the judgment model will output the foreign object judgment result and the type judgment result, and, depending on the situation, the location judgment result, providing a basis for subsequent operations.

[0065] In this embodiment, after acquiring an image of the affected area using a laryngoscope or similar instrument, the device can rapidly process the image to obtain a foreign body identification result corresponding to the image. Because it utilizes a judgment model trained on numerous examples, the device can accurately determine whether the obstruction is due to a foreign body or a pathological obstruction, identifying the specific type among four possible types and its location. Compared to manual judgment, it provides a rapid, automated, and highly accurate result. Medical personnel can then perform appropriate actions based on the judgment results, resulting in a low operational threshold and significantly improving the success rate of emergency care. Furthermore, preprocessing greatly enhances the accuracy of the judgment results. The technology is highly scalable, allowing for the addition of new weighting coefficients and corresponding judgment conditions to the judgment model to identify new obstruction types, demonstrating long-term technological iteration potential.

[0066] The above describes a method for determining airway obstruction according to an embodiment of this application. The following describes a device for determining airway obstruction according to an embodiment of this application. Please refer to... Figure 3 One embodiment of an airway obstruction detection device in this application includes: The acquisition unit 301 is used to acquire an image of the affected area to be processed. The segmentation unit 302 is used to perform image segmentation processing on the image to be processed based on the pixel features of the pixels in the image to be processed, and to obtain a segmentation mask map including the obstruction region and the airway region. Extraction unit 303 is used to extract the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the gray-level mean difference between the obstruction region and the airway region, and the gray-level contrast of the obstruction region from the segmentation mask image. The acquisition unit 301 is also used to acquire the trained judgment model, which includes a foreign object judgment module; The processing unit 304 is used to input the shape regularity, adhesion, grayscale mean difference and grayscale contrast into the foreign body judgment module to obtain the corresponding foreign body judgment result, wherein the foreign body judgment result is foreign body obstruction or pathological obstruction.

[0067] In this embodiment, the acquisition unit 301 first acquires the image to be processed for the affected area. Then, the segmentation unit 302 performs image segmentation processing on the image to be processed based on the pixel features of the pixels in the image to be processed using an image segmentation algorithm, and obtains a segmentation mask image including the obstruction region and the airway region. Next, the extraction unit 303 extracts the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the gray-level mean difference between the obstruction region and the airway region, and the gray-level contrast of the obstruction region from the segmentation mask image. Then, the trained judgment model is acquired. Finally, the processing unit 304 inputs the shape regularity, adhesion, gray-level mean difference, and gray-level contrast into the foreign body judgment module of the judgment model to obtain the corresponding foreign body judgment result. The method of this application is applied to an airway obstruction detection device. After obtaining an image of the affected area through a laryngoscope or other means, the device can quickly process the image to obtain a foreign body detection result (foreign body obstruction or pathological obstruction) corresponding to the image. Because it uses a judgment model trained with a large number of examples, the device can accurately determine whether it is a foreign body obstruction or a pathological obstruction. Compared with manual judgment, it can quickly provide results in a short time with high accuracy, which greatly improves the success rate of emergency treatment.

[0068] The following is a detailed description of an airway obstruction detection device according to an embodiment of this application. Another embodiment of the airway obstruction detection device according to an embodiment of this application includes: The acquisition unit is used to acquire the image to be processed of the affected area. The segmentation unit is used to perform image segmentation processing on the image to be processed based on the pixel features of the pixels in the image to be processed, and to obtain a segmentation mask map including the obstruction region and the airway region. The extraction unit is used to extract the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the gray-level mean difference between the obstruction region and the airway region, and the gray-level contrast of the obstruction region from the segmented mask image. The acquisition unit is also used to acquire the trained judgment model, which includes a foreign object judgment module; The processing unit is used to input shape regularity, adhesion, grayscale mean difference and grayscale contrast into the foreign body judgment module to obtain the corresponding foreign body judgment result, wherein the foreign body judgment result is foreign body obstruction or pathological obstruction.

[0069] The extraction unit is specifically used for: Based on the segmentation mask image, calculate the area and perimeter of the obstruction region in the segmentation mask image, and substitute the area and perimeter into the preset roundness formula to calculate the roundness of the obstruction region. The curvature of the contour points of the obstruction region in the segmentation mask is calculated based on the segmentation mask image. The edge smoothness of the obstruction region is calculated based on the obstruction curvature variation coefficient and the preset smoothness coefficient, according to the preset smoothness formula. The obstruction curvature variation coefficient is the ratio of the standard deviation to the mean of the curvature of all contour points in the obstruction region. The shape regularity of the obstruction region is obtained by weighted fusion calculation based on roundness and edge smoothness. In the segmentation mask image, the area ratio of the contact area to the airway area is determined as the adhesion between the obstruction area and the airway area. The contact area is the area where every pixel in the obstruction area is a pixel in the airway area. The mean grayscale difference between the obstruction region and the airway region, and the grayscale contrast of the obstruction region are obtained from the segmentation mask image.

[0070] Segmentation units, specifically used for: Extract pixel features for each pixel from the image to be processed. The pixel features include grayscale features, color features, texture features, and semantic features. The grayscale features, color features, texture features, and semantic features corresponding to each pixel are concatenated and fused to obtain the fused features; For each pixel, the weighted sum of the fusion features and the preset fusion weights is determined as the pixel's probability vector, and the category label corresponding to the dimension with the highest probability in the probability vector is determined as the pixel's target category label, thereby obtaining a segmentation mask map corresponding to the image to be processed, including the obstruction region and the airway region.

[0071] The judgment model also includes a type judgment module, and the device also includes a first determining unit, used for: The gray values ​​of all pixels in the obstructed region are obtained from the segmentation mask image, and the standard deviation of the gray values ​​calculated from the gray values ​​of all pixels in the obstructed region is determined as the density uniformity. The gray values ​​of all pixels in the obstructed region are obtained from the segmentation mask image, and the average gray value calculated from the gray values ​​of all pixels in the obstructed region is determined as the transmittance. The area of ​​the obstruction region and the cross-sectional area of ​​the airway are determined based on the segmentation mask image, and the ratio of the area of ​​the obstruction region to the cross-sectional area of ​​the airway is determined as the airway mucosal redness rate. The minimum effective ventilation diameter and airway diameter of the airway stenosis segment are determined based on the segmentation mask diagram, and the ratio of the minimum effective ventilation diameter and airway diameter of the airway stenosis segment is determined as the stenosis rate. The minimum diameter of all cross sections within the airway stenosis is determined based on the segmentation mask image, and the ratio of the standard deviation to the mean of the minimum diameter of all cross sections within the airway stenosis is determined as the diameter variation coefficient; the stenosis uniformity is obtained based on the diameter variation coefficient, where the diameter variation coefficient is inversely proportional to the stenosis uniformity. The curvature of the narrow segment contour points in the airway region in the segmentation mask is calculated based on the segmentation mask image, and the ratio of the standard deviation to the mean of the curvature of the narrow segment contour points in the airway region is determined as the narrow segment edge regularity. The results of foreign body identification, density uniformity, light transmittance, airway mucosal redness and swelling rate, stenosis rate, stenosis uniformity, and stenosis segment edge regularity are input into the type identification module to obtain the corresponding type identification results, which are hard foreign bodies, soft foreign bodies, laryngeal edema, or tumor compression.

[0072] The judgment model also includes a position judgment module, and the segmentation mask image also includes a glottis region. The device also includes a second determination unit, used for: Based on the calibration relationship between pixel coordinates and actual size, the target distance between the obstruction region and the glottis region is calculated according to the segmentation mask image; If the type determination result is a hard foreign object, the target distance is input to the position determination module to obtain the corresponding position determination result, which includes high-level obstruction, mid-level obstruction, or low-level obstruction.

[0073] The device also includes a preprocessing unit for: The image to be processed is denoised using a Gaussian filtering algorithm, and the image after denoising is contrast-enhanced using a histogram equalization algorithm to obtain the first image. The first image is processed by an edge detection algorithm to obtain a second image that displays the outlines of different objects. The segmentation unit is specifically used for: Based on the image segmentation algorithm, the image to be processed is segmented according to the pixel features of the pixels in the second image, resulting in a segmentation mask image including the obstruction region and the airway region.

[0074] The device also includes a training unit for: Acquire a set of labeled training images, a set of labeled verification images, and a set of labeled test images according to a preset ratio; The training image set, the verification image set, and the test image set to be processed are subjected to denoising, contrast enhancement, and edge detection respectively to obtain the target training image set, the target verification image set, and the target test image set; The image segmentation algorithm based on semantic segmentation performs image segmentation processing on the target training image set, the target verification image set, and the target test image set respectively, and then performs feature extraction to obtain the target features corresponding to the target training image set, the target verification image set, and the target test image set respectively. Among them, the target features include the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the gray-level mean difference of the airway region, and the gray-level contrast of the obstruction region. The target features corresponding to the target training image set are input into the preset initial model to iteratively train the weight matrix of the initial model. The target features corresponding to the target verification image set are input into the initial model to adjust the hyperparameters and structure of the initial model. The initial model is then tested based on the target features corresponding to the target test image set until the initial model meets the preset convergence condition and becomes the judgment model.

[0075] The functions and procedures performed by each unit in the airway obstruction detection device of this embodiment are the same as those described above. Figures 1 to 2 The function and process of the middle airway obstruction detection device are similar, and will not be described in detail here.

[0076] Figure 4 This is a schematic diagram of an airway obstruction detection device provided in an embodiment of this application. The airway obstruction detection device 400 may include one or more central processing units (CPUs) 401 and a memory 405, in which one or more application programs or data are stored.

[0077] The memory 405 can be volatile or persistent storage. The program stored in the memory 405 can include one or more modules, each module including a series of instruction operations on the airway obstruction detection device 400. Furthermore, the central processing unit 401 can be configured to communicate with the memory 405 and execute the series of instruction operations in the memory 405 on the airway obstruction detection device 400.

[0078] The airway obstruction detection device 400 may also include one or more power supplies 402, one or more wired or wireless network interfaces 403, one or more input / output interfaces 404, and / or one or more operating systems, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0079] The central processing unit 401 can perform the aforementioned... Figures 1 to 2 The specific operations performed by the airway obstruction detection device in the illustrated embodiment will not be described in detail here.

[0080] This application also provides a computer-readable storage medium including instructions that, when executed on a computer, cause the computer to perform the methods described in the foregoing embodiments.

[0081] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0082] It should be noted that although the steps in the flowcharts of the various embodiments are drawn sequentially according to the arrows, unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the various embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps.

[0083] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the embodiments of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0084] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0085] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural transformations made using the description and drawings of this application under the inventive concept of this application, or direct / indirect applications in other related technical fields, are included within the patent protection scope of this application.

Claims

1. A method for determining airway obstruction, characterized in that, include: Acquire the image of the affected area to be processed; Based on the image segmentation algorithm, the image to be processed is segmented according to the pixel features of the pixels in the image to be processed, and a segmentation mask image including the obstruction region and the airway region is obtained. Extract the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the gray-scale mean difference between the obstruction region and the airway region, and the gray-scale contrast of the obstruction region from the segmentation mask image. Obtain a trained judgment model, the judgment model including a foreign object judgment module; The shape regularity, adhesion, grayscale mean difference, and grayscale contrast are input into the foreign body judgment module to obtain the corresponding foreign body judgment result, wherein the foreign body judgment result is foreign body obstruction or pathological obstruction.

2. The method for determining airway obstruction according to claim 1, characterized in that, The step of extracting the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the difference in the mean grayscale value between the obstruction region and the airway region, and the grayscale contrast of the obstruction region from the segmented mask image includes: Based on the segmentation mask image, calculate the area and perimeter of the obstruction region in the segmentation mask image, and substitute the area and perimeter into a preset roundness formula to calculate the roundness of the obstruction region. The curvature of the contour points of the obstruction region in the segmentation mask is calculated based on the segmentation mask image, and the edge smoothness of the obstruction region is calculated based on the obstruction curvature variation coefficient and the preset smoothness coefficient according to the preset smoothness formula. The obstruction curvature variation coefficient is the ratio of the standard deviation to the mean of the curvature of all contour points in the obstruction region. The shape regularity of the obstruction region is obtained by weighted fusion calculation based on the circularity and the edge smoothness. In the segmentation mask image, the area ratio of the contact area to the airway area is determined as the adhesion degree between the obstruction area and the airway area, wherein the contact area is the area where the neighboring pixels of each pixel in the obstruction area are pixels of the airway area. The mean grayscale difference between the obstruction region and the airway region, and the grayscale contrast of the obstruction region are obtained from the segmentation mask image.

3. The method for determining airway obstruction according to claim 1, characterized in that, The image segmentation algorithm performs image segmentation processing on the image to be processed based on the pixel features of the pixels in the image to be processed, and obtains a segmentation mask image including the obstruction region and the airway region, including: Pixel features of each pixel are extracted from the image to be processed, wherein the pixel features include grayscale features, color features, texture features and semantic features; The grayscale feature, color feature, texture feature, and semantic feature corresponding to each pixel are concatenated and fused to obtain the fused feature; For each pixel, the weighted sum of the fusion features and the preset fusion weights is determined as the probability vector of the pixel, and the category label corresponding to the dimension with the highest probability in the probability vector is determined as the target category label of the pixel, thereby obtaining the segmentation mask map corresponding to the image to be processed, including the obstruction region and the airway region.

4. The method for determining airway obstruction according to claim 1, characterized in that, The judgment model further includes a type judgment module. After inputting the shape regularity, adhesion, grayscale mean difference, and grayscale contrast into the foreign object judgment module to obtain the corresponding foreign object judgment result, the method further includes: The gray values ​​of all pixels in the obstruction region are obtained according to the segmentation mask image, and the standard deviation of the gray values ​​calculated based on the gray values ​​of all pixels in the obstruction region is determined as the density uniformity. The grayscale values ​​of all pixels in the obstruction region are obtained according to the segmentation mask image, and the average grayscale value calculated based on the grayscale values ​​of all pixels in the obstruction region is determined as the transmittance. The area of ​​the obstruction region and the cross-sectional area of ​​the airway are determined based on the segmentation mask image, and the ratio of the area of ​​the obstruction region to the cross-sectional area of ​​the airway is determined as the airway mucosal redness rate. The minimum effective ventilation diameter of the airway stenosis segment and the airway diameter are determined based on the segmented mask diagram, and the ratio of the minimum effective ventilation diameter of the airway stenosis segment to the airway diameter is determined as the stenosis rate. The minimum diameter of all cross sections within the airway stenosis is determined based on the segmentation mask image, and the ratio of the standard deviation to the mean of the minimum diameter of all cross sections within the airway stenosis is determined as the diameter variation coefficient; the stenosis uniformity is obtained based on the diameter variation coefficient, wherein the diameter variation coefficient is inversely proportional to the stenosis uniformity. The curvature of the narrow segment contour points of the airway region in the segmentation mask is calculated based on the segmentation mask, and the ratio of the standard deviation to the mean of the curvature of the narrow segment contour points of the airway region is determined as the narrow segment edge regularity. The foreign body judgment result, density uniformity, light transmittance, airway mucosal redness and swelling rate, stenosis rate, stenosis uniformity, and stenosis segment edge regularity are input into the type judgment module to obtain the corresponding type judgment result, which is a hard foreign body, a soft foreign body, laryngeal edema, or tumor compression.

5. The method for determining airway obstruction according to claim 4, characterized in that, The judgment model further includes a location judgment module, and the segmentation mask image further includes a glottic region. After inputting the foreign body judgment result, density uniformity, light transmittance, airway mucosal redness and swelling rate, stenosis rate, stenosis uniformity, and stenosis segment edge regularity into the type judgment module to obtain the corresponding type judgment result, the method further includes: Based on the calibration relationship between pixel coordinates and actual size, the target distance between the obstruction region and the glottic region is calculated according to the segmentation mask image; If the type determination result is a hard foreign object, the target distance is input to the position determination module to obtain the corresponding position determination result, which includes high-level obstruction, mid-level obstruction, or low-level obstruction.

6. The method for determining airway obstruction according to claim 1, characterized in that, After acquiring the image to be processed targeting the affected area, and before performing image segmentation processing on the image to be processed based on the pixel features of the pixels in the image to obtain a segmentation mask image including the obstruction region and the airway region, the method further includes: The image to be processed is subjected to noise removal based on the Gaussian filtering algorithm, and the image after noise removal is subjected to contrast enhancement based on the histogram equalization algorithm to obtain the first image. The first image is processed by an edge detection algorithm to obtain a second image that displays the contour boundaries of different objects. The image segmentation algorithm performs image segmentation processing on the image to be processed based on the pixel features of the pixels in the image to be processed, and obtains a segmentation mask image including the obstruction region and the airway region, including: Based on the image segmentation algorithm, the image to be processed is segmented according to the pixel features of the pixels in the second image to obtain a segmentation mask image including the obstruction region and the airway region.

7. The method for determining airway obstruction according to claim 1, characterized in that, Before obtaining the trained judgment model, the method further includes: Acquire a set of labeled training images, a set of labeled verification images, and a set of labeled test images according to a preset ratio; The training image set, the verification image set, and the test image set to be processed are respectively subjected to denoising, contrast enhancement, and edge detection processing to obtain the target training image set, the target verification image set, and the target test image set; The semantic segmentation-based image segmentation algorithm performs image segmentation processing on the target training image set, the target verification image set, and the target test image set respectively, and then performs feature extraction to obtain the target features corresponding to each of the target training image set, the target verification image set, and the target test image set. The target features include the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the gray-level mean difference of the airway region, and the gray-level contrast of the obstruction region. The target features corresponding to the target training image set are input into a preset initial model to iteratively train the weight matrix of the initial model. The target features corresponding to the target verification image set are input into the initial model to adjust the hyperparameters and structure of the initial model. The initial model is then tested based on the target features corresponding to the target test image set until the initial model meets the preset convergence condition and becomes the judgment model.

8. A device for determining airway obstruction, characterized in that, include: The acquisition unit is used to acquire the image to be processed of the affected area. The segmentation unit is used to perform image segmentation processing on the image to be processed based on the pixel features of the pixels in the image to be processed, based on the image segmentation algorithm, to obtain a segmentation mask image including the obstruction region and the airway region. The extraction unit is used to extract from the segmentation mask image the shape regularity of the obstruction region, the adhesion between the obstruction region and the airway region, the gray-level mean difference between the obstruction region and the airway region, and the gray-level contrast of the obstruction region. The acquisition unit is also used to acquire a trained judgment model, the judgment model including a foreign object judgment module; The processing unit is used to input the shape regularity, the adhesion, the grayscale mean difference and the grayscale contrast into the foreign body judgment module to obtain the corresponding foreign body judgment result, wherein the foreign body judgment result is foreign body obstruction or pathological obstruction.

9. A device for determining airway obstruction, characterized in that, include: Central processing unit, memory, and input / output interfaces; The memory is either a short-term storage memory or a persistent storage memory; The central processing unit is configured to communicate with the memory and execute instructions in the memory to perform the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Includes instructions that, when executed on a computer, cause the computer to perform the method as described in any one of claims 1 to 7.

Images